diff --git a/translations/ar/.co-op-translator.json b/translations/ar/.co-op-translator.json index d42fe3237..fa6108e7e 100644 --- a/translations/ar/.co-op-translator.json +++ b/translations/ar/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "ar" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:33:30+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:46:48+00:00", "source_file": "README.md", "language_code": "ar" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "ar" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:47:41+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "ar" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:29:36+00:00", diff --git a/translations/ar/README.md b/translations/ar/README.md index 1b8d77117..ee7a28945 100644 --- a/translations/ar/README.md +++ b/translations/ar/README.md @@ -1,74 +1,74 @@ -# كتاب طبخ Phi: أمثلة تطبيقية مع نماذج Phi من مايكروسوفت +# كتاب فاراي: أمثلة عملية مع نماذج Phi من مايكروسوفت [![افتح واستخدم العينات في GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![افتح في حاويات التطوير](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![مساهمو GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![قضايا GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![طلبات سحب GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![مشاكل GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![طلبات السحب في GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![مرحب بطلبات السحب](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![مراقبو GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![فروع GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![مشاهدو GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![تفريعات GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![نجوم GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) -[![مايكروسوفت فاوندري Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) +[![ديسكورد Microsoft Foundry](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi هي سلسلة من نماذج الذكاء الاصطناعي مفتوحة المصدر تم تطويرها من قبل مايكروسوفت. +Phi هي سلسلة من نماذج الذكاء الاصطناعي مفتوحة المصدر التي طورتها مايكروسوفت. -Phi حالياً هو أقوى وأكفأ نموذج لغة صغير (SLM)، مع مؤشرات أداء جيدة جداً في دعم لغات متعددة، والتفكير، وتوليد النصوص/الدردشة، والبرمجة، والصور، والصوت وسيناريوهات أخرى. +Phi هو حالياً أصغر نموذج لغة قوي وفعال من حيث التكلفة (SLM)، مع مؤشرات أداء جيدة جداً في تعدد اللغات، الاستدلال، إنشاء النصوص/الدردشة، الترميز، الصور، الصوت وسيناريوهات أخرى. -يمكنك نشر Phi على السحابة أو على أجهزة الطرف النهائي، ويمكنك بسهولة بناء تطبيقات ذكاء اصطناعي توليدية مع قدرة حسابية محدودة. +يمكنك نشر Phi على السحابة أو على أجهزة الحافة، ويمكنك بسهولة بناء تطبيقات الذكاء الاصطناعي التوليدية مع قدرة حسابية محدودة. -اتبع هذه الخطوات للبدء باستخدام هذه الموارد: -1. **افتح نسخة من المستودع**: انقر [![فروع GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **انسخ المستودع محليًا**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**انضم إلى مجتمع Microsoft AI على Discord وتعرف على الخبراء والمطورين الآخرين**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +اتبع هذه الخطوات لتبدأ باستخدام هذه الموارد: +1. **افتح مستودع النسخة**: اضغط [![تفريعات GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **انسخ المستودع**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**انضم إلى مجتمع مايكروسوفت AI على ديسكورد والتق بخبراء ومطوّرين آخرين**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/ar/cover.eb18d1b9605d754b.webp) ### 🌐 دعم متعدد اللغات -#### مدعوم عبر GitHub Action (آلي ومحدث دائماً) +#### مدعوم عبر GitHub Action (آلي ومُحدّث دائماً) -[العربية](./README.md) | [البنغالية](../bn/README.md) | [البلغارية](../bg/README.md) | [البورمية (ميانمار)](../my/README.md) | [الصينية (مبسطة)](../zh-CN/README.md) | [الصينية (تقليدية، هونغ كونغ)](../zh-HK/README.md) | [الصينية (تقليدية، ماكاو)](../zh-MO/README.md) | [الصينية (تقليدية، تايوان)](../zh-TW/README.md) | [الكرواتية](../hr/README.md) | [التشيكية](../cs/README.md) | [الدانماركية](../da/README.md) | [الهولندية](../nl/README.md) | [الإستونية](../et/README.md) | [الفنلندية](../fi/README.md) | [الفرنسية](../fr/README.md) | [الألمانية](../de/README.md) | [اليونانية](../el/README.md) | [العبرية](../he/README.md) | [الهندية](../hi/README.md) | [الهنغارية](../hu/README.md) | [الإندونيسية](../id/README.md) | [الإيطالية](../it/README.md) | [اليابانية](../ja/README.md) | [الكانادا](../kn/README.md) | [الخميرية](../km/README.md) | [الكورية](../ko/README.md) | [اللتوانية](../lt/README.md) | [الماليزية](../ms/README.md) | [المالايالامية](../ml/README.md) | [الماراثية](../mr/README.md) | [النيبالية](../ne/README.md) | [البيجدين النيجيرية](../pcm/README.md) | [النرويجية](../no/README.md) | [الفارسية (اللغة الفارسية)](../fa/README.md) | [البولندية](../pl/README.md) | [البرتغالية (البرازيل)](../pt-BR/README.md) | [البرتغالية (البرتغال)](../pt-PT/README.md) | [البنجابية (غرمخي)](../pa/README.md) | [الرومانية](../ro/README.md) | [الروسية](../ru/README.md) | [الصربية (السيريلية)](../sr/README.md) | [السلوفاكية](../sk/README.md) | [السلوفينية](../sl/README.md) | [الإسبانية](../es/README.md) | [السواحيلية](../sw/README.md) | [السويدية](../sv/README.md) | [التاغالوج (الفلبينية)](../tl/README.md) | [التاميلية](../ta/README.md) | [التيلوغو](../te/README.md) | [التايلاندية](../th/README.md) | [التركية](../tr/README.md) | [الأوكرانية](../uk/README.md) | [الأردية](../ur/README.md) | [الفيتنامية](../vi/README.md) +[العربية](./README.md) | [البنغالية](../bn/README.md) | [البلغارية](../bg/README.md) | [البورمية (ميانمار)](../my/README.md) | [الصينية (المبسطة)](../zh-CN/README.md) | [الصينية (التقليدية، هونغ كونغ)](../zh-HK/README.md) | [الصينية (التقليدية، ماكاو)](../zh-MO/README.md) | [الصينية (التقليدية، تايوان)](../zh-TW/README.md) | [الكرواتية](../hr/README.md) | [التشيكية](../cs/README.md) | [الدانمركية](../da/README.md) | [الهولندية](../nl/README.md) | [الإستونية](../et/README.md) | [الفنلندية](../fi/README.md) | [الفرنسية](../fr/README.md) | [الألمانية](../de/README.md) | [اليونانية](../el/README.md) | [العبرية](../he/README.md) | [الهندية](../hi/README.md) | [الهنغارية](../hu/README.md) | [الإندونيسية](../id/README.md) | [الإيطالية](../it/README.md) | [اليابانية](../ja/README.md) | [الكانتادية](../kn/README.md) | [الخمرية](../km/README.md) | [الكورية](../ko/README.md) | [الليتوانية](../lt/README.md) | [الملايوية](../ms/README.md) | [المالايالامية](../ml/README.md) | [الماراثية](../mr/README.md) | [النيبالية](../ne/README.md) | [النيجيرية بيدجين](../pcm/README.md) | [النرويجية](../no/README.md) | [الفارسية (الفرسية)](../fa/README.md) | [البولندية](../pl/README.md) | [البرتغالية (البرازيل)](../pt-BR/README.md) | [البرتغالية (البرتغال)](../pt-PT/README.md) | [البنجابية (غورموخي)](../pa/README.md) | [الرومانية](../ro/README.md) | [الروسية](../ru/README.md) | [الصربية (السيريلية)](../sr/README.md) | [السلوفاكية](../sk/README.md) | [السلوفينية](../sl/README.md) | [الإسبانية](../es/README.md) | [السواحيلية](../sw/README.md) | [السويدية](../sv/README.md) | [التاغالوغ (الفلبينية)](../tl/README.md) | [التاميلية](../ta/README.md) | [التيلجو](../te/README.md) | [التايلاندية](../th/README.md) | [التركية](../tr/README.md) | [الأوكرانية](../uk/README.md) | [الأردية](../ur/README.md) | [الفيتنامية](../vi/README.md) -> **هل تفضل النسخ محليًا؟** +> **هل تفضل النسخ محلياً؟** > -> يحتوي هذا المستودع على أكثر من 50 ترجمة للغات مختلفة مما يزيد بشكل كبير من حجم التنزيل. للنسخ بدون الترجمات، استخدم السحب الانتقائي: +> يشمل هذا المستودع أكثر من 50 ترجمة لغوية مما يزيد بشكل كبير من حجم التنزيل. للنسخ بدون الترجمات، استخدم الفحص المتفرق: > -> **باش / ماك / لينكس:** +> **باش / ماك أو إس / لينكس:** > ```bash > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` > -> **CMD (ويندوز):** +> **موجه الأوامر (ويندوز):** > ```cmd > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> هذا يمنحك كل ما تحتاجه لإتمام الدورة مع تنزيل أسرع بكثير. +> هذا يوفر لك كل ما تحتاجه لإكمال الدورة مع تحميل أسرع بكثير. ## جدول المحتويات -- المقدمة - - [مرحبًا بك في عائلة Phi](./md/01.Introduction/01/01.PhiFamily.md) +- مقدمة + - [مرحباً بكم في عائلة Phi](./md/01.Introduction/01/01.PhiFamily.md) - [إعداد بيئتك](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [فهم التقنيات الرئيسية](./md/01.Introduction/01/01.Understandingtech.md) + - [فهم التقنيات الأساسية](./md/01.Introduction/01/01.Understandingtech.md) - [أمان الذكاء الاصطناعي لنماذج Phi](./md/01.Introduction/01/01.AISafety.md) - - [دعم العتاد لنماذج Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [نماذج Phi وتوفرها عبر المنصات](./md/01.Introduction/01/01.Edgeandcloud.md) + - [دعم الأجهزة لنماذج Phi](./md/01.Introduction/01/01.Hardwaresupport.md) + - [نماذج Phi وتوافرها عبر المنصات](./md/01.Introduction/01/01.Edgeandcloud.md) - [استخدام Guidance-ai و Phi](./md/01.Introduction/01/01.Guidance.md) - [نماذج سوق GitHub](https://github.com/marketplace/models) - [كتالوج نماذج Azure AI](https://ai.azure.com) -- استدلال Phi في بيئات مختلفة +- الاستدلال باستخدام Phi في بيئات مختلفة - [Hugging face](./md/01.Introduction/02/01.HF.md) - [نماذج GitHub](./md/01.Introduction/02/02.GitHubModel.md) - [كتالوج نماذج Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) @@ -78,106 +78,108 @@ Phi حالياً هو أقوى وأكفأ نموذج لغة صغير (SLM)، م - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - استدلال عائلة Phi - - [استدلال Phi في iOS](./md/01.Introduction/03/iOS_Inference.md) - - [استدلال Phi في Android](./md/01.Introduction/03/Android_Inference.md) - - [استدلال Phi في Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [استدلال Phi في AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [استدلال Phi مع إطار عمل Apple MLX](./md/01.Introduction/03/MLX_Inference.md) - - [استدلال Phi في الخادم المحلي](./md/01.Introduction/03/Local_Server_Inference.md) - - [استدلال Phi في الخادم البعيد باستخدام أدوات AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [استدلال Phi باستخدام Rust](./md/01.Introduction/03/Rust_Inference.md) - - [استدلال Phi-Vision في المحلي](./md/01.Introduction/03/Vision_Inference.md) - - [استدلال Phi مع Kaito AKS، حاويات Azure (الدعم الرسمي)](./md/01.Introduction/03/Kaito_Inference.md) -- [كمية عائلة Phi](./md/01.Introduction/04/QuantifyingPhi.md) + - [الاستدلال باستخدام Phi في iOS](./md/01.Introduction/03/iOS_Inference.md) + - [الاستدلال باستخدام Phi في Android](./md/01.Introduction/03/Android_Inference.md) + - [الاستدلال باستخدام Phi في Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [الاستدلال باستخدام Phi في الحاسوب AI](./md/01.Introduction/03/AIPC_Inference.md) + - [الاستدلال باستخدام Phi مع إطار Apple MLX](./md/01.Introduction/03/MLX_Inference.md) + - [الاستدلال باستخدام Phi في الخادم المحلي](./md/01.Introduction/03/Local_Server_Inference.md) + - [الاستدلال باستخدام Phi في الخادم البعيد باستخدام AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [الاستدلال باستخدام Phi مع Rust](./md/01.Introduction/03/Rust_Inference.md) + - [الاستدلال برؤية Phi محلياً](./md/01.Introduction/03/Vision_Inference.md) + - [الاستدلال باستخدام Phi مع Kaito AKS، حاويات Azure (الدعم الرسمي)](./md/01.Introduction/03/Kaito_Inference.md) +- [تكميم عائلة Phi](./md/01.Introduction/04/QuantifyingPhi.md) - [تكميم Phi-3.5 / 4 باستخدام llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [تكميم Phi-3.5 / 4 باستخدام امتدادات Generative AI لـ onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [تكميم Phi-3.5 / 4 باستخدام امتدادات الذكاء الاصطناعي التوليدي لـ onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [تكميم Phi-3.5 / 4 باستخدام Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [تكميم Phi-3.5 / 4 باستخدام إطار عمل Apple MLX](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [تكميم Phi-3.5 / 4 باستخدام إطار Apple MLX](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - تقييم Phi - - [الذكاء الاصطناعي المسؤول](./md/01.Introduction/05/ResponsibleAI.md) - - [Microsoft Foundry للتقييم](./md/01.Introduction/05/AIFoundry.md) + - [ذكاء اصطناعي مسؤول](./md/01.Introduction/05/ResponsibleAI.md) + - [مايكروسوفت فوندري للتقييم](./md/01.Introduction/05/AIFoundry.md) - [استخدام Promptflow للتقييم](./md/01.Introduction/05/Promptflow.md) -- استرجاع المعرفة مع بحث Azure AI - - [كيفية استخدام Phi-4-mini و Phi-4-multimodal (RAG) مع بحث Azure AI](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- RAG مع Azure AI Search + - [كيفية استخدام Phi-4-mini و Phi-4-multimodal(RAG) مع Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG هجين محلي بدون سحابة مع SQLite FTS5 و phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - عينات تطوير تطبيقات Phi - تطبيقات النص والدردشة - - عينات Phi-4 - - [📓] [الدردشة مع نموذج Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [الدردشة مع نموذج Phi-4 المحلي ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [تطبيق الدردشة .NET Console مع Phi-4 ONNX باستخدام Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - عينات Phi-4 + - [📓] [دردشة مع نموذج Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [دردشة مع نموذج Phi-4 محلي ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [تطبيق دردشة .NET Console مع Phi-4 ONNX باستخدام السيمانتك كيرنل](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - عينات Phi-3 / 3.5 - - [روبوت دردشة محلي في المتصفح باستخدام Phi3، ONNX Runtime Web و WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [دردشة بوت محلي في المتصفح باستخدام Phi3، ONNX Runtime Web و WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [دردشة OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [النموذج المتعدد - Phi-3-mini التفاعلي و OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - بناء مغلف واستخدام Phi-3 مع MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [نموذج متعدد - Phi-3-mini تفاعلي و OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - بناء غلاف واستخدام Phi-3 مع MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [تحسين النموذج - كيفية تحسين نموذج Phi-3-min لـ ONNX Runtime Web باستخدام Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [تطبيق WinUI3 مع Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[مثال تطبيق ملاحظات AI متعدد النماذج مع WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [تحسين دقيق ودمج نماذج Phi-3 المخصصة مع Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [تحسين دقيق ودمج نماذج Phi-3 المخصصة مع Prompt flow في Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [تقييم نموذج Phi-3 / Phi-3.5 المحسن بدقة في Microsoft Foundry مع التركيز على مبادئ الذكاء الاصطناعي المسؤول من مايكروسوفت](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [نموذج توقع اللغة Phi-3.5-mini-instruct (صيني/إنجليزي)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [روبوت الدردشة Phi-3.5-Instruct WebGPU RAG](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [استخدام GPU ويندوز لإنشاء حل Prompt flow مع Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + -[مثال لتطبيق ملاحظات متعدد النماذج مدعوم بالذكاء الاصطناعي WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [ضبط دقيق ودمج نماذج Phi-3 مخصصة مع Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [ضبط دقيق ودمج نماذج Phi-3 مخصصة مع Prompt flow في Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [تقييم نموذج Phi-3 / Phi-3.5 المضبوط بدقة في Microsoft Foundry مع التركيز على مبادئ الذكاء الاصطناعي المسؤول من Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [مثال تنبؤ لغوي بـ Phi-3.5-mini-instruct (بالصينية/الإنجليزية)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [بوت الدردشة Phi-3.5-Instruct WebGPU RAG](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [استخدام GPU في ويندوز لإنشاء حل Prompt flow مع Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [استخدام Microsoft Phi-3.5 tflite لإنشاء تطبيق أندرويد](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [مثال سؤال وجواب .NET باستخدام نموذج ONNX محلي Phi-3 مع Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [تطبيق دردشة وحدة تحكم .NET مع Semantic Kernel و Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [مثال أسئلة وأجوبة .NET باستخدام نموذج ONNX Phi-3 المحلي بواسطة Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [تطبيق دردشة وحدة التحكم .NET مع Semantic Kernel و Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - عينات SDK استنتاج Azure AI المعتمدة على الكود - - عينات Phi-4 - - [📓] [إنشاء كود مشروع باستخدام Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - عينات SDK الاستدلال لـ Azure AI + - عينات Phi-4 + - [📓] [توليد رمز المشروع باستخدام Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - عينات Phi-3 / 3.5 - - [بناء مساعد دردشة GitHub في Visual Studio Code باستخدام عائلة Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [إنشاء وكيل دردشة مساعد في Visual Studio Code مع Phi-3.5 بواسطة نماذج GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [بناء دردشة GitHub Copilot الخاصة بك في Visual Studio Code مع عائلة Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [إنشاء وكيل دردشة GitHub Copilot خاص بك في Visual Studio Code مع Phi-3.5 بواسطة نماذج GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - عينات الاستدلال المتقدمة - - عينات Phi-4 + - عينات التفكير المتقدم + - عينات Phi-4 - [📓] [عينات Phi-4-mini-reasoning أو Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [تحسين دقيق لـ Phi-4-mini-reasoning مع Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [تحسين دقيق لـ Phi-4-mini-reasoning مع Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [ضبط دقيق لـ Phi-4-mini-reasoning مع Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [ضبط دقيق لـ Phi-4-mini-reasoning مع Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-reasoning مع نماذج GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-reasoning مع نماذج Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - العروض التوضيحية + - عروض توضيحية - [عروض Phi-4-mini مستضافة على Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - [عروض Phi-4-multimodal مستضافة على Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - عينات الرؤية - - عينات Phi-4 - - [📓] [استخدام Phi-4-multimodal لقراءة الصور وإنشاء الكود](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - عينات رؤية + - عينات Phi-4 + - [📓] [استخدام Phi-4-multimodal لقراءة الصور وتوليد الشفرة](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - عينات Phi-3 / 3.5 - - [📓][Phi-3-vision-نص الصورة إلى نص](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-تحويل نص الصورة إلى نص](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision تشفير CLIP](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [تجريبي: إعادة تدوير Phi-3](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - مساعد اللغة البصرية - مع Phi3-Vision و OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [📓][Phi-3-vision تضمين CLIP](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [عرض توضيحي: تدوير Phi-3](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - مساعد لغة بصري - مع Phi3-Vision و OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][عينة رؤية متعددة الإطارات أو متعددة الصور Phi-3.5 Vision](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [نموذج ONNX المحلي لرؤية Phi-3 باستخدام Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [نموذج ONNX المحلي لرؤية Phi-3 مع قائمة باستخدام Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][عينة رؤى متعددة الإطارات أو متعددة الصور بـ Phi-3.5 Vision](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [نموذج ONNX محلي لرؤية Phi-3 باستخدام Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [نموذج ONNX محلي لرؤية Phi-3 قائم على القائمة باستخدام Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - عينات الرؤية والاستدلال - - Phi-4-Reasoning-Vision-15B - - [📓] [استخدام Phi-4-Reasoning-Vision-15B للكشف عن عبور المارة بشكل غير قانوني](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - عينات التفكير والرؤية + - Phi-4-Reasoning-Vision-15B + - [📓] [استخدام Phi-4-Reasoning-Vision-15B لاكتشاف عبور المشاة غير القانوني](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [استخدام Phi-4-Reasoning-Vision-15B في الرياضيات](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [استخدام Phi-4-Reasoning-Vision-15B للكشف عن واجهة المستخدم](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [استخدام Phi-4-Reasoning-Vision-15B لاكتشاف واجهة المستخدم](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - عينات الرياضيات - - عينات Phi-4-Mini-Flash-Reasoning-Instruct [عرض توضيحي رياضي باستخدام Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - عينات Phi-4-Mini-Flash-Reasoning-Instruct [عرض تجريبي للرياضيات مع Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - عينات الصوت - - عينات Phi-4 + - عينات Phi-4 - [📓] [استخراج نصوص الصوت باستخدام Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [عينة صوت Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [عينة ترجمة الكلام Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [تطبيق وحدة تحكم .NET باستخدام Phi-4-multimodal لتحليل ملف صوتي وإنشاء النص](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [تطبيق وحدة التحكم .NET باستخدام Phi-4-multimodal لتحليل ملف صوتي وتوليد نص](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - عينات MOE - عينات Phi-3 / 3.5 - - [📓] [نماذج مزيج الخبراء Phi-3.5 (MoEs) لعينة وسائل التواصل الاجتماعي](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [بناء خط أنابيب توليد معزّز بالاستدعاء (RAG) مع NVIDIA NIM Phi-3 MOE، Azure AI Search، و LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [نماذج خليط الخبراء Phi-3.5 (MoEs) عينة وسائل التواصل الاجتماعي](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [بناء خط إنتاج التوليد المعزز بالاستخراج (RAG) باستخدام NVIDIA NIM Phi-3 MOE، Azure AI Search، و LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - عينات استدعاء الوظائف - عينات Phi-4 🆕 @@ -185,69 +187,70 @@ Phi حالياً هو أقوى وأكفأ نموذج لغة صغير (SLM)، م - [📓] [استخدام استدعاء الوظائف لإنشاء وكلاء متعددين مع Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [استخدام استدعاء الوظائف مع Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [استخدام استدعاء الوظائف مع ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - عينات المزج المتعدد الوسائط + - عينات الخلط المتعدد الأنماط - عينات Phi-4 🆕 - [📓] [استخدام Phi-4-multimodal كصحفي تقني](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [تطبيق وحدة تحكم .NET باستخدام Phi-4-multimodal لتحليل الصور](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- عينات التحسين الدقيق Phi - - [سيناريوهات التحسين الدقيق](./md/03.FineTuning/FineTuning_Scenarios.md) - - [التحسين الدقيق مقابل RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [دع Phi-3 يصبح خبيرًا صناعيًا - تحسين دقيق](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [التحسين الدقيق لـ Phi-3 باستخدام AI Toolkit لـ VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [التحسين الدقيق لـ Phi-3 باستخدام خدمة تعلم الآلة Azure](./md/03.FineTuning/Introduce_AzureML.md) - - [التحسين الدقيق لـ Phi-3 باستخدام Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [التحسين الدقيق لـ Phi-3 باستخدام QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [التحسين الدقيق لـ Phi-3 باستخدام Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [التحسين الدقيق لـ Phi-3 باستخدام Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [التحسين الدقيق باستخدام Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [مختبر عملي للتحسين الدقيق باستخدام Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [التحسين الدقيق لـ Phi-3-vision باستخدام Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [التحسين الدقيق لـ Phi-3 باستخدام إطار عمل Apple MLX](./md/03.FineTuning/FineTuning_MLX.md) - - [التحسين الدقيق لـ Phi-3-vision (الدعم الرسمي)](./md/03.FineTuning/FineTuning_Vision.md) - - [ضبط دقيق لـ Phi-3 باستخدام Kaito AKS وحاويات Azure (الدعم الرسمي)](./md/03.FineTuning/FineTuning_Kaito.md) - - [ضبط دقيق لـ Phi-3 و Phi-3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + - [تطبيق وحدة التحكم .NET باستخدام Phi-4-multimodal لتحليل الصور](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- ضبط دقيق لعينة Phi + - [سيناريوهات الضبط الدقيق](./md/03.FineTuning/FineTuning_Scenarios.md) + - [الضبط الدقيق مقابل RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [اضبط Phi-3 ليصبح خبيرًا في الصناعة](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [ضبط دقيق لـ Phi-3 باستخدام مجموعة أدوات AI لـ VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [ضبط دقيق لـ Phi-3 باستخدام خدمة تعلم الآلة Azure](./md/03.FineTuning/Introduce_AzureML.md) + - [ضبط دقيق لـ Phi-3 مع Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [ضبط دقيق لـ Phi-3 مع QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [ضبط دقيق لـ Phi-3 مع Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [ضبط دقيق لـ Phi-3 باستخدام Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [الضبط الدقيق مع Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [الضبط الدقيق مع مختبر عملي Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) + - [ضبط دقيق لرؤية Phi-3 باستخدام Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [تخصيص Phi-3 باستخدام إطار عمل Apple MLX](./md/03.FineTuning/FineTuning_MLX.md) + - [تخصيص Phi-3-vision (الدعم الرسمي)](./md/03.FineTuning/FineTuning_Vision.md) + - [تخصيص Phi-3 باستخدام Kaito AKS ، حاويات Azure (الدعم الرسمي)](./md/03.FineTuning/FineTuning_Kaito.md) + - [تخصيص Phi-3 و 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - مختبر عملي - - [استكشاف أحدث النماذج المتقدمة: LLMs، SLMs، التطوير المحلي والمزيد](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [فتح إمكانيات معالجة اللغة الطبيعية: الضبط الدقيق باستخدام Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - -- الأبحاث الأكاديمية والأوراق المنشورة - - [الكتب الدراسية هي كل ما تحتاجه II: تقرير فني عن phi-1.5](https://arxiv.org/abs/2309.05463) - - [تقرير فني عن Phi-3: نموذج لغة عالي القدرة يعمل محليًا على هاتفك](https://arxiv.org/abs/2404.14219) - - [تقرير فني عن Phi-4](https://arxiv.org/abs/2412.08905) - - [تقرير فني عن Phi-4-Mini: نماذج لغوية متعددة الوسائط مدمجة لكنها قوية عبر مزيج LoRAs](https://arxiv.org/abs/2503.01743) - - [تحسين نماذج اللغة الصغيرة لاستدعاء الوظائف داخل المركبة](https://arxiv.org/abs/2501.02342) - - [(لماذا PHI) ضبط دقيق لـ PHI-3 للإجابة على أسئلة اختيار من متعدد: المنهجية، النتائج، والتحديات](https://arxiv.org/abs/2501.01588) - - [تقرير فني عن استدلال Phi-4](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [تقرير فني عن استدلال Phi-4-mini](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [استكشاف النماذج المتطورة: LLMs، SLMs، التطوير المحلي والمزيد](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [إطلاق إمكانيات معالجة اللغة الطبيعية: التخصيص باستخدام Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + +- أوراق وأبحاث أكاديمية ومنشورات + - [الكتب الدراسية هي كل ما تحتاجه II: تقرير فني phi-1.5](https://arxiv.org/abs/2309.05463) + - [تقرير فني Phi-3: نموذج لغة عالي الكفاءة محليًا على هاتفك](https://arxiv.org/abs/2404.14219) + - [تقرير فني Phi-4](https://arxiv.org/abs/2412.08905) + - [تقرير فني Phi-4-Mini: نماذج لغوية متعددة الوسائط مدمجة وقوية من خلال خليط من LoRAs](https://arxiv.org/abs/2503.01743) + - [تحسين النماذج اللغوية الصغيرة لاستدعاء الوظائف في المركبات](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) تخصيص PHI-3 للإجابة على أسئلة الاختيار من متعدد: المنهجية، النتائج، والتحديات](https://arxiv.org/abs/2501.01588) + - [تقرير فني Phi-4 للتفكير](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [تقرير فني Phi-4-mini للتفكير](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## استخدام نماذج Phi ### Phi على Microsoft Foundry -يمكنك تعلم كيفية استخدام Microsoft Phi وكيفية بناء حلول شاملة في أجهزتك المختلفة. لتجربة Phi بنفسك، ابدأ بالتعامل مع النماذج وتخصيص Phi لسيناريوهاتك باستخدام [كتالوج نماذج Azure AI من Microsoft Foundry](https://aka.ms/phi3-azure-ai) ويمكنك معرفة المزيد في البدء مع [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +يمكنك معرفة كيفية استخدام Microsoft Phi وكيفية بناء حلول شاملة في أجهزتك المختلفة. لتجربة Phi بنفسك، ابدأ بالتفاعل مع النماذج وتخصيص Phi لسيناريوهاتك باستخدام [كتالوج نماذج Microsoft Foundry Azure AI](https://aka.ms/phi3-azure-ai) يمكنك معرفة المزيد في البدء مع [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**ساحة اللعب** -لكل نموذج ساحة لعب مخصصة لاختبار النموذج على [ساحة لعب Azure AI](https://aka.ms/try-phi3). +**ميدان اللعب** +لكل نموذج ميدان لعب مخصص لاختبار النموذج [Azure AI Playground](https://aka.ms/try-phi3). ### Phi على نماذج GitHub -يمكنك تعلم كيفية استخدام Microsoft Phi وكيفية بناء حلول شاملة في أجهزتك المختلفة. لتجربة Phi بنفسك، ابدأ بالتعامل مع النموذج وتخصيص Phi لسيناريوهاتك باستخدام [كتالوج نماذج GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ويمكنك معرفة المزيد في البدء مع [كتالوج نماذج GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) +يمكنك تعلم كيفية استخدام Microsoft Phi وكيفية بناء حلول شاملة في أجهزتك المختلفة. لتجربة Phi بنفسك، ابدأ باللعب مع النموذج وتخصيص Phi لسيناريوهاتك باستخدام [كتالوج نماذج GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) يمكنك معرفة المزيد في البدء مع [كتالوج نماذج GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) -**ساحة اللعب** -لكل نموذج ساحة لعب مخصصة لاختبار النموذج [ساحة اللعب]( /md/02.QuickStart/GitHubModel_QuickStart.md). +**ميدان اللعب** +لكل نموذج [ميدان لعب مخصص لاختبار النموذج](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi على Hugging Face يمكنك أيضًا العثور على النموذج على [Hugging Face](https://huggingface.co/microsoft) -**ساحة اللعب** - [ساحة لعب Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**ميدان اللعب** + [ميدان لعب Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 دورات أخرى + ## 🎒 دورات أخرى -فريقنا ينتج دورات أخرى! تفقد: +يقوم فريقنا بإنتاج دورات أخرى! اطلع على: ### LangChain @@ -273,42 +276,44 @@ Phi حالياً هو أقوى وأكفأ نموذج لغة صغير (SLM)، م --- ### التعلم الأساسي -[![تعلم الآلة للمبتدئين](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![التعلم الآلي للمبتدئين](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![علوم البيانات للمبتدئين](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![الذكاء الاصطناعي للمبتدئين](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![الأمن السيبراني للمبتدئين](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![تطوير الويب للمبتدئين](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![إنترنت الأشياء للمبتدئين](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![انترنت الأشياء للمبتدئين](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![تطوير XR للمبتدئين](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### سلسلة المساعد الذكي -[![المساعد الذكي للبرمجة الزوجية بالذكاء الاصطناعي](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![المساعد الذكي لـ C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![مغامرة المساعد الذكي](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +### سلسلة Copilot +[![Copilot للبرمجة المصاحبة بالذكاء الاصطناعي](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot لـ C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![مغامرة Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## الذكاء الاصطناعي المسؤول +## الذكاء الاصطناعي المسؤول + +تلتزم مايكروسوفت بمساعدة عملائنا على استخدام منتجات الذكاء الاصطناعي لدينا بمسؤولية، ومشاركة خبراتنا، وبناء شراكات قائمة على الثقة من خلال أدوات مثل ملاحظات الشفافية وتقييمات التأثير. يمكن العثور على العديد من هذه الموارد في [https://aka.ms/RAI](https://aka.ms/RAI). +يعتمد نهج مايكروسوفت تجاه الذكاء الاصطناعي المسؤول على مبادئ الذكاء الاصطناعي لدينا المتعلقة بالإنصاف، والموثوقية والسلامة، والخصوصية والأمان، والشمولية، والشفافية، والمسائلة. + +يمكن للنماذج الطبيعية واسعة النطاق للغة والصورة والصوت - مثل تلك المستخدمة في هذا المثال - أن تتصرف ربما بطرق غير عادلة أو غير موثوقة أو مسيئة، مما قد يسبب أضرارًا. يرجى الاطلاع على [ملاحظة الشفافية لخدمة Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) للتعرف على المخاطر والقيود. -تلتزم مايكروسوفت بمساعدة عملائنا على استخدام منتجات الذكاء الاصطناعي الخاصة بنا بمسؤولية، ومشاركة ما تعلمناه، وبناء شراكات مبنية على الثقة من خلال أدوات مثل ملاحظات الشفافية وتقييمات التأثير. يمكن العثور على العديد من هذه الموارد في [https://aka.ms/RAI](https://aka.ms/RAI). -تعتمد منهجية مايكروسوفت تجاه الذكاء الاصطناعي المسؤول على مبادئ الذكاء الاصطناعي لدينا التي تشمل العدالة، والموثوقية والسلامة، والخصوصية والأمان، والشمولية، والشفافية، والمساءلة. -يمكن لنماذج اللغة الطبيعية واسعة النطاق، والنماذج القائمة على الصور والصوت - مثل تلك المستخدمة في هذا المثال - أن تتصرف بطرق قد تكون غير عادلة أو غير موثوقة أو مسيئة، مما قد يسبب أضرارًا. يرجى مراجعة [مذكرة الشفافية لخدمة Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) للاطلاع على المخاطر والقيود. -النهج الموصى به للتخفيف من هذه المخاطر هو تضمين نظام أمان في هندستك المعمارية يمكنه الكشف عن السلوك الضار ومنعه. يوفر [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) طبقة حماية مستقلة، قادرة على الكشف عن المحتوى الضار الذي ينشئه المستخدم أو الذكاء الاصطناعي في التطبيقات والخدمات. يتضمن Azure AI Content Safety واجهات برمجة تطبيقات للنصوص والصور تتيح لك الكشف عن المواد الضارة. ضمن Microsoft Foundry، تتيح لك خدمة Content Safety عرض واستكشاف وتجربة أمثلة التعليمات البرمجية للكشف عن المحتوى الضار عبر أنماط مختلفة. يوجهك [وثائق البداية السريعة التالية](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) لكيفية إجراء الطلبات إلى الخدمة. +النهج الموصى به للتقليل من هذه المخاطر هو تضمين نظام أمان في البنية التحتية الخاصة بك يمكنه اكتشاف ومنع السلوك الضار. يوفر [أمان المحتوى في Azure AI](https://learn.microsoft.com/azure/ai-services/content-safety/overview) طبقة حماية مستقلة، قادرة على اكتشاف المحتوى الضار الذي يُنتجه المستخدمون أو الذكاء الاصطناعي في التطبيقات والخدمات. يشمل أمان المحتوى في Azure AI واجهات برمجة التطبيقات للنصوص والصور التي تتيح لك اكتشاف المواد الضارة. داخل Microsoft Foundry، تتيح لك خدمة أمان المحتوى عرض واستكشاف وتجربة أكواد تجريبية لاكتشاف المحتوى الضار عبر أنماط متعددة. الوثائق [التمهيدية السريعة](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) التالية ترشدك خلال كيفية إرسال الطلبات إلى الخدمة. -جانب آخر يجب أخذه في الاعتبار هو الأداء العام للتطبيق. مع التطبيقات متعددة الأنماط والنماذج، نعتبر الأداء يعني أن النظام يعمل كما تتوقع أنت ومستخدموك، بما في ذلك عدم توليد مخرجات ضارة. من المهم تقييم أداء تطبيقك العام باستخدام [مقاييس التقييم للأداء والجودة والمخاطر والسلامة](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). لديك أيضًا القدرة على إنشاء وتقييم باستخدام [مقاييس تقييم مخصصة](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +جانب آخر يجب أخذه في الاعتبار هو الأداء العام للتطبيق. بالنسبة للتطبيقات متعددة الأنماط والنماذج، نعتبر أن الأداء يعني أن النظام يعمل كما تتوقع أنت ومستخدموك، بما في ذلك عدم توليد مخرجات ضارة. من المهم تقييم أداء تطبيقك العام باستخدام [مقاييس الأداء والجودة والمخاطر والسلامة](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). لديك أيضًا القدرة على الإنشاء والتقييم باستخدام [مقاييس مخصصة](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -يمكنك تقييم تطبيق الذكاء الاصطناعي الخاص بك في بيئة التطوير باستخدام [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). بالنظر إلى مجموعة بيانات اختبار أو هدف، يتم قياس أجيال تطبيقك التوليدي للذكاء الاصطناعي كميةً باستخدام مقاييس مدمجة أو مقاييس مخصصة من اختيارك. للبدء مع azure ai evaluation sdk لتقييم نظامك، يمكنك اتباع [دليل البداية السريعة](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). بمجرد تنفيذ تشغيل التقييم، يمكنك [عرض النتائج في Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +يمكنك تقييم تطبيق الذكاء الاصطناعي الخاص بك في بيئة التطوير باستخدام [مجموعة أدوات تقييم Azure AI](https://microsoft.github.io/promptflow/index.html). استنادًا إلى مجموعة بيانات اختبار أو هدف، يتم قياس توليدات تطبيق الذكاء الاصطناعي التوليدي لديك بشكل كمي باستخدام المقاييس المدمجة أو المقاييس المخصصة التي تختارها. لبدء استخدام مجموعة أدوات تقييم Azure AI لتقييم نظامك، يمكنك اتباع [دليل البدء السريع](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). بمجرد تنفيذ تشغيل التقييم، يمكنك [عرض النتائج في Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## العلامات التجارية -قد يحتوي هذا المشروع على علامات تجارية أو شعارات لمشاريع أو منتجات أو خدمات. يخضع الاستخدام المصرح به لعلامات Microsoft التجارية أو شعاراتها ويجب أن يتبع [إرشادات العلامات التجارية والعلامات التجارية الخاصة بـ Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -يجب ألا يسبب استخدام علامات Microsoft التجارية أو شعاراتها في نسخ معدلة من هذا المشروع أي لبس أو يوحي برعاية من Microsoft. وأي استخدام لعلامات تجارية أو شعارات لأطراف ثالثة يخضع لسياسات تلك الأطراف. +قد يحتوي هذا المشروع على علامات تجارية أو شعارات لمشاريع أو منتجات أو خدمات. الاستخدام المصرح به لعلامات أو شعارات Microsoft يجب أن يتبع ويخضع لـ [إرشادات العلامة التجارية والهوية التجارية لشركة Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +يجب ألا يتسبب استخدام علامات أو شعارات Microsoft في نسخ معدلة من هذا المشروع في حدوث ارتباك أو يوحي برعاية Microsoft. أي استخدام لعلامات أو شعارات تتبع طرفًا ثالثًا يخضع لسياسات ذلك الطرف الثالث. ## الحصول على المساعدة -إذا علقت أو كان لديك أي أسئلة حول بناء تطبيقات الذكاء الاصطناعي، انضم إلى: +إذا واجهت صعوبة أو كان لديك أي أسئلة حول بناء تطبيقات الذكاء الاصطناعي، انضم إلى: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) @@ -319,6 +324,6 @@ Phi حالياً هو أقوى وأكفأ نموذج لغة صغير (SLM)، م --- -**تنويه**: -تمت ترجمة هذا المستند باستخدام خدمة الترجمة الآلية [Co-op Translator](https://github.com/Azure/co-op-translator). بينما نسعى للجودة والدقة، يرجى ملاحظة أن الترجمات الآلية قد تحتوي على أخطاء أو عدم دقة. يجب اعتبار المستند الأصلي بلغته الأصلية المصدر الموثوق به. للمعلومات الهامة أو الحساسة، يُنصح بالترجمة المحترفة بواسطة مترجم بشري. نحن غير مسؤولين عن أي سوء فهم أو تفسير خاطئ قد ينشأ عن استخدام هذه الترجمة. +**تنويه**: +تمت ترجمة هذا المستند باستخدام خدمة الترجمة بالذكاء الاصطناعي [Co-op Translator](https://github.com/Azure/co-op-translator). بينما نسعى للدقة، يرجى العلم أن الترجمات الآلية قد تحتوي على أخطاء أو عدم دقة. يجب اعتبار المستند الأصلي بلغته الأصلية المصدر الرسمي والمعتمد. للمعلومات الهامة، يُنصح بالاستعانة بترجمة بشرية محترفة. نحن غير مسؤولين عن أي سوء فهم أو تفسير ناتج عن استخدام هذه الترجمة. \ No newline at end of file diff --git a/translations/ar/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/ar/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..9a2433375 --- /dev/null +++ b/translations/ar/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 استرجاع هجين مع Microsoft phi-4-mini و SQLite FTS5 (نماذج لغة صغيرة بدون سحابة)\n", + "\n", + "> **المؤلف:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **التركيز:** نماذج اللغة الصغيرة (SLMs)، SQLite FTS5 BM25، التضمينات الكثيفة، دمج الترتيب العكسي (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. الدافع: معضلة استرجاع الكلمات المفتاحية في نماذج اللغة الصغيرة المحلية\n", + "غالبًا ما تفشل هندسات RAG القياسية التي تعتمد فقط على التضمينات المتجهية الكثيفة في استرجاع الرموز العددية الدقيقة (مثلاً، `2,340,000 TL`، رموز العقود، أرقام الحسابات). \n", + "بالمقابل، تفوت عمليات البحث اللفظي المتناثر (BM25) المترادفات الدلالية والأسئلة المعاد صياغتها.\n", + "\n", + "توضح هذه الوصفة كيفية تنفيذ **محرك استرجاع هجين عالي السرعة داخل الذاكرة** يجمع بين:\n", + "1. **المتجهات الكثيفة** (تشابه جيب التمام)\n", + "2. **البحث اللفظي المتناثر** (SQLite FTS5 BM25)\n", + "3. **دمج الترتيب العكسي (RRF, $k=60$)**\n", + "4. **توليد الاقتباسات المبنية (`[1]`, `[2]`)** مع Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. مخطط SQLite مزدوج (متجهات كثيفة + جدول FTS5 افتراضي بتصنيف BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. قياس أداء استيعاب وتنفيذ العينة\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. صياغة مطالبات مرتكزة لنموذج Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**تنويه**:\nتمت ترجمة هذا المستند باستخدام خدمة الترجمة بالذكاء الاصطناعي [Co-op Translator](https://github.com/Azure/co-op-translator). بينما نسعى للدقة، يرجى العلم أن الترجمات الآلية قد تحتوي على أخطاء أو عدم دقة. يجب اعتبار المستند الأصلي بلغته الأصلية المصدر الرسمي والمعتمد. للمعلومات الهامة، يُنصح بالاستعانة بترجمة بشرية محترفة. نحن غير مسؤولين عن أي سوء فهم أو تفسير ناتج عن استخدام هذه الترجمة.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/bg/.co-op-translator.json b/translations/bg/.co-op-translator.json index 2cfd7a654..eeb2c0a18 100644 --- a/translations/bg/.co-op-translator.json +++ b/translations/bg/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "bg" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T23:51:01+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:51:42+00:00", "source_file": "README.md", "language_code": "bg" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "bg" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:52:44+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "bg" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:32:11+00:00", diff --git a/translations/bg/README.md b/translations/bg/README.md index fb5bae4b3..f670d7989 100644 --- a/translations/bg/README.md +++ b/translations/bg/README.md @@ -1,42 +1,42 @@ -# Рецептурник Phi: Практически примери с Phi моделите на Microsoft +# Книга с рецепти за Phi: Практически примери с Phi моделите на Microsoft [![Отворете и използвайте примерите в GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Отваряне в Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Отвори в Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![Приносители в GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![Проблеми в GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Молби за промени в GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![Добре дошли PR-и](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Поискай промени в GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Добре дошли](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![Наблюдатели в GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![Вилици в GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![Форкове в GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![Звезди в GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) Phi е серия от отворени AI модели, разработени от Microsoft. -В момента Phi е най-мощният и икономичен малък езиков модел (SLM), с много добри показатели в многоезични, логически, текстови/чат генериращи, кодиращи, изображения, аудио и други сценарии. +Phi е в момента най-мощният и ефективен по отношение на разходите малък езиков модел (SLM), с отлични резултати в многоезични, логически, текстови/чат генерации, кодиране, изображения, аудио и други сценарии. -Можете да разположите Phi в облака или на крайни устройства, и лесно да изграждате генеративни AI приложения с ограничени изчислителни ресурси. +Можете да разгърнете Phi в облак или на крайни устройства и лесно да изграждате генеративни AI приложения с ограничена изчислителна мощ. Следвайте тези стъпки, за да започнете да използвате тези ресурси: -1. **Форкирайте хранилището**: Натиснете [![Вилици в GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +1. **Форкнете хранилището**: Кликнете [![Форкове в GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Клонирайте хранилището**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Присъединете се към Discord общността на Microsoft AI и срещнете експерти и други разработчици**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Присъединете се към Microsoft AI Discord Общността и се срещнете с експерти и други разработчици**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/bg/cover.eb18d1b9605d754b.webp) -### 🌐 Многоезична поддръжка +### 🌐 Поддръжка на много езици -#### Поддържана чрез GitHub Action (Автоматизирано и винаги актуално) +#### Поддържано чрез GitHub Action (Автоматично и винаги актуално) -[Арабски](../ar/README.md) | [Бенгалски](../bn/README.md) | [Български](./README.md) | [Бирмански (Мианмар)](../my/README.md) | [Китайски (опростен)](../zh-CN/README.md) | [Китайски (традиционен, Хонг Конг)](../zh-HK/README.md) | [Китайски (традиционен, Макао)](../zh-MO/README.md) | [Китайски (традиционен, Тайван)](../zh-TW/README.md) | [Хърватски](../hr/README.md) | [Чешки](../cs/README.md) | [Датски](../da/README.md) | [Холандски](../nl/README.md) | [Естонски](../et/README.md) | [Фински](../fi/README.md) | [Френски](../fr/README.md) | [Немски](../de/README.md) | [Гръцки](../el/README.md) | [Иврит](../he/README.md) | [Хинди](../hi/README.md) | [Унгарски](../hu/README.md) | [Индонезийски](../id/README.md) | [Италиански](../it/README.md) | [Японски](../ja/README.md) | [Каннада](../kn/README.md) | [Кхмер](../km/README.md) | [Корейски](../ko/README.md) | [Литовски](../lt/README.md) | [Малайски](../ms/README.md) | [Малаялам](../ml/README.md) | [Маратхи](../mr/README.md) | [Непалски](../ne/README.md) | [Нигерийски пиджин](../pcm/README.md) | [Норвежки](../no/README.md) | [Персийски (фарси)](../fa/README.md) | [Полски](../pl/README.md) | [Португалски (Бразилия)](../pt-BR/README.md) | [Португалски (Португалия)](../pt-PT/README.md) | [Пенджабски (гурмукхи)](../pa/README.md) | [Румънски](../ro/README.md) | [Руски](../ru/README.md) | [Сръбски (кирилица)](../sr/README.md) | [Словашки](../sk/README.md) | [Словенски](../sl/README.md) | [Испански](../es/README.md) | [Суахили](../sw/README.md) | [Шведски](../sv/README.md) | [Тагалог (Филипински)](../tl/README.md) | [Тамилски](../ta/README.md) | [Телугу](../te/README.md) | [Тайски](../th/README.md) | [Турски](../tr/README.md) | [Украински](../uk/README.md) | [Урду](../ur/README.md) | [Виетнамски](../vi/README.md) +[Арабски](../ar/README.md) | [Бенгалски](../bn/README.md) | [Български](./README.md) | [Бирмански (Мианмар)](../my/README.md) | [Китайски (опростен)](../zh-CN/README.md) | [Китайски (традиционен, Хонконг)](../zh-HK/README.md) | [Китайски (традиционен, Макао)](../zh-MO/README.md) | [Китайски (традиционен, Тайван)](../zh-TW/README.md) | [Хърватски](../hr/README.md) | [Чешки](../cs/README.md) | [Датски](../da/README.md) | [Холандски](../nl/README.md) | [Естонски](../et/README.md) | [Фински](../fi/README.md) | [Френски](../fr/README.md) | [Немски](../de/README.md) | [Гръцки](../el/README.md) | [Иврит](../he/README.md) | [Хинди](../hi/README.md) | [Унгарски](../hu/README.md) | [Индонезийски](../id/README.md) | [Италиански](../it/README.md) | [Японски](../ja/README.md) | [Каннада](../kn/README.md) | [Кхмерски](../km/README.md) | [Корейски](../ko/README.md) | [Литовски](../lt/README.md) | [Малайски](../ms/README.md) | [Малаялам](../ml/README.md) | [Маратхи](../mr/README.md) | [Непалски](../ne/README.md) | [Нигерийски Пиджин](../pcm/README.md) | [Норвежки](../no/README.md) | [Фарси (Персийски)](../fa/README.md) | [Полски](../pl/README.md) | [Португалски (Бразилия)](../pt-BR/README.md) | [Португалски (Португалия)](../pt-PT/README.md) | [Пенджаби (Гурмукхи)](../pa/README.md) | [Румънски](../ro/README.md) | [Руски](../ru/README.md) | [Сръбски (кирилица)](../sr/README.md) | [Словашки](../sk/README.md) | [Словенски](../sl/README.md) | [Испански](../es/README.md) | [Суахили](../sw/README.md) | [Шведски](../sv/README.md) | [Тагалог (Филипински)](../tl/README.md) | [Тамилски](../ta/README.md) | [Телугу](../te/README.md) | [Тайски](../th/README.md) | [Турски](../tr/README.md) | [Украински](../uk/README.md) | [Урду](../ur/README.md) | [Виетнамски](../vi/README.md) > **Предпочитате да клонирате локално?** > -> Това хранилище включва над 50 езикови превода, което значително увеличава размера за сваляне. За да клонирате без преводи, използвайте sparse checkout: +> Това хранилище включва над 50 езикови превода, които значително увеличават размера на изтеглянето. За клониране без преводи, използвайте sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,7 +52,7 @@ Phi е серия от отворени AI модели, разработени > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Това ви дава всичко необходимо за завършване на курса с много по-бързо сваляне. +> Това ви осигурява всичко необходимо, за да завършите курса с по-бързо изтегляне. ## Съдържание @@ -60,192 +60,195 @@ Phi е серия от отворени AI модели, разработени - Въведение - [Добре дошли във фамилията Phi](./md/01.Introduction/01/01.PhiFamily.md) - [Настройване на вашата среда](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Разбиране на ключови технологии](./md/01.Introduction/01/01.Understandingtech.md) - - [AI безопасност за Phi модели](./md/01.Introduction/01/01.AISafety.md) - - [Поддръжка на хардуер Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi модели и наличност на платформи](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Разбиране на ключовите технологии](./md/01.Introduction/01/01.Understandingtech.md) + - [Безопасност при AI за Phi моделите](./md/01.Introduction/01/01.AISafety.md) + - [Поддръжка на хардуер за Phi](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Phi модели и наличност през платформите](./md/01.Introduction/01/01.Edgeandcloud.md) - [Използване на Guidance-ai и Phi](./md/01.Introduction/01/01.Guidance.md) - [GitHub Marketplace модели](https://github.com/marketplace/models) - - [Каталог на Azure AI модели](https://ai.azure.com) + - [Azure AI каталог на модели](https://ai.azure.com) -- Инференция на Phi в различни среди +- Извеждане на Phi в различна среда - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [GitHub модели](./md/01.Introduction/02/02.GitHubModel.md) - - [Каталог Microsoft Foundry Model](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [GitHub Модели](./md/01.Introduction/02/02.GitHubModel.md) + - [Microsoft Foundry Каталог с модели](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Инференция Phi фамилия - - [Инференция Phi в iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Инференция Phi в Android](./md/01.Introduction/03/Android_Inference.md) - - [Инференция Phi в Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Инференция Phi в AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Инференция Phi с Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Инференция Phi в локален сървър](./md/01.Introduction/03/Local_Server_Inference.md) - - [Инференция Phi в отдалечен сървър с AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Инференция Phi с Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Инференция Phi - Vision локално](./md/01.Introduction/03/Vision_Inference.md) - - [Инференция Phi с Kaito AKS, Azure Контейнери (официална поддръжка)](./md/01.Introduction/03/Kaito_Inference.md) -- [Квантоване Phi фамилия](./md/01.Introduction/04/QuantifyingPhi.md) - - [Квантоване Phi-3.5 / 4 с llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Квантоване Phi-3.5 / 4 с разширения за генеративен AI за onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Квантоване Phi-3.5 / 4 с Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Квантоване Phi-3.5 / 4 с Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Оценка Phi - - [Отговорствен AI](./md/01.Introduction/05/ResponsibleAI.md) - - [Microsoft Foundry за оценка](./md/01.Introduction/05/AIFoundry.md) - - [Използване на Promptflow за оценка](./md/01.Introduction/05/Promptflow.md) +- Извеждане на Phi фамилията + - [Извеждане на Phi в iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Извеждане на Phi в Android](./md/01.Introduction/03/Android_Inference.md) + - [Извеждане на Phi в Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Извеждане на Phi в AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Извеждане на Phi с Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Извеждане на Phi в локален сървър](./md/01.Introduction/03/Local_Server_Inference.md) + - [Извеждане на Phi в отдалечен сървър с AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Извеждане на Phi с Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Извеждане на Phi--Визия локално](./md/01.Introduction/03/Vision_Inference.md) + - [Извеждане на Phi с Kaito AKS, Azure Контейнери (официална поддръжка)](./md/01.Introduction/03/Kaito_Inference.md) +- [Квантизиране на Phi фамилията](./md/01.Introduction/04/QuantifyingPhi.md) + - [Квантизиране на Phi-3.5 / 4 с llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Квантизиране на Phi-3.5 / 4 с разширения за генеративен AI за onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Квантизиране на Phi-3.5 / 4 с Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Квантизиране на Phi-3.5 / 4 с Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- Оценяване на Phi + - [Отговорен AI](./md/01.Introduction/05/ResponsibleAI.md) + - [Microsoft Foundry за оценяване](./md/01.Introduction/05/AIFoundry.md) + - [Използване на Promptflow за оценяване](./md/01.Introduction/05/Promptflow.md) -- RAG с Azure AI Търсене - - [Как да използваме Phi-4-mini и Phi-4-мултимодален (RAG) с Azure AI Търсене](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- RAG с Azure AI Search + - [Как да използвате Phi-4-mini и Phi-4-мултимодален (RAG) с Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud локален хибриден RAG с SQLite FTS5 и phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Примери за разработка на приложения с Phi +- Примери за разработване на Phi приложения - Текстови и чат приложения - - Примери Phi-4 + - Phi-4 Примери - [📓] [Чат с Phi-4-mini ONNX модел](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Чат с Phi-4 локален ONNX модел на .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Чат .NET конзолно приложение с Phi-4 ONNX използвайки Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - [Чат с Phi-4 локален ONNX модел .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Чат конзолно приложение на .NET с Phi-4 ONNX използващо Semnatic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Примери Phi-3 / 3.5 - - [Локален чатбот в браузър с Phi3, ONNX Runtime Web и WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Локален чатбот в браузъра с Phi3, ONNX Runtime Web и WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Мулти Модел - Интерактивен Phi-3-mini и OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [Мулти модел - интерактивни Phi-3-mini и OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Създаване на обвивка и използване на Phi-3 с MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [Оптимизация на модел - Как да оптимизирате модел Phi-3-min за ONNX Runtime Web с Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [WinUI3 приложение с Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Мулти Модел AI Захранван Пример за Бележки приложение](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Фина настройка и интеграция на персонализирани Phi-3 модели с Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Фина настройка и интеграция на персонализирани Phi-3 модели с Prompt flow в Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Оценка на фината настройка на Phi-3 / Phi-3.5 модел в Microsoft Foundry с акцент върху отговорните принципи на AI на Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct пример за предсказване на език (китайски/английски)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + -[WinUI3 мултимоделно приложение за бележки с AI](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Фино настройване и интеграция на персонализирани Phi-3 модели с Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Фино настройване и интеграция на персонализирани Phi-3 модели с Prompt flow в Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Оценка на фино настроения модел Phi-3 / Phi-3.5 в Microsoft Foundry с фокус върху отговорните AI принципи на Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct пример за езиково предсказване (китайски/английски)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG чатбот](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Използване на Windows GPU за създаване на Prompt flow решение с Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Използване на Microsoft Phi-3.5 tflite за създаване на Android приложение](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Въпроси и отговори .NET пример използващ локален ONNX Phi-3 модел с Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Конзолно чат .NET приложение с Semantic Kernel и Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Пример за Q&A .NET с локален ONNX Phi-3 модел чрез Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Конзолно чат .NET приложение със Semantic Kernel и Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Примери за код с Azure AI Inference SDK + - Примери за Azure AI Inference SDK, базирани на код - Примери Phi-4 - - [📓] [Генериране на код на проект използвайки Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [📓] [Генериране на проектен код с Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Примери Phi-3 / 3.5 - - [Създайте собствен Visual Studio Code GitHub Copilot Chat с Microsoft Phi-3 Семейство](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Създаване на собствен Visual Studio Code Chat Copilot агент с Phi-3.5 чрез GitHub модели](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Създайте свой собствен Visual Studio Code GitHub Copilot Chat с Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Създайте свой собствен Visual Studio Code Chat Copilot Agent с Phi-3.5 чрез GitHub модели](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - Примери за разширено разсъждение + - Примери за усъвършенствано разсъждение - Примери Phi-4 - - [📓] [Phi-4-mini-reasoning или Phi-4-reasoning примери](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Фина настройка на Phi-4-mini-reasoning с Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Фина настройка на Phi-4-mini-reasoning с Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning с GitHub Модели](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning с Microsoft Foundry Модели](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [📓] [Примери Phi-4-mini-reasoning или Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Фино настройване на Phi-4-mini-reasoning със Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Фино настройване на Phi-4-mini-reasoning с Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-reasoning с GitHub модели](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-reasoning с Microsoft Foundry модели](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Демонстрации - - [Phi-4-mini демонстрации хоствани на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal демонстрации хоствани на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Примери за визия + - [Phi-4-mini демонстрации, хоствани на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal демонстрации, хоствани на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Примери във визията - Примери Phi-4 - [📓] [Използване на Phi-4-multimodal за четене на изображения и генериране на код](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Примери Phi-3 / 3.5 - - [📓][Phi-3-видение - превод от текст на изображение към текст](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-видение-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-видение CLIP Вграждане](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [ДЕМО: Phi-3 Рециклиране](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-видение - Визуален езиков асистент - с Phi3-Vision и OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Видение мулти-фрейм или мулти-изображение пример](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision локален ONNX модел с Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Меню базиран Phi-3 Vision локален ONNX модел с Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Примери за разсъждение с визия - - Phi-4-Reasoning-Vision-15B - - [📓] [Използване на Phi-4-Reasoning-Vision-15B за детекция на jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓][Phi-3-vision-Image текст към текст](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vision CLIP_embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [ДЕМО: Phi-3 рециклиране](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - Визуален езиков асистент - с Phi3-Vision и OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Визия Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Визия OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Визия мултирамков или мултиизображен пример](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Визия Локален ONNX модел чрез Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Меню базиран Phi-3 Визия Локален ONNX модел чрез Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Примери за разсъждения и визия + - Phi-4-Reasoning-Vision-15B + - [📓] [Използване на Phi-4-Reasoning-Vision-15B за откриване на неправилно пресичане на пешеходна пътека](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [Използване на Phi-4-Reasoning-Vision-15B за математика](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Използване на Phi-4-Reasoning-Vision-15B за детекция на UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [Използване на Phi-4-Reasoning-Vision-15B за откриване на UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Примери за математика - - Phi-4-Mini-Flash-Reasoning-Instruct Примери [Демонстрация на математика с Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-Mini-Flash-Reasoning-Instruct Примери [Демо математика с Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - Примери за аудио - Примери Phi-4 - [📓] [Извличане на аудио транскрипции с Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal аудио пример](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal пример за превод на реч](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET конзолно приложение използващо Phi-4-multimodal аудио за анализ на аудио файл и генериране на транскрипция](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [📓] [Пример за Phi-4-multimodal аудио](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Пример за Phi-4-multimodal превод на реч](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET конзолно приложение, използващо Phi-4-multimodal Audio за анализ на аудио файл и генериране на транскрипт](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - Примери за MOE + - Примери MOE - Примери Phi-3 / 3.5 - - [📓] [Phi-3.5 смес от експерти (MoEs) пример с социални медии](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Phi-3.5 модели Mixture of Experts (MoEs) пример в социалните медии](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - [📓] [Създаване на Retrieval-Augmented Generation (RAG) pipeline с NVIDIA NIM Phi-3 MOE, Azure AI Search и LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Примери за повикване на функции + - Примери за извикване на функции - Примери Phi-4 🆕 - - [📓] [Използване на Function Calling с Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Използване на Function Calling за създаване на мулти-агенти с Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Използване на Function Calling с Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Използване на Function Calling с ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - [📓] [Използване на извикване на функции с Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Използване на извикване на функции за създаване на мулти-агенти с Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Използване на извикване на функции с Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Използване на извикване на функции с ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - Примери за мултимодално смесване - Примери Phi-4 🆕 - [📓] [Използване на Phi-4-multimodal като технологичен журналист](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET конзолно приложение използващо Phi-4-multimodal за анализ на изображения](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Фина настройка на Phi Примери - - [Сценарии за фина настройка](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Фина настройка срещу RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Фина настройка: Нека Phi-3 стане индустриален експерт](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Фина настройка на Phi-3 с AI Toolkit за VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Фина настройка на Phi-3 с Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Фина настройка на Phi-3 с Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Фина настройка на Phi-3 с QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Фина настройка на Phi-3 с Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Фина настройка на Phi-3 с Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Фина настройка с Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Практическа лаборатория за фина настройка с Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [Фина настройка на Phi-3-vision с Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Фина настройка на Phi-3 с Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Фина настройка на Phi-3-vision (официална поддръжка)](./md/03.FineTuning/FineTuning_Vision.md) - - [Фино настройване на Phi-3 с Kaito AKS, Azure контейнери (официална поддръжка)](./md/03.FineTuning/FineTuning_Kaito.md) + - [.NET конзолно приложение, използващо Phi-4-multimodal за анализ на изображения](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Фино настройване на Phi примери + - [Сценарии за фино настройване](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Фино настройване срещу RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Фино настройване: Нека Phi-3 стане индустриален експерт](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Фино настройване на Phi-3 с AI Toolkit за VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Фино настройване на Phi-3 с Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Фино настройване на Phi-3 с Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Фино настройване на Phi-3 с QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Фино настройване на Phi-3 с Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Фино настройване на Phi-3 с Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Фино настройване с Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Фино настройване с Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) + - [Фино настройване на Phi-3-vision с Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Фино настройване на Phi-3 с Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Фино настройване на Phi-3-vision (официална поддръжка)](./md/03.FineTuning/FineTuning_Vision.md) + - [Фино настройване на Phi-3 с Kaito AKS, Azure Containers (официална поддръжка)](./md/03.FineTuning/FineTuning_Kaito.md) - [Фино настройване на Phi-3 и 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) -- Практически лабораторен курс - - [Изследване на най-съвременните модели: LLM, SLM, локална разработка и още](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Отваряне на потенциала на NLP: Фино настройване с Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) +- Практическо обучение + - [Изследване на най-новите модели: LLMs, SLMs, локална разработка и други](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Отключване на потенциала на NLP: Фино настройване с Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Академични изследователски статии и публикации - - [Учебници са всичко, от което се нуждаете II: технически доклад phi-1.5](https://arxiv.org/abs/2309.05463) - - [Технически доклад Phi-3: Високо способен езиков модел локално на вашия телефон](https://arxiv.org/abs/2404.14219) +- Академични изследователски доклади и публикации + - [Учебниците са всичко, от което се нуждаете II: технически доклад phi-1.5](https://arxiv.org/abs/2309.05463) + - [Технически доклад Phi-3: Висококапацитивен езиков модел локално на телефона ви](https://arxiv.org/abs/2404.14219) - [Технически доклад Phi-4](https://arxiv.org/abs/2412.08905) - - [Технически доклад Phi-4-Mini: Компактни, но мощни мултимодални езикови модели чрез Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Оптимизиране на малки езикови модели за извикване на функции в автомобил](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Фино настройване на PHI-3 за отговор на въпроси с множествен избор: методология, резултати и предизвикателства](https://arxiv.org/abs/2501.01588) - - [Технически доклад Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Технически доклад Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Технически доклад Phi-4-Mini: Компактни, но мощни мултимодални езикови модели чрез смесване на LoRA](https://arxiv.org/abs/2503.01743) + - [Оптимизиране на малки езикови модели за функция за повикване в превозното средство](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Фино настройване на PHI-3 за въпроси с множествен избор: методология, резултати и предизвикателства](https://arxiv.org/abs/2501.01588) + - [Технически доклад Phi-4-разсъждения](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Технически доклад Phi-4-mini-разсъждения](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Използване на Phi модели ### Phi в Microsoft Foundry -Можете да научите как да използвате Microsoft Phi и как да изграждате крайни края решения на различни хардуерни устройства. За да изпитате Phi лично, започнете с експериментиране с моделите и персонализиране на Phi за вашите сценарии, като използвате [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Можете да научите повече от Ръководството за започване с [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Можете да научите как да използвате Microsoft Phi и как да изграждате комплексни решения (E2E) на различни хардуерни устройства. За да усетите Phi сами, започнете с експериментиране с моделите и персонализация на Phi за вашите сценарии, използвайки [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Можете да научите повече в Започване с [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Пясъчник** -Всеки модел има специален пясъчник, където можете да тествате модела в [Azure AI Playground](https://aka.ms/try-phi3). +**Playground** +Всеки модел има специално игрално поле за тестове [Azure AI Playground](https://aka.ms/try-phi3). ### Phi в GitHub модели -Можете да научите как да използвате Microsoft Phi и как да изграждате крайно решения на различни хардуерни устройства. За да изпитате Phi лично, започнете с експериментиране с модела и персонализиране на Phi за вашите сценарии, като използвате [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Можете да научите повече от Ръководството за започване с [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Можете да научите как да използвате Microsoft Phi и как да изграждате комплексни решения (E2E) на различни хардуерни устройства. За да усетите Phi сами, започнете с експериментиране с модела и персонализация на Phi за вашите сценарии, използвайки [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Можете да научите повече в Започване с [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Пясъчник** -Всеки модел има специален [пясъчник за тестване](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Playground** +Всеки модел има специално [игрално поле за тест на модела](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi в Hugging Face -Моделът също може да бъде намерен в [Hugging Face](https://huggingface.co/microsoft) +Можете също да намерите модела на [Hugging Face](https://huggingface.co/microsoft) -**Пясъчник** -[Hugging Chat пясъчник](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Playground** + [Игрално поле Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Други курсове + ## 🎒 Други курсове Нашият екип произвежда и други курсове! Разгледайте: @@ -273,8 +276,8 @@ Phi е серия от отворени AI модели, разработени --- ### Основно обучение -[![ML за начинаещи](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Наука за данни за начинаещи](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![Машинно обучение за начинаещи](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Данни наука за начинаещи](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI за начинаещи](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![Киберсигурност за начинаещи](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Уеб разработка за начинаещи](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) @@ -284,41 +287,43 @@ Phi е серия от отворени AI модели, разработени --- ### Серия Copilot -[![Copilot за програмиране с AI в тандем](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot за AI двойно програмиране](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot за C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![Приключение с Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## Отговорен AI -Microsoft е ангажиран да помага на клиентите си да използват AI продуктите ни отговорно, споделяйки нашите знания и изграждайки партньорства базирани на доверие чрез инструменти като Transparency Notes и Impact Assessments. Много от тези ресурси могат да бъдат намерени на [https://aka.ms/RAI](https://aka.ms/RAI). -Подходът на Microsoft към отговорния AI се основава на нашите AI принципи за справедливост, надеждност и безопасност, поверителност и сигурност, приобщаване, прозрачност и отчетност. +Microsoft се ангажира да помогне на клиентите си да използват AI продуктите отговорно, да споделя нашите знания и да изгражда партньорства, базирани на доверие, чрез инструменти като Бележки за прозрачност и Оценки на въздействието. Много от тези ресурси могат да бъдат намерени на [https://aka.ms/RAI](https://aka.ms/RAI). +Подходът на Microsoft към отговорния AI е основан на нашите AI принципи за справедливост, надеждност и безопасност, поверителност и сигурност, приобщаване, прозрачност и отчетност. + +Големите модели за естествен език, изображения и реч – като тези, използвани в този пример – могат потенциално да се държат по начини, които са несправедливи, ненадеждни или обидни, като по този начин причиняват вреди. Моля, консултирайте се с [Бележката за прозрачност на услугата Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), за да се информирате за рисковете и ограниченията. + -Големи модели за естествен език, изображения и реч — като тези, използвани в този пример — потенциално могат да се държат несправедливо, ненадеждно или обидно, причинявайки вреди. Моля, прегледайте [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), за да бъдете информирани за рисковете и ограниченията. -Препоръчителният подход за намаляване на тези рискове е да включите система за безопасност във вашата архитектура, която може да открива и предотвратява вредно поведение. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) предоставя независим слой на защита, който може да открива вредно съдържание, генерирано от потребители и AI, в приложения и услуги. Azure AI Content Safety включва текстови и графични API, които ви позволяват да откривате материал, който е вреден. В Microsoft Foundry, услугата Content Safety ви позволява да преглеждате, изследвате и изпробвате примерен код за откриване на вредно съдържание в различни модалности. Следващата [документация за бърз старт](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ви насочва през процеса на изпращане на заявки към услугата. +Препоръчителният подход за смекчаване на тези рискове е да включите система за безопасност във вашата архитектура, която може да открива и предотвратява вредно поведение. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) предоставя независим защитен слой, способен да открива вредно съдържание, генерирано от потребители и от AI, в приложения и услуги. Azure AI Content Safety включва текстови и изображенчески API, които ви позволяват да откривате материал, който е вреден. В Microsoft Foundry услугата Content Safety ви позволява да виждате, разглеждате и изпробвате примерен код за откриване на вредно съдържание в различни модалности. Следващата [документация за бърз старт](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ви води през процеса на изпращане на заявки към услугата. -Друг аспект, който трябва да се вземе предвид, е цялостната производителност на приложението. При мултимодални и мултимоделни приложения, считаме производителността за това, че системата работи както вие и вашите потребители очаквате, включително и да не генерира вредни изходи. Важно е да оцените производителността на вашето цялостно приложение, използвайки [оценители за производителност, качество, риск и безопасност](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Също така имате възможност да създавате и оценявате с [персонализирани оценители](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Друг аспект, който трябва да се вземе предвид, е общата производителност на приложението. При мултимодални и мултимоделови приложения, считаме, че производителността означава, че системата работи така, както вие и вашите потребители очаквате, включително да не генерира вредни резултати. Важно е да оцените производителността на цялото си приложение, като използвате [оценители за производителност, качество, риск и безопасност](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Също така имате възможност да създавате и оценявате с [персонализирани оценители](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Можете да оцените вашето AI приложение в средата за разработка, използвайки [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). След предоставяне на тестов набор от данни или цел, генерирането на вашето генеративно AI приложение се измерва количествено с вградени оценители или персонализирани оценители по ваш избор. За да започнете с azure ai evaluation sdk за оценка на вашата система, можете да следвате [ръководството за бърз старт](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). След като изпълните оценъчно изпълнение, можете да [визуализирате резултатите в Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Можете да оцените своето AI приложение в средата за разработка, като използвате [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Независимо дали имате тестов набор от данни или цел, генерираните резултати на вашето генеративно AI приложение се измерват количествено с вградени оценители или персонализирани оценители по ваш избор. За да започнете с azure ai evaluation sdk за оценка на вашата система, можете да следвате [ръководството за бърз старт](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). След като изпълните оценъчен цикъл, можете да [визуализирате резултатите в Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Търговски марки -Този проект може да съдържа търговски марки или лога за проекти, продукти или услуги. Употребата на търговски марки или лога на Microsoft е разрешена и подлежи на спазване на [Правилата за търговски марки и бранд на Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Използването на търговски марки или лога на Microsoft в модифицирани версии на този проект не трябва да предизвиква объркване или да подразбира спонсорство на Microsoft. Всяко използване на търговски марки или лога на трети страни подлежи на политиките на тези трети страни. +Този проект може да съдържа търговски марки или лога на проекти, продукти или услуги. Пълномощното използване на търговски марки или лога на Microsoft се подчинява на [Правилата за търговски марки и бранд на Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Използването на търговски марки или лога на Microsoft в модифицирани версии на този проект не трябва да създава объркване или да предполага спонсорство от Microsoft. Всяко използване на търговски марки или лога на трети страни се подчинява на правилата на тези трети страни. ## Получаване на помощ -Ако срещнете затруднения или имате въпроси относно създаването на AI приложения, присъединете се към: +Ако се затрудните или имате въпроси относно изграждането на AI приложения, присъединете се към: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Ако имате обратна връзка за продукта или грешки по време на разработка, посетете: +Ако имате обратна връзка за продукта или грешки при изграждането, посетете: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Отказ от отговорност**: -Този документ е преведен с помощта на AI преводаческа услуга [Co-op Translator](https://github.com/Azure/co-op-translator). Въпреки че се стремим към точност, моля, имайте предвид, че автоматизираните преводи могат да съдържат грешки или неточности. Оригиналният документ на неговия първичен език трябва да се счита за авторитетен източник. За критична информация се препоръчва професионален човешки превод. Не носим отговорност за никакви недоразумения или неправилни интерпретации, произтичащи от използването на този превод. +**Отказ от отговорност**: +Този документ е преведен с помощта на AI преводачески услуга [Co-op Translator](https://github.com/Azure/co-op-translator). Въпреки че се стремим към точност, моля имайте предвид, че автоматизираните преводи могат да съдържат грешки или неточности. Оригиналният документ на неговия роден език трябва да се счита за авторитетен източник. За критична информация се препоръчва професионален човешки превод. Ние не носим отговорност за каквито и да е недоразумения или неправилни тълкувания, произтичащи от използването на този превод. \ No newline at end of file diff --git a/translations/bg/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/bg/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..5c7494dff --- /dev/null +++ b/translations/bg/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Хибриден RAG с Microsoft phi-4-mini и SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Автор:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Фокус:** Малки езикови модели (SLMs), SQLite FTS5 BM25, Плътни вграждания, Обединение на реципрочен ранг (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Мотивация: Дилемата с възстановяването на ключови думи в локални SLM\n", + "Стандартните архитектури RAG, разчитащи само на плътни векторни вграждания, често не успяват да възстановят точни числови токени (например `2,340,000 TL`, кодове на договори, номера на сметки). \n", + "Обратно, разреденият лексикален търсене (BM25) пропуска семантични синоними и преписани въпроси.\n", + "\n", + "Тази готварска книга показва как да се реализира **високоскоростен, в паметта Хибриден Retrieval engine**, който комбинира:\n", + "1. **Плътни вектори** (Косинусова подобност)\n", + "2. **Разредено лексикално търсене** (SQLite FTS5 BM25)\n", + "3. **Обединение на реципрочен ранг (RRF, $k=60$)**\n", + "4. **Генериране на обосновани цитати (`[1]`, `[2]`)** с Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Двоен SQLite Схема (Гъсти Вектори + Виртуална FTS5 BM25 Таблица)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Бенчмарк за приемане и изпълнение на проби\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Формулиране на Подкрепена Подсещалка за Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Отказ от отговорност**:\nТози документ е преведен с помощта на AI преводачески услуга [Co-op Translator](https://github.com/Azure/co-op-translator). Въпреки че се стремим към точност, моля имайте предвид, че автоматизираните преводи могат да съдържат грешки или неточности. Оригиналният документ на неговия роден език трябва да се счита за авторитетен източник. За критична информация се препоръчва професионален човешки превод. Ние не носим отговорност за каквито и да е недоразумения или неправилни тълкувания, произтичащи от използването на този превод.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/bn/.co-op-translator.json b/translations/bn/.co-op-translator.json index e9c0aa5f8..2aa8ad722 100644 --- a/translations/bn/.co-op-translator.json +++ b/translations/bn/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "bn" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T23:57:11+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:37:54+00:00", "source_file": "README.md", "language_code": "bn" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "bn" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:43:25+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "bn" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:18+00:00", diff --git a/translations/bn/README.md b/translations/bn/README.md index 90130a518..aadf3c5dd 100644 --- a/translations/bn/README.md +++ b/translations/bn/README.md @@ -1,42 +1,42 @@ -# Phi কুকবুক: Microsoft-এর Phi মডেলের সাথে হাতেকলমে উদাহরণসমূহ +# ফাই কুকবুক: মাইক্রোসফটের ফাই মডেল নিয়ে হাতেকলমে উদাহরণ -[![GitHub Codespaces-এ নমুনা খুলুন এবং ব্যবহার করুন](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![GitHub কোডস্পেসে নমুনাগুলো খুলুন এবং ব্যবহার করুন](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Dev Containers-এ খুলুন](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub অবদানকারী](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ইস্যু](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub পুল রিকোয়েস্ট](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![পিআর-স্বাগত](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub অবদানকারীরা](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub সমস্যা](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub পুল অনুরোধ](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![পুল অনুরোধ স্বাগতম](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![GitHub পর্যবেক্ষক](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub ফর্ক](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub তারকা](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) -[![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) +[![মাইক্রোসফট ফাউন্ড্রি ডিসকর্ড](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi হলো Microsoft দ্বারা উন্নত করা ওপেন সোর্স AI মডেলসের একটি সিরিজ। +ফাই হল মাইক্রোসফট কর্তৃক বিকাশকৃত একটি ওপেন সোর্স AI মডেল সিরিজ। -Phi বর্তমানে সবচেয়ে শক্তিশালী এবং খরচ-কার্যকর ছোট ভাষা মডেল (SLM), যেটি বহু-ভাষায়, যুক্তি, টেক্সট/চ্যাট উৎপাদন, কোডিং, ছবি, অডিও এবং অন্যান্য বিভিন্ন পরিস্থিতিতে খুব ভালো বেন্চমার্কধারী। +ফাই বর্তমানে সবচেয়ে শক্তিশালী এবং খরচ সাশ্রয়ী ছোট ভাষার মডেল (SLM), যা বহু-ভাষা, যুক্তিবিচার, টেক্সট/চ্যাট সৃষ্টিসহ, কোডিং, ছবি, অডিও এবং অন্যান্য পরিস্থিতিতে খুব ভালো পারফরম্যান্স প্রদান করে। -আপনি Phi-কে ক্লাউড অথবা এজ ডিভাইসে ডিপ্লয় করতে পারেন, এবং সীমিত কম্পিউটিং ক্ষমতা নিয়ে সহজেই জেনারেটিভ AI অ্যাপ্লিকেশন তৈরি করতে পারেন। +আপনি ফাই ক্লাউডে বা এজ ডিভাইসে ডিপ্লয় করতে পারেন, এবং সীমিত কম্পিউটিং শক্তি দিয়ে সহজেই জেনারেটিভ AI অ্যাপ্লিকেশন তৈরি করতে পারেন। -এই রিসোর্স ব্যবহার শুরু করতে নিম্নলিখিত ধাপগুলো অনুসরণ করুন: -1. **রিপোজিটরি ফর্ক করুন**: ক্লিক করুন [![GitHub ফর্ক](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **রিপোজিটরি ক্লোন করুন**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Microsoft AI Discord কমিউনিটিতে যোগ দিন এবং বিশেষজ্ঞ ও অন্যান্য ডেভেলপারদের সাথে পরিচিত হন**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +এই রিসোর্সগুলো ব্যবহার শুরু করতে নিম্নলিখিত ধাপগুলো অনুসরণ করুন: +1. **রেপোসিটরি ফর্ক করুন**: ক্লিক করুন [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **রেপোসিটরি ক্লোন করুন**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**মাইক্রোসফট AI ডিসকর্ড কমিউনিটিতে যোগ দিন এবং বিশেষজ্ঞ এবং সহ-разработчиковদের সাথে পরিচিত হন**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/bn/cover.eb18d1b9605d754b.webp) ### 🌐 বহুভাষিক সমর্থন -#### GitHub Action এর মাধ্যমে সমর্থিত (স্বয়ংক্রিয় ও সর্বদা আপ-টু-ডেট) +#### GitHub অ্যাকশনের মাধ্যমে সমর্থিত (স্বয়ংক্রিয় এবং সর্বদা আপ-টু-ডেট) -[আরবি](../ar/README.md) | [বাংলা](./README.md) | [বুলগেরিয়ান](../bg/README.md) | [বর্মিজ (মায়ানমার)](../my/README.md) | [চীনা (সরলীকৃত)](../zh-CN/README.md) | [চীনা (প্রথাগত, হংকং)](../zh-HK/README.md) | [চীনা (প্রথাগত, ম্যাকাও)](../zh-MO/README.md) | [চীনা (প্রথাগত, তাইওয়ান)](../zh-TW/README.md) | [ক্রোয়েশিয়ান](../hr/README.md) | [চেক](../cs/README.md) | [ড্যানিশ](../da/README.md) | [ডাচ](../nl/README.md) | [এস্তোনিয়ান](../et/README.md) | [ফিনিশ](../fi/README.md) | [ফরাসি](../fr/README.md) | [জার্মান](../de/README.md) | [গ্রিক](../el/README.md) | [হিব্রু](../he/README.md) | [হিন্দি](../hi/README.md) | [হাঙ্গেরিয়ান](../hu/README.md) | [ইন্দোনেশীয়](../id/README.md) | [ইতালিয়ান](../it/README.md) | [জাপানি](../ja/README.md) | [কন্নড়](../kn/README.md) | [খমের](../km/README.md) | [কোরিয়ান](../ko/README.md) | [লিথুয়ানিয়ান](../lt/README.md) | [মালয়](../ms/README.md) | [মালায়ালাম](../ml/README.md) | [মরাঠী](../mr/README.md) | [নেপালি](../ne/README.md) | [নাইজেরিয়ান পিজিন](../pcm/README.md) | [নরওয়েজীয়](../no/README.md) | [পারসীয় (ফারসি)](../fa/README.md) | [পোলিশ](../pl/README.md) | [পর্তুগিজ (ব্রাজিল)](../pt-BR/README.md) | [পর্তুগিজ (পর্তুগাল)](../pt-PT/README.md) | [পাঞ্জাবী (গুরুমুখী)](../pa/README.md) | [রোমানিয়ান](../ro/README.md) | [রুশিয়ান](../ru/README.md) | [সার্বিয়ান (সিরিলিক)](../sr/README.md) | [স্লোভাক](../sk/README.md) | [স্লোভেনীয়](../sl/README.md) | [স্প্যানিশ](../es/README.md) | [সোয়াহিলি](../sw/README.md) | [সুইডিশ](../sv/README.md) | [টাগালগ (ফিলিপিনো)](../tl/README.md) | [তামিল](../ta/README.md) | [তেলুগু](../te/README.md) | [থাই](../th/README.md) | [তুর্কি](../tr/README.md) | [ইউক্রেনিয়ান](../uk/README.md) | [উর্দু](../ur/README.md) | [ভিয়েতনামীজ](../vi/README.md) +[Arabic](../ar/README.md) | [বাংলা](./README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **লোকাল ক্লোন করতে বাছাই করুন?** +> **স্থানীয়ভাবে ক্লোন করতে ইচ্ছুক?** > -> এই রিপোজিটরিতে ৫০+ ভাষার অনুবাদ অন্তর্ভুক্ত রয়েছে যা ডাউনলোড সাইজ অনেক বড় করে। অনুবাদ ছাড়া ক্লোন করতে sparse checkout ব্যবহার করুন: +> এই রেপোসিটরিতে ৫০+ ভাষার অনুবাদ রয়েছে যা ডাউনলোড সাইজ অনেক মাত্রায় বৃদ্ধি করে। অনুবাদ ছাড়া ক্লোন করতে স্পার্স চেকআউট ব্যবহার করুন: > > **Bash / macOS / Linux:** > ```bash @@ -52,266 +52,272 @@ Phi বর্তমানে সবচেয়ে শক্তিশালী > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> এর ফলে কোর্স সম্পন্ন করার জন্য যা যা দরকার সবকিছু অনেক দ্রুত ডাউনলোড হবে। +> এটি আপনাকে দ্রুত ডাউনলোডের মাধ্যমে কোর্স সম্পন্ন করার জন্য প্রয়োজনীয় সবকিছু দেবে। -## বিষয়বস্তু তালিকা +## বিষয়বস্তু সূচি -- ভূমিকা - - [Phi পরিবারে স্বাগতম](./md/01.Introduction/01/01.PhiFamily.md) - - [আপনার পরিবেশ সেটআপ করা](./md/01.Introduction/01/01.EnvironmentSetup.md) +- পরিচিতি + - [ফাই পরিবারে স্বাগতম](./md/01.Introduction/01/01.PhiFamily.md) + - [আপনার পরিবেশ স্থাপন](./md/01.Introduction/01/01.EnvironmentSetup.md) - [মূল প্রযুক্তিগুলো বোঝা](./md/01.Introduction/01/01.Understandingtech.md) - - [Phi মডেলের জন্য AI নিরাপত্তা](./md/01.Introduction/01/01.AISafety.md) - - [Phi হার্ডওয়্যার সমর্থন](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi মডেল ও প্ল্যাটফর্ম জুড়ে সুবিধা](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Guidance-ai এবং Phi ব্যবহার](./md/01.Introduction/01/01.Guidance.md) - - [GitHub মার্কেটপ্লেস মডেলসমূহ](https://github.com/marketplace/models) + - [ফাই মডেলগুলোর জন্য AI নিরাপত্তা](./md/01.Introduction/01/01.AISafety.md) + - [ফাই হার্ডওয়্যার সমর্থন](./md/01.Introduction/01/01.Hardwaresupport.md) + - [প্ল্যাটফর্ম জুড়ে ফাই মডেল ও প্রাপ্যতা](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Guidance-ai এবং ফাই ব্যবহার](./md/01.Introduction/01/01.Guidance.md) + - [GitHub মার্কেটপ্লেস মডেল](https://github.com/marketplace/models) - [Azure AI মডেল ক্যাটালগ](https://ai.azure.com) -- বিভিন্ন পরিবেশে Phi ইনফারেন্স +- বিভিন্ন পরিবেশে ফাই ইনফারেন্স - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [GitHub মডেলসমূহ](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry মডেল ক্যাটালগ](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [GitHub মডেল](./md/01.Introduction/02/02.GitHubModel.md) + - [মাইক্রোসফট ফাউন্ড্রি মডেল ক্যাটালগ](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [AI টুলকিট VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - -- Phi পরিবার ইনফারেন্স - - [iOS-এ Phi ইনফারেন্স](./md/01.Introduction/03/iOS_Inference.md) - - [অ্যান্ড্রয়েড-এ Phi ইনফারেন্স](./md/01.Introduction/03/Android_Inference.md) - - [Jetson-এ Phi ইনফারেন্স](./md/01.Introduction/03/Jetson_Inference.md) - - [AI পিসিতে Phi ইনফারেন্স](./md/01.Introduction/03/AIPC_Inference.md) - - [Apple MLX ফ্রেমওয়ার্ক দিয়ে Phi ইনফারেন্স](./md/01.Introduction/03/MLX_Inference.md) - - [লোকাল সার্ভারে Phi ইনফারেন্স](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI Toolkit দিয়ে রিমোট সার্ভারে Phi ইনফারেন্স](./md/01.Introduction/03/Remote_Interence.md) - - [Rust-এর সাথে Phi ইনফারেন্স](./md/01.Introduction/03/Rust_Inference.md) - - [লোকালে Phi--ভিশন ইনফারেন্স](./md/01.Introduction/03/Vision_Inference.md) - - [Kaito AKS, Azure Containers (সরকারি সমর্থন) দিয়ে Phi ইনফারেন্স](./md/01.Introduction/03/Kaito_Inference.md) -- [Phi পরিবারকে পরিমাণগত করা](./md/01.Introduction/04/QuantifyingPhi.md) - - [llama.cpp দিয়ে Phi-3.5 / 4 পরিমাণগত করা](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime এর জন্য জেনারেটিভ AI এক্সটেনশনের মাধ্যমে Phi-3.5 / 4 পরিমাণগত করা](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Intel OpenVINO ব্যবহার করে Phi-3.5 / 4 পরিমাণগত করা](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Apple MLX ফ্রেমওয়ার্ক ব্যবহার করে Phi-3.5 / 4 পরিমাণগত করা](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Phi মূল্যায়ন - - [Responsible AI](./md/01.Introduction/05/ResponsibleAI.md) - - [Microsoft Foundry মূল্যায়নের জন্য](./md/01.Introduction/05/AIFoundry.md) + - [ফাউন্ড্রি লোকাল](./md/01.Introduction/02/07.FoundryLocal.md) + +- ফাই পরিবারে ইনফারেন্স + - [iOS-এ ফাই ইনফারেন্স](./md/01.Introduction/03/iOS_Inference.md) + - [অ্যান্ড্রয়েডে ফাই ইনফারেন্স](./md/01.Introduction/03/Android_Inference.md) + - [জেটসনে ফাই ইনফারেন্স](./md/01.Introduction/03/Jetson_Inference.md) + - [AI PC-এ ফাই ইনফারেন্স](./md/01.Introduction/03/AIPC_Inference.md) + - [Apple MLX Framework দিয়ে ফাই ইনফারেন্স](./md/01.Introduction/03/MLX_Inference.md) + - [লোকাল সার্ভারে ফাই ইনফারেন্স](./md/01.Introduction/03/Local_Server_Inference.md) + - [AI টুলকিট ব্যবহার করে রিমোট সার্ভারে ফাই ইনফারেন্স](./md/01.Introduction/03/Remote_Interence.md) + - [রাস্ট দিয়ে ফাই ইনফারেন্স](./md/01.Introduction/03/Rust_Inference.md) + - [লোকালে ফাই--Vision ইনফারেন্স](./md/01.Introduction/03/Vision_Inference.md) + - [Kaito AKS, Azure Containers (সরকারি সমর্থন) সহ ফাই ইনফারেন্স](./md/01.Introduction/03/Kaito_Inference.md) +- [ফাই পরিবার কোয়ান্টিফিকেশন](./md/01.Introduction/04/QuantifyingPhi.md) + - [llama.cpp ব্যবহার করে Phi-3.5 / 4 কোয়ান্টাইজিং](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntime-এর জন্য জেনারেটিভ AI এক্সটেনশন ব্যবহার করে Phi-3.5 / 4 কোয়ান্টাইজিং](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Intel OpenVINO ব্যবহার করে Phi-3.5 / 4 কোয়ান্টাইজিং](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Apple MLX Framework ব্যবহার করে Phi-3.5 / 4 কোয়ান্টাইজিং](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- ফাই মূল্যায়ন + - [দায়িত্বশীল AI](./md/01.Introduction/05/ResponsibleAI.md) + - [মূল্যায়নের জন্য মাইক্রোসফট ফাউন্ড্রি](./md/01.Introduction/05/AIFoundry.md) - [মূল্যায়নের জন্য Promptflow ব্যবহার](./md/01.Introduction/05/Promptflow.md) -- Azure AI সার্চ সহ RAG - - [Phi-4-mini এবং Phi-4-multimodal(RAG) Azure AI সার্চের সাথে কীভাবে ব্যবহার করবেন](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- Azure AI Search সহ RAG + - [Phi-4-mini এবং Phi-4-multimodal (RAG) Azure AI সার্চের সঙ্গে ব্যবহার করার উপায়](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [SQLite FTS5 এবং phi-4-mini দিয়ে জিরো-ক্লাউড লোকাল হাইব্রিড RAG](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi অ্যাপ্লিকেশন উন্নয়ন নমুনাসমূহ - - টেক্সট ও চ্যাট অ্যাপ্লিকেশন +- ফাই অ্যাপ্লিকেশন উন্নয়ন নমুনা + - টেক্সট এবং চ্যাট অ্যাপ্লিকেশন - Phi-4 নমুনা - [📓] [Phi-4-mini ONNX মডেলের সাথে চ্যাট](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 লোকাল ONNX মডেল .NET দিয়ে চ্যাট](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Semantic Kernel ব্যবহার করে Phi-4 ONNX সহ .NET কনসোল অ্যাপ চ্যাট](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 নমুনাসমূহ - - [Phi3, ONNX Runtime Web এবং WebGPU ব্যবহার করে ব্রাউজারে লোকাল চ্যাটবোট](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino চ্যাট](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [মাল্টি মডেল - ইন্টারঅ্যাকটিভ Phi-3-মিনি এবং OpenAI উইস্পার](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - একটি র‍্যাপার তৈরি করা এবং Phi-3 এর সাথে MLFlow ব্যবহার করা](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [মডেল অপটিমাইজেশন - Olive দিয়ে ONNX Runtime ওয়েবের জন্য Phi-3-মিন মডেল কীভাবে অপটিমাইজ করবেন](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [Phi-3 মিনি-4k-instruct-onnx সহ WinUI3 অ্যাপ](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 মাল্টি মডেল AI চালিত নোটস অ্যাপ নমুনা](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Prompt flow সহ কাস্টম Phi-3 মডেল ফাইন-টিউন এবং ইন্টিগ্রেশন](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Microsoft Foundry তে Prompt flow সহ কাস্টম Phi-3 মডেল ফাইন-টিউন এবং ইন্টিগ্রেশন](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Microsoft এর দায়িত্বশীল AI নীতিমালা উপর মনোযোগ দিয়ে Microsoft Foundry তে Fine-tuned Phi-3 / Phi-3.5 মডেল মূল্যায়ন](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct ভাষা ভবিষ্যদ্বাণী নমুনা (চীনা/ইংরেজি)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG চ্যাটবট](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Phi-3.5-Instruct ONNX দিয়ে উইন্ডোজ GPU ব্যবহার করে Prompt flow সমাধান তৈরি](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Microsoft Phi-3.5 tflite ব্যবহার করে অ্যান্ড্রয়েড অ্যাপ তৈরি](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [স্থানীয় ONNX Phi-3 মডেল ব্যবহার করে Microsoft.ML.OnnxRuntime দিয়ে Q&A .NET উদাহরণ](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Semantic Kernel এবং Phi-3 সহ কনসোল চ্যাট .NET অ্যাপ](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI ইনফারেন্স SDK কোড ভিত্তিক নমুনা - - Phi-4 নমুনা - - [📓] [Phi-4-multimodal ব্যবহার করে প্রকল্প কোড তৈরি](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 নমুনা - - [Microsoft Phi-3 পরিবার ব্যবহার করে আপনার নিজস্ব Visual Studio Code GitHub Copilot চ্যাট তৈরি করুন](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub মডেল সহ Phi-3.5 দিয়ে আপনার নিজস্ব Visual Studio Code চ্যাট কপাইলট এজেন্ট তৈরি করুন](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Phi-4 লোকাল ONNX মডেলের সাথে চ্যাট .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Phi-4 ONNX ব্যবহার করে Sementic Kernel-এর মাধ্যমে চ্যাট .NET কনসোল অ্যাপ](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - ফাই-৩ / ৩.৫ নমুনা + - [Local Chatbot in the browser using Phi3, ONNX Runtime Web and WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Multi Model - Interactive Phi-3-mini and OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Building a wrapper and using Phi-3 with MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Model Optimization - How to optimize Phi-3-min model for ONNX Runtime Web with Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 App with Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 Multi Model AI Powered Notes App Sample](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Fine-tune and Integrate custom Phi-3 models with Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Fine-tune and Integrate custom Phi-3 models with Prompt flow in Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Evaluate the Fine-tuned Phi-3 / Phi-3.5 Model in Microsoft Foundry Focusing on Microsoft's Responsible AI Principles](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct language prediction sample (Chinese/English)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Using Windows GPU to create Prompt flow solution with Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Using Microsoft Phi-3.5 tflite to create Android app](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Q&A .NET Example using local ONNX Phi-3 model using the Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Console chat .NET app with Semantic Kernel and Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + + - আজুর এআই ইনফারেন্স SDK কোড ভিত্তিক নমুনা + - ফাই-৪ নমুনা + - [📓] [Phi-4-multimodal ব্যবহার করে প্রজেক্ট কোড জেনারেট করা](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - ফাই-৩ / ৩.৫ নমুনা + - [Microsoft Phi-3 Family দিয়ে নিজের Visual Studio Code GitHub Copilot Chat তৈরি করুন](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub মডেল দিয়ে Phi-3.5 ব্যবহার করে আপনার নিজস্ব Visual Studio Code Chat Copilot Agent তৈরি করুন](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - উন্নত যুক্তি নমুনা - - Phi-4 নমুনা + - ফাই-৪ নমুনা - [📓] [Phi-4-mini-reasoning বা Phi-4-reasoning নমুনা](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [মাইক্রোসফট Olive দিয়ে Phi-4-mini-reasoning ফাইন-টিউনিং](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Apple MLX দিয়ে Phi-4-mini-reasoning ফাইন-টিউনিং](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [GitHub মডেল সহ Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Microsoft Foundry মডেল সহ Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [📓] [Microsoft Olive ব্যবহার করে Phi-4-mini-reasoning ফাইন-টিউনিং](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLX ব্যবহার করে Phi-4-mini-reasoning ফাইন-টিউনিং](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [GitHub মডেল দিয়ে Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Microsoft Foundry মডেল দিয়ে Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - ডেমো - - [Phi-4-mini ডেমো যা Hugging Face স্পেসে হোস্ট করা হয়েছে](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal ডেমো যা Hugging Face স্পেসে হোস্ট করা হয়েছে](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-mini ডেমো যা Hugging Face Spaces এ হোস্ট করা হয়েছে](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal ডেমো যা Hugginge Face Spaces এ হোস্ট করা হয়েছে](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - ভিশন নমুনা - - Phi-4 নমুনা + - ফাই-৪ নমুনা - [📓] [ছবি পড়তে এবং কোড তৈরি করতে Phi-4-multimodal ব্যবহার করুন](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 নমুনা + - ফাই-৩ / ৩.৫ নমুনা - [📓][Phi-3-vision-ছবি টেক্সট থেকে টেক্সট](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP এম্বেডিং](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [ডেমো: Phi-3 রিসাইক্লিং](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - ভিজ্যুয়াল ভাষা সহকারী - Phi3-Vision এবং OpenVINO সহ](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3-vision - ভিজ্যুয়াল ল্যাঙ্গুয়েজ অ্যাসিস্ট্যান্ট - Phi3-Vision এবং OpenVINO সহ](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 ভিশন Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 ভিশন OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - [📓][Phi-3.5 ভিশন মাল্টি-ফ্রেম বা মাল্টি-ইমেজ নমুনা](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Microsoft.ML.OnnxRuntime .NET ব্যবহার করে Phi-3 ভিশন লোকাল ONNX মডেল](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Phi-3 ভিশন লোকাল ONNX মডেল Microsoft.ML.OnnxRuntime .NET ব্যবহার করে](../../md/04.HOL/dotnet/src/LabsPhi303) - [মেনু ভিত্তিক Phi-3 ভিশন লোকাল ONNX মডেল Microsoft.ML.OnnxRuntime .NET ব্যবহার করে](../../md/04.HOL/dotnet/src/LabsPhi304) - যুক্তি-ভিশন নমুনা - Phi-4-Reasoning-Vision-15B - - [📓] [Jaywalking সনাক্ত করতে Phi-4-Reasoning-Vision-15B ব্যবহার করা হচ্ছে](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [গণিত করতে Phi-4-Reasoning-Vision-15B ব্যবহার](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [UI সনাক্ত করতে Phi-4-Reasoning-Vision-15B ব্যবহার](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B ব্যবহার করে jaywalking সনাক্তকরণ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B ব্যবহার করে গণিত](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B ব্যবহার করে UI সনাক্তকরণ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - গণিত নমুনা - - Phi-4-Mini-Flash-Reasoning-Instruct নমুনা [Phi-4-Mini-Flash-Reasoning-Instruct এর সাথে গণিত ডেমো](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-Mini-Flash-Reasoning-Instruct নমুনা [Phi-4-Mini-Flash-Reasoning-Instruct দিয়ে গণিত ডেমো](./md/02.Application/09.Math/MathDemo.ipynb) - অডিও নমুনা - - Phi-4 নমুনা + - ফাই-৪ নমুনা - [📓] [Phi-4-multimodal ব্যবহার করে অডিও ট্রান্সক্রিপ্ট বের করা](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Phi-4-multimodal অডিও নমুনা](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Phi-4-multimodal স্পিচ ট্রান্সলেশন নমুনা](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET কনসোল অ্যাপ্লিকেশন Phi-4-multimodal অডিও ব্যবহার করে একটি অডিও ফাইল বিশ্লেষণ এবং ট্রান্সক্রিপ্ট তৈরি](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [.NET কনসোল অ্যাপ্লিকেশন যা Phi-4-multimodal অডিও ব্যবহার করে একটি অডিও ফাইল বিশ্লেষণ এবং ট্রান্সক্রিপ্ট তৈরি করে](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE নমুনা - - Phi-3 / 3.5 নমুনা - - [📓] [Phi-3.5 মিশ্র বিশেষজ্ঞ মডেল (MoEs) সামাজিক মিডিয়া নমুনা](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI সার্চ এবং LlamaIndex দিয়ে রিট্রিভাল-অগমেন্টেড জেনারেশন (RAG) পাইপলাইন তৈরি করা](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - ফাই-৩ / ৩.৫ নমুনা + - [📓] [Phi-3.5 মিক্সচার অফ এক্সপার্টস (MoEs) সোশ্যাল মিডিয়া নমুনা](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search এবং LlamaIndex দিয়ে রিট্রিভাল-অগমেন্টেড জেনারেশন (RAG) পাইপলাইন তৈরি](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - ফাংশন কলিং নমুনা - - Phi-4 নমুনা 🆕 + - ফাই-৪ নমুনা 🆕 - [📓] [Phi-4-mini দিয়ে ফাংশন কলিং ব্যবহার](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - [📓] [Phi-4-mini দিয়ে মাল্টি-এজেন্ট তৈরি করতে ফাংশন কলিং ব্যবহার](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Ollama দিয়ে ফাংশন কলিং ব্যবহার](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [ONNX সহ ফাংশন কলিং ব্যবহার](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - মাল্টিমডাল মিক্সিং নমুনা - - Phi-4 নমুনা 🆕 - - [📓] [প্রযুক্তি সাংবাদিক হিসেবে Phi-4-multimodal ব্যবহার](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET কনসোল অ্যাপ্লিকেশন Phi-4-multimodal ব্যবহার করে ছবি বিশ্লেষণ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Phi ফাইন-টিউনিং নমুনা - - [ফাইন-টিউনিং সিনারিও](./md/03.FineTuning/FineTuning_Scenarios.md) + - [📓] [ONNX দিয়ে ফাংশন কলিং ব্যবহার](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - মাল্টিমোডাল মিক্সিং নমুনা + - ফাই-৪ নমুনা 🆕 + - [📓] [Phi-4-multimodal কে প্রযুক্তি সাংবাদিক হিসেবে ব্যবহার করা](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET কনসোল অ্যাপ্লিকেশন যা Phi-4-multimodal ব্যবহার করে ছবি বিশ্লেষণ করে](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- ফাইন-টিউনিং ফাই নমুনা + - [ফাইন-টিউনিং পরিস্থিতি](./md/03.FineTuning/FineTuning_Scenarios.md) - [ফাইন-টিউনিং বনাম RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Phi-3 কে একটি শিল্প বিশেষজ্ঞ তৈরি করা ফাইন-টিউনিং](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [AI Toolkit for VS Code দিয়ে Phi-3 ফাইন-টিউনিং](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Azure Machine Learning Service দিয়ে Phi-3 ফাইন-টিউনিং](./md/03.FineTuning/Introduce_AzureML.md) - - [Lora দিয়ে Phi-3 ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora দিয়ে Phi-3 ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_Qlora.md) - - [Microsoft Foundry দিয়ে Phi-3 ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Azure ML CLI/SDK দিয়ে Phi-3 ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_MLSDK.md) + - [ফাইন-টিউনিং: ফাই-৩ কে একটি ইন্ডাস্ট্রি বিশেষজ্ঞ হতে দিন](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Visual Studio Code এর জন্য AI Toolkit দিয়ে ফাই-৩ ফাইন-টিউনিং](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure Machine Learning Service দিয়ে ফাই-৩ ফাইন-টিউনিং](./md/03.FineTuning/Introduce_AzureML.md) + - [Lora দিয়ে ফাই-৩ ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_Lora.md) + - [QLora দিয়ে ফাই-৩ ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_Qlora.md) + - [Microsoft Foundry দিয়ে ফাই-৩ ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Azure ML CLI/SDK দিয়ে ফাই-৩ ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_MLSDK.md) - [Microsoft Olive দিয়ে ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Microsoft Olive Hands-On Lab দিয়ে ফাইন-টিউনিং](./md/03.FineTuning/olive-lab/readme.md) - - [Weights and Bias দিয়ে Phi-3-vision ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Apple MLX Framework দিয়ে Phi-3 ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-vision ফাইন-টিউনিং (সরকারী সমর্থন)](./md/03.FineTuning/FineTuning_Vision.md) - - [কাইতো AKS, Azure Containers (আধিকারিক সমর্থন) সহ Phi-3 ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_Kaito.md) + - [Weights and Bias দিয়ে ফাই-৩-ভিশন ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [অ্যাপল MLX ফ্রেমওয়ার্ক দিয়ে Phi-3 ফাইন-টিউনিং](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision ফাইন-টিউনিং (অফিসিয়াল সমর্থন)](./md/03.FineTuning/FineTuning_Vision.md) + - [Kaito AKS , Azure Containers দিয়ে Phi-3 ফাইন-টিউনিং (অফিসিয়াল সমর্থন)](./md/03.FineTuning/FineTuning_Kaito.md) - [Phi-3 এবং 3.5 Vision ফাইন-টিউনিং](https://github.com/2U1/Phi3-Vision-Finetune) -- হাতে কলমে ল্যাব - - [কাটিং-এজ মডেলগুলি অন্বেষণ: LLMs, SLMs, স্থানীয় উন্নয়ন এবং আরও অনেক কিছু](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [এনএলপি সম্ভাবনা উন্মোচন: Microsoft Olive এর সাথে ফাইন-টিউনিং](https://github.com/azure/Ignite_FineTuning_workshop) +- হাতেকলমে ল্যাব + - [কাটিং-এজ মডেলস অন্বেষণ: LLMs, SLMs, লোকাল ডেভেলপমেন্ট এবং আরও অনেক কিছু](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP সম্ভাবনা উন্মুক্ত: Microsoft Olive দিয়ে ফাইন-টিউনিং](https://github.com/azure/Ignite_FineTuning_workshop) -- একাডেমিক গবেষণাপত্র ও প্রকাশনা +- একাডেমিক গবেষণা পত্র ও প্রকাশনা - [Textbooks Are All You Need II: phi-1.5 প্রযুক্তিগত প্রতিবেদন](https://arxiv.org/abs/2309.05463) - [Phi-3 প্রযুক্তিগত প্রতিবেদন: আপনার ফোনে স্থানীয়ভাবে একটি অত্যন্ত সক্ষম ভাষা মডেল](https://arxiv.org/abs/2404.14219) - [Phi-4 প্রযুক্তিগত প্রতিবেদন](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini প্রযুক্তিগত প্রতিবেদন: Mixture-of-LoRAs এর মাধ্যমে কমপ্যাক্ট কিন্তু শক্তিশালী মাল্টিমোডাল ভাষা মডেল](https://arxiv.org/abs/2503.01743) - - [ইন-ভেহিকেল ফাংশন-কলিংয়ের জন্য ছোট ভাষা মডেল অপ্টিমাইজেশন](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) মাল্টিপল-চয়েস প্রশ্নোত্তরের জন্য PHI-3 ফাইন-টিউনিং: পদ্ধতি, ফলাফল, এবং চ্যালেঞ্জ](https://arxiv.org/abs/2501.01588) + - [Phi-4-Mini প্রযুক্তিগত প্রতিবেদন: Mixture-of-LoRAs এর মাধ্যমে কমপ্যাক্ট কিন্তু শক্তিশালী মাল্টিমোডাল ভাষা মডেলস](https://arxiv.org/abs/2503.01743) + - [গাড়ির মধ্যে কার্যকলাপ কল করার জন্য ছোট ভাষা মডেলস অপ্টিমাইজেশন](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) একাধিক-চয়েস প্রশ্নোত্তর পরীক্ষার জন্য PHI-3 ফাইন-টিউনিং: পদ্ধতি, ফলাফল এবং চ্যালেঞ্জ](https://arxiv.org/abs/2501.01588) - [Phi-4-রিজনিং প্রযুক্তিগত প্রতিবেদন](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Phi-4-মিনি-রিজনিং প্রযুক্তিগত প্রতিবেদন](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Phi মডেল ব্যবহারের পদ্ধতি +## Phi মডেল ব্যবহারের উপায় ### Microsoft Foundry তে Phi -আপনি শিখতে পারবেন কীভাবে Microsoft Phi ব্যবহার করতে হয় এবং বিভিন্ন হার্ডওয়্যার ডিভাইসে E2E সমাধান তৈরি করতে হয়। Phi নিজের জন্য ব্যবহার করার জন্য, মডেলগুলির সাথে খেলে শুরু করুন এবং আপনার পরিস্থিতির জন্য Phi কাস্টমাইজ করুন [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) ব্যবহার করে। আপনি আরও জানতে পারেন [Microsoft Foundry](./md/02.QuickStart/AzureAIFoundry_QuickStart.md) থেকে। +আপনি শিখতে পারেন কীভাবে Microsoft Phi ব্যবহার করবেন এবং কীভাবে আপনার বিভিন্ন হার্ডওয়্যার ডিভাইসগুলিতে E2E সমাধান তৈরি করবেন। Phi নিজেই অনুভব করতে, শুরু করুন মডেলগুলি নিয়ে খেলা করে এবং আপনার ক্ষেত্রে Phi কাস্টমাইজ করে [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) ব্যবহার করে আপনি আরও জানতে পারবেন Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**প্লেযোগ্রাউন্ড** -প্রতিটি মডেলের জন্য একটি নির্দিষ্ট প্লেযোগ্রাউন্ড আছে মডেল পরীক্ষার জন্য [Azure AI Playground](https://aka.ms/try-phi3)। +**প্লেিহাগ্রাউন্ড** +প্রতিটি মডেলের জন্য একটি নির্দিষ্ট প্লেিহাগ্রাউন্ড রয়েছে মডেল পরীক্ষা করার জন্য [Azure AI Playground](https://aka.ms/try-phi3)। -### GitHub মডেলে Phi +### GitHub মডেলসে Phi -আপনি শিখতে পারবেন কীভাবে Microsoft Phi ব্যবহার করতে হয় এবং বিভিন্ন হার্ডওয়্যার ডিভাইসে E2E সমাধান তৈরি করতে হয়। Phi নিজে ব্যবহার করার জন্য, মডেলের সাথে খেলে শুরু করুন এবং আপনার পরিস্থিতির জন্য Phi কাস্টমাইজ করুন [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ব্যবহার করে। আরো জানার জন্য [GitHub Model Catalog](./md/02.QuickStart/GitHubModel_QuickStart.md) শুরু করুন। +আপনি শিখতে পারেন কীভাবে Microsoft Phi ব্যবহার করবেন এবং কীভাবে আপনার বিভিন্ন হার্ডওয়্যার ডিভাইসগুলিতে E2E সমাধান তৈরি করবেন। Phi নিজেই অনুভব করতে, শুরু করুন মডেল নিয়ে খেলা করে এবং আপনার ক্ষেত্রে Phi কাস্টমাইজ করে [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ব্যবহার করে আপনি আরও জানতে পারবেন Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**প্লেযোগ্রাউন্ড** -প্রতিটি মডেলের জন্য একটি নির্দিষ্ট [প্লেযোগ্রাউন্ড মডেল পরীক্ষা করার জন্য](./md/02.QuickStart/GitHubModel_QuickStart.md) আছে। +**প্লেিহাগ্রাউন্ড** +প্রতিটি মডেলের জন্য একটি নির্দিষ্ট [প্লেিহাগ্রাউন্ড যা মডেল পরীক্ষা করে](/md/02.QuickStart/GitHubModel_QuickStart.md)। -### Hugging Face তে Phi +### Hugging Face এ Phi -আপনি মডেলটি [Hugging Face](https://huggingface.co/microsoft) এও পেতে পারেন। +আপনি মডেলটি [Hugging Face](https://huggingface.co/microsoft) তেও পেতে পারেন -**প্লেযোগ্রাউন্ড** -[Hugging Chat প্লেযোগ্রাউন্ড](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**প্লেিহাগ্রাউন্ড** + [Hugging Chat প্লেিহাগ্রাউন্ড](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 অন্যান্য কোর্সসমূহ + ## 🎒 অন্যান্য কোর্স -আমাদের দল অন্যান্য কোর্সসমূহ তৈরি করে! দেখুন: +আমাদের দল অন্যান্য কোর্স তৈরি করে! দেখুন: -### LangChain -[![শুরুদের জন্য LangChain4j](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![শুরুদের জন্য LangChain.js](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![শুরুদের জন্য LangChain](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +### LangChain +[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain for Beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / এজেন্টসমূহ -[![শুরুদের জন্য AZD](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![শুরুদের জন্য Edge AI](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![শুরুদের জন্য MCP](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![শুরুদের জন্য AI এজেন্ট](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agents +[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### জেনেরেটিভ AI সিরিজ -[![শুরুদের জন্য জেনেরেটিভ AI](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![জেনেরেটিভ AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![জেনেরেটিভ AI (জাভা)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![জেনেরেটিভ AI (জাভাস্ক্রিপ্ট)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### Generative AI সিরিজ +[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### মূল শিক্ষা -[![শুরুদের জন্য এমএল](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![শুরুদের জন্য ডাটা সায়েন্স](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![শুরুদের জন্য AI](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![শুরুকদের জন্য সাইবারসিকিউরিটি](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![শুরুদের জন্য ওয়েব ডেভেলপমেন্ট](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![শুরুদের জন্য IoT](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![শুরুদের জন্য XR ডেভেলপমেন্ট](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) + +### মূল শিক্ষা +[![শিখতে পারেন মেশিন লার্নিং (ML) beginners জন্য](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![ডেটা সায়েন্স for beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![কৃত্রিম বুদ্ধিমত্তা (AI) for beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![সাইবারসিকিউরিটি for beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![ওয়েব ডেভ for beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![আইওটি for beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![এক্সআর ডেভেলপমেন্ট for beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### কপিলট সিরিজ -[![কপিলট ফর AI পেয়ারড প্রোগ্রামিং](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![কপিলট ফর C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![কপিলট অ্যাডভেঞ্চার](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + +### কোপাইলট সিরিজ +[![AI Paired Programming জন্য কোপাইলট](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![C#/.NET জন্য কোপাইলট](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![কোপাইলট অ্যাডভেঞ্চার](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## দায়িত্বশীল AI -Microsoft আমাদের গ্রাহকদের সাহায্য করতে প্রতিজ্ঞাবদ্ধ যাতে তারা আমাদের AI পণ্যগুলি দায়িত্বসহ ব্যবহার করতে পারে, আমাদের শেখাগুলি শেয়ার করে, এবং ট্রান্সপারেন্সি নোটস এবং প্রভাব মূল্যায়ন মত টুলের মাধ্যমে বিশ্বাসভিত্তিক অংশীদারিত্ব তৈরি করে। এই অনেক রিসোর্স [https://aka.ms/RAI](https://aka.ms/RAI) তে পাওয়া যায়। -Microsoft এর দায়িত্বশীল AI পদ্ধতি আমাদের AI নীতির উপর ভিত্তি করে যা হ'ল ন্যায্যতা, নির্ভরযোগ্যতা ও নিরাপত্তা, গোপনীয়তা ও নিরাপত্তা, অন্তর্ভুক্তি, স্বচ্ছতা এবং জবাবদিহিতা। +Microsoft আমাদের গ্রাহকদের দায়িত্বশীলভাবে আমাদের AI পণ্যগুলি ব্যবহার করতে সাহায্য করার জন্য প্রতিশ্রুতিবদ্ধ, আমাদের শেখাগুলি শেয়ার করে এবং ট্রান্সপারেন্সি নোট এবং প্রভাব মূল্যায়নের মতো সরঞ্জামগুলির মাধ্যমে বিশ্বাস-ভিত্তিক অংশীদারিত্ব তৈরি করে। এই অনেকগুলি সম্পদ আপনি [https://aka.ms/RAI](https://aka.ms/RAI) থেকে পেতে পারেন। +Microsoft এর দায়িত্বশীল AI এর পদ্ধতি আমাদের AI ন্যায্যতা, নির্ভরযোগ্যতা ও নিরাপত্তা, গোপনীয়তা ও নিরাপত্তা, অন্তর্ভুক্তি, স্বচ্ছতা এবং জবাবদিহিতার নীতিমালার উপর ভিত্তি করে। + +বড়-পরিমাণের প্রাকৃতিক ভাষা, ছবি এবং বক্তৃতা মডেল - এই নমুনায় ব্যবহৃতগুলির মতো - সম্ভবত এমন আচরণ করতে পারে যা ন্যায়সঙ্গত নয়, অপ্রত্যাশিত, অথবা অপমানজনক, যা ক্ষতি করতে পারে। দয়া করে [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) পরামর্শ করুন বিপদ এবং সীমাবদ্ধতা সম্পর্কে অবগত থাকার জন্য। + -বড় আকারের প্রাকৃতিক ভাষা, ছবি, এবং স্পীচ মডেলসমূহ - যেমন এই স্যাম্পলটিতে ব্যবহৃত - সম্ভাব্যভাবে এমনভাবে আচরণ করতে পারে যা অন্যায়, অবিচার্য, বা আপত্তিকর হতে পারে, যা ক্ষতি করতে পারে। অনুগ্রহ করে ঝুঁকি ও সীমাবদ্ধতা সম্পর্কে অবহিত হতে [Azure OpenAI পরিষেবার ট্রান্সপারেন্সি নোট](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) দেখুন। -এই ঝুঁকিগুলি মোকাবেলার জন্য সুপারিশকৃত পদ্ধতি হল আপনার আর্কিটেকচারে একটি সেফটি সিস্টেম অন্তর্ভুক্ত করা যা ক্ষতিকর আচরণ সনাক্ত এবং প্রতিরোধ করতে পারে। [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) একটি স্বতন্ত্র সুরক্ষা স্তর প্রদান করে, যা অ্যাপ্লিকেশন এবং পরিষেবাগুলিতে ক্ষতিকর ব্যবহারকারী-উত্পাদিত এবং AI-উত্পাদিত বিষয়বস্তু সনাক্ত করতে সক্ষম। Azure AI Content Safety-তে টেক্সট এবং ইমেজ API রয়েছে যা আপনাকে ক্ষতিকর উপাদান সনাক্ত করতে দেয়। Microsoft Foundry-র মধ্যে, Content Safety সেবা আপনাকে বিভিন্ন মোডালিটির ক্ষতিকর বিষয়বস্তু সনাক্ত করার জন্য উদাহরণ কোড দেখতে, অনুসন্ধান করতে এবং চেষ্টা করতে দেয়। নিম্নলিখিত [quickstart ডকুমেন্টেশন](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) আপনাকে সেবা ব্যবহার করে অনুরোধ করার প্রক্রিয়াতে পথনির্দেশ দেয়। +এই ঝুঁকিগুলো কমানোর জন্য প্রস্তাবিত পদ্ধতি হল আপনার স্থাপত্যে একটি সুরক্ষা ব্যবস্থা অন্তর্ভুক্ত করা যা ক্ষতিকারক আচরণ সনাক্ত করতে এবং প্রতিরোধ করতে পারে। [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) একটি স্বতন্ত্র সুরক্ষা স্তর প্রদান করে, যা অ্যাপ্লিকেশন এবং সার্ভিসে ক্ষতিকারক ব্যবহারকারী-উৎপন্ন এবং AI-উৎপন্ন সামগ্রী সনাক্ত করতে সক্ষম। Azure AI Content Safety-তে টেক্সট এবং ইমেজ API রয়েছে যা আপনাকে ক্ষতিকর উপাদান সনাক্ত করতে দেয়। Microsoft Foundry-এর মধ্যে, Content Safety সার্ভিস আপনাকে বিভিন্ন মোডালিটিজ জুড়ে ক্ষতিকর সামগ্রী সনাক্ত করার জন্য নমুনা কোড দেখা, অন্বেষণ এবং চেষ্টা করার সুযোগ দেয়। নিম্নলিখিত [quickstart ডকুমেন্টেশন](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) আপনাকে সার্ভিসে অনুরোধ পাঠানো সম্পর্কে গাইড করে। -আরেকটি দিক বিবেচনা করার বিষয় হল সর্বমোট অ্যাপ্লিকেশন পারফরম্যান্স। মাল্টি-মোডাল এবং মাল্টি-মডেল অ্যাপ্লিকেশনগুলিতে, আমরা পারফরম্যান্স বোঝাতে চাই যে সিস্টেমটি আপনি এবং আপনার ব্যবহারকারীরা প্রত্যাশা করেন তেমন কাজ করে, যার মধ্যে ক্ষতিকর ফলাফল তৈরি না করাও অন্তর্ভুক্ত। আপনার সর্বমোট অ্যাপ্লিকেশনের পারফরম্যান্স মূল্যায়ন করা গুরুত্বপূর্ণ যা [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) ব্যবহার করে করা যায়। আপনার আরও ক্ষমতা আছে [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) তৈরি এবং মূল্যায়ন করার। +বিবেচনা করার আরেকটি দিক হল সামগ্রিক অ্যাপ্লিকেশন কর্মক্ষমতা। মাল্টি-মোডাল এবং মাল্টি-মডেল অ্যাপ্লিকেশনগুলির ক্ষেত্রে, কর্মক্ষমতা অর্থ হয় যে সিস্টেমটি আপনি এবং আপনার ব্যবহারকারীরা যেমন প্রত্যাশা করেন তেমন কাজ করে, যার মধ্যে ক্ষতিকর আউটপুট তৈরি না করাও অন্তর্ভুক্ত। আপনার সামগ্রিক অ্যাপ্লিকেশনের কর্মক্ষমতা মূল্যায়ন করা গুরুত্বপূর্ণ [পারফরম্যান্স এবং কোয়ালিটি এবং ঝুঁকি এবং সুরক্ষা ইভ্যালুয়েটর](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) ব্যবহার করে। আপনার কাছে [কাস্টম ইভ্যালুয়েটর](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) তৈরি এবং মূল্যায়নের সুযোগও রয়েছে। -আপনি আপনার উন্নয়ন পরিবেশে [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ব্যবহার করে আপনার AI অ্যাপ্লিকেশন মূল্যায়ন করতে পারেন। একটি পরীক্ষা ডেটাসেট বা একটি লক্ষ্য প্রদান করলে, আপনার জেনারেটিভ AI অ্যাপ্লিকেশনের জেনারেশনগুলি Built-in evaluators বা আপনার পছন্দসই custom evaluators দ্বারা количеত্মকভাবে মাপা হয়। Azure AI Evaluation SDK দিয়ে আপনার সিস্টেম মূল্যায়ন শুরু করতে, আপনি [quickstart গাইড](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) অনুসরণ করতে পারেন। একবার আপনি একটি মূল্যায়ন রান সম্পাদন করলে, আপনি [Microsoft Foundry-তে ফলাফলগুলো ভিজ্যুয়ালাইজ করতে পারেন](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)। +আপনি আপনার ডেভেলপমেন্ট পরিবেশে [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ব্যবহার করে আপনার AI অ্যাপ্লিকেশন মূল্যায়ন করতে পারেন। একটি টেস্ট ডেটাসেট বা একটি লক্ষ্য দেওয়া হলে, আপনার জেনারেটিভ AI অ্যাপ্লিকেশনের উৎপাদনগুলি built-in ইভ্যালুয়েটর বা আপনার পছন্দের কাস্টম ইভ্যালুয়েটর দিয়ে পরিমাণগতভাবে পরিমাপ করা হয়। আপনার সিস্টেম মূল্যায়ন করার জন্য Azure AI Evaluation SDK দিয়ে শুরু করতে, আপনি [quickstart গাইড](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) অনুসরণ করতে পারেন। একবার আপনি একটি ইভ্যালুয়েশন রন সম্পন্ন করলে, আপনি [Microsoft Foundry-তে ফলাফল দেখতে পারেন](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)। ## ট্রেডমার্ক -এই প্রকল্পে প্রকল্প, পণ্য বা পরিষেবাগুলোর ট্রেডমার্ক বা লোগো থাকতে পারে। Microsoft ট্রেডমার্ক বা লোগোগুলোর স্বীকৃত ব্যবহার [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) অনুসরণ করতে হবে এবং তার বাধ্যতামূলক। Microsoft ট্রেডমার্ক বা লোগোর সংশোধিত সংস্করণে ব্যবহার বিভ্রান্তি সৃষ্টি বা Microsoft স্পন্সরশিপের ধারণা দিতে পারবে না। তৃতীয় পক্ষের ট্রেডমার্ক বা লোগোর যে কোনো ব্যবহার ঐ তৃতীয় পক্ষের নীতিমালা অনুসারে হবে। +এই প্রকল্পে প্রোজেক্ট, পণ্য বা সার্ভিসের ট্রেডমার্ক বা লোগো থাকতে পারে। Microsoft ট্রেডমার্ক বা লোগোর অনুমোদিত ব্যবহার অবশ্যই [Microsoft এর ট্রেডমার্ক ও ব্র্যান্ড গাইডলাইনস](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) অনুসরণ করতে হবে। +পরিবর্তিত সংস্করণে Microsoft ট্রেডমার্ক বা লোগোর ব্যবহার বিভ্রান্তি সৃষ্টিকারী বা Microsoft স্পন্সরশিপ নির্দেশকারী হতে পারবে না। তৃতীয় পক্ষের ট্রেডমার্ক বা লোগোর কোনো ব্যবহার সেই তৃতীয় পক্ষের নীতিমালা অনুসরণ করবে। -## সাহায্য নেওয়া +## সাহায্য পাওয়া -আপনি আটকে গেলে বা AI অ্যাপ নির্মাণ সম্পর্কে কোনো প্রশ্ন থাকলে, যোগ দিন: +আপনি যদি আটকে যান বা AI অ্যাপ তৈরির বিষয়ে কোনো প্রশ্ন থাকে, তাহলে যোগ দিন: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -আপনার যদি পণ্য প্রতিক্রিয়া বা গঠনকালে কোনো ত্রুটি থাকে, তাহলে যান: +আপনি যদি পণ্যের প্রতিক্রিয়া বা ত্রুটি সম্পর্কে জানাতে চান, তাহলে যান: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) @@ -319,5 +325,5 @@ Microsoft এর দায়িত্বশীল AI পদ্ধতি আম **অস্বীকৃতি**: -এই নথিটি AI অনুবাদ পরিষেবা [Co-op Translator](https://github.com/Azure/co-op-translator) ব্যবহার করে অনূদিত হয়েছে। আমরা সঠিকতার জন্য চেষ্টা করি, তবে দয়া করে মনে রাখবেন যে স্বয়ংক্রিয় অনুবাদে ভুল বা অসঙ্গতি থাকতে পারে। মূল নথিটি তার স্বাভাবিক ভাষায়ই কর্তৃত্বপ্রাপ্ত উৎস হিসেবে বিবেচনা করা উচিত। গুরুত্বপূর্ণ তথ্যের জন্য পেশাদার মানব অনুবাদ সুপারিশ করা হয়। এই অনুবাদের ব্যবহারে সৃষ্ট কোনও ভুল বোঝাবুঝি বা ভুল ব্যাখ্যার জন্য আমরা দায়ী নই। +এই নথিটি AI অনুবাদ পরিষেবা [Co-op Translator](https://github.com/Azure/co-op-translator) ব্যবহার করে অনূদিত হয়েছে। যদিও আমরা শুদ্ধতার জন্য চেষ্টা করি, অনুগ্রহ করে মনে রাখবেন যে স্বয়ংক্রিয় অনুবাদে ত্রুটি বা অসঙ্গতি থাকতে পারে। মূল নথিটি তার স্বভাষায় কর্তৃত্বপূর্ণ উৎস হিসেবে বিবেচিত হওয়া উচিত। গুরুত্বপূর্ণ তথ্যের জন্য পেশাদার মানব অনুবাদ সুপারিশ করা হয়। এই অনুবাদের ব্যবহারে প্রয়োজনীয় ভুল বোঝাবুঝি বা ভুল ব্যাখ্যার জন্য আমরা দায়বদ্ধ নই। \ No newline at end of file diff --git a/translations/bn/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/bn/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..37f0b4dec --- /dev/null +++ b/translations/bn/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 মাইক্রোসফ্ট phi-4-mini & SQLite FTS5 সহ হাইব্রিড RAG (জিরো-ক্লাউড SLM)\n", + "\n", + "> **লেখক:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **ফোকাস:** স্মল ল্যাঙ্গুয়েজ মডেলস (SLMs), SQLite FTS5 BM25, ডেন্স এমবেডিংস, রিসিপ্রোক্যাল র‍্যাঙ্ক ফিউশন (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. প্রেরণা: লোকাল SLMs এ কীওয়ার্ড রিকল সমস্যাটি\n", + "শুধুমাত্র ডেন্স ভেক্টর এমবেডিংসের ওপর নির্ভরশীল স্ট্যান্ডার্ড RAG আর্কিটেকচারগুলি প্রায়শই সঠিক সংখ্যাসূচক টোকেন (যেমন, `2,340,000 TL`, চুক্তি কোড, অ্যাকাউন্ট নম্বর) উদ্ধার করতে ব্যর্থ হয়। \n", + "অপরদিকে, স্পার্স লেক্সিকাল সার্চ (BM25) সেমান্টিক স্পর্শক শব্দ এবং প্যারাফ্রেজড প্রশ্নগুলি মিস করে।\n", + "\n", + "এই কুকবুকটি প্রদর্শন করে কীভাবে একটি **উচ্চ-গতির, ইন-মেমরি হাইব্রিড রিট্রিভাল ইঞ্জিন** বাস্তবায়ন করা যায় যা সংমিশ্রণ করে:\n", + "1. **ডেন্স ভেক্টরস** (কসাইন সাদৃশ্য)\n", + "2. **স্পার্স লেক্সিকাল সার্চ** (SQLite FTS5 BM25)\n", + "3. **রিসিপ্রোক্যাল র‍্যাঙ্ক ফিউশন (RRF, $k=60$)**\n", + "4. মাইক্রোসফ্ট `phi-4-mini` সহ **গ্রাউন্ডেড সাইটেশন জেনারেশন (`[1]`, `[2]`)**\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ ২. দ্বৈত SQLite স্কিমা (ঘন ভেক্টর + ভার্চুয়াল FTS5 BM25 টেবিল)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 ৩. উদাহরণ গ্রহণ ও কার্যকরী বেন্চমার্ক\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. মাইক্রোসফ্ট phi-4-mini এর জন্য গ্রাউন্ডেড প্রম্পট ফরমুলেশন\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**অস্বীকৃতি**:\nএই নথিটি AI অনুবাদ পরিষেবা [Co-op Translator](https://github.com/Azure/co-op-translator) ব্যবহার করে অনূদিত হয়েছে। যদিও আমরা শুদ্ধতার জন্য চেষ্টা করি, অনুগ্রহ করে মনে রাখবেন যে স্বয়ংক্রিয় অনুবাদে ত্রুটি বা অসঙ্গতি থাকতে পারে। মূল নথিটি তার স্বভাষায় কর্তৃত্বপূর্ণ উৎস হিসেবে বিবেচিত হওয়া উচিত। গুরুত্বপূর্ণ তথ্যের জন্য পেশাদার মানব অনুবাদ সুপারিশ করা হয়। এই অনুবাদের ব্যবহারে প্রয়োজনীয় ভুল বোঝাবুঝি বা ভুল ব্যাখ্যার জন্য আমরা দায়বদ্ধ নই।\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/cs/.co-op-translator.json b/translations/cs/.co-op-translator.json index 399d6d80e..3f43af57c 100644 --- a/translations/cs/.co-op-translator.json +++ b/translations/cs/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "cs" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:27:26+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:41:15+00:00", "source_file": "README.md", "language_code": "cs" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "cs" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:42:13+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "cs" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:58+00:00", diff --git a/translations/cs/README.md b/translations/cs/README.md index feecd3c6f..c4a031853 100644 --- a/translations/cs/README.md +++ b/translations/cs/README.md @@ -1,42 +1,42 @@ # Phi Cookbook: Praktické příklady s modely Phi od Microsoftu -[![Otevřít a používat ukázky v GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Otevřít v Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Otevřít a použít ukázky v GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Otevřít ve vývojových kontejnerech](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub přispěvatelé](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub issues](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub pull-requests](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![Přispěvatelé na GitHubu](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![Problémy na GitHubu](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![Pull requesty na GitHubu](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub sledující](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub forků](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub hvězd](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub watchers](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub stars](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi je série open source AI modelů vyvinutých společností Microsoft. +Phi je řada open source AI modelů vyvinutých společností Microsoft. -Phi je v současnosti nejvýkonnější a nejefektivnější malý jazykový model (SLM) s velmi dobrými benchmarky v mnoha jazycích, v oblasti uvažování, generování textu/chatů, kódování, obrazu, zvuku a dalších scénářů. +Phi je v současné době nejsilnější a nejefektivnější malý jazykový model (SLM) s velmi dobrými výsledky v oblasti vícejazyčnosti, uvažování, generování textu/chatů, kódování, obrázků, audia a dalších scénářů. -Phi můžete nasadit do cloudu nebo na edge zařízení a snadno tak vytvořit generativní AI aplikace s omezenou výpočetní kapacitou. +Phi můžete nasadit na cloud nebo na edge zařízení a snadno si postavit generativní AI aplikace s omezeným výpočetním výkonem. -Postupujte podle těchto kroků, abyste mohli začít používat tyto zdroje: -1. **Vytvořte Fork Repozitáře**: Klikněte na [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Klonujte Repozitář**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Připojte se ke komunitě Microsoft AI na Discordu a potkejte experty a další vývojáře**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +Postupujte podle těchto kroků, jak začít používat tyto zdroje: +1. **Vytvořte fork repozitáře**: Klikněte na [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Klonujte repozitář**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Přidejte se do Microsoft AI Discord komunity a potkejte odborníky a další vývojáře**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/cs/cover.eb18d1b9605d754b.webp) ### 🌐 Podpora více jazyků -#### Podporováno přes GitHub Action (automatické a vždy aktuální) +#### Podporováno pomocí GitHub Action (Automatické a vždy aktuální) -[Arabština](../ar/README.md) | [Bengálština](../bn/README.md) | [Bulharština](../bg/README.md) | [Barmština (Myanmar)](../my/README.md) | [Čínština (zjednodušená)](../zh-CN/README.md) | [Čínština (tradiční, Hong Kong)](../zh-HK/README.md) | [Čínština (tradiční, Macau)](../zh-MO/README.md) | [Čínština (tradiční, Taiwan)](../zh-TW/README.md) | [Chorvatština](../hr/README.md) | [Čeština](./README.md) | [Dánština](../da/README.md) | [Nizozemština](../nl/README.md) | [Estonština](../et/README.md) | [Finština](../fi/README.md) | [Francouzština](../fr/README.md) | [Němčina](../de/README.md) | [Řečtina](../el/README.md) | [Hebrejština](../he/README.md) | [Hindština](../hi/README.md) | [Maďarština](../hu/README.md) | [Indonéština](../id/README.md) | [Italština](../it/README.md) | [Japonština](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korejština](../ko/README.md) | [Litevština](../lt/README.md) | [Malajština](../ms/README.md) | [Malajalámština](../ml/README.md) | [Maráthština](../mr/README.md) | [Nepálština](../ne/README.md) | [Nigerijský pidžin](../pcm/README.md) | [Norština](../no/README.md) | [Perština (Fársí)](../fa/README.md) | [Polština](../pl/README.md) | [Portugalština (Brazílie)](../pt-BR/README.md) | [Portugalština (Portugalsko)](../pt-PT/README.md) | [Paňdžábština (Gurmukhi)](../pa/README.md) | [Rumunština](../ro/README.md) | [Ruština](../ru/README.md) | [Srbština (cyrilice)](../sr/README.md) | [Slovenština](../sk/README.md) | [Slovinština](../sl/README.md) | [Španělština](../es/README.md) | [Svahilština](../sw/README.md) | [Švédština](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugština](../te/README.md) | [Thajština](../th/README.md) | [Turečtina](../tr/README.md) | [Ukrajinština](../uk/README.md) | [Urdština](../ur/README.md) | [Vietnamština](../vi/README.md) +[Arabština](../ar/README.md) | [Bengálština](../bn/README.md) | [Bulharština](../bg/README.md) | [Barmsky (Myanmar)](../my/README.md) | [Čínština (zjednodušená)](../zh-CN/README.md) | [Čínština (tradiční, Hongkong)](../zh-HK/README.md) | [Čínština (tradiční, Macao)](../zh-MO/README.md) | [Čínština (tradiční, Taiwan)](../zh-TW/README.md) | [Chorvatština](../hr/README.md) | [Čeština](./README.md) | [Dánština](../da/README.md) | [Nizozemština](../nl/README.md) | [Estonština](../et/README.md) | [Finština](../fi/README.md) | [Francouzština](../fr/README.md) | [Němčina](../de/README.md) | [Řečtina](../el/README.md) | [Hebrejština](../he/README.md) | [Hindština](../hi/README.md) | [Maďarština](../hu/README.md) | [Indonéština](../id/README.md) | [Italština](../it/README.md) | [Japonština](../ja/README.md) | [Kannadština](../kn/README.md) | [Khmerština](../km/README.md) | [Korejština](../ko/README.md) | [Litevština](../lt/README.md) | [Malajština](../ms/README.md) | [Malajalámština](../ml/README.md) | [Maráthština](../mr/README.md) | [Nepálština](../ne/README.md) | [Nigerijská pidžin](../pcm/README.md) | [Norština](../no/README.md) | [Perština (Farsi)](../fa/README.md) | [Polština](../pl/README.md) | [Portugalština (Brazílie)](../pt-BR/README.md) | [Portugalština (Portugalsko)](../pt-PT/README.md) | [Paňdžábština (Gurmukhi)](../pa/README.md) | [Rumunština](../ro/README.md) | [Ruština](../ru/README.md) | [Srbština (cyrilice)](../sr/README.md) | [Slovenština](../sk/README.md) | [Slovinština](../sl/README.md) | [Španělština](../es/README.md) | [Svahilština](../sw/README.md) | [Švédština](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamilština](../ta/README.md) | [Telugština](../te/README.md) | [Thajština](../th/README.md) | [Turečtina](../tr/README.md) | [Ukrajinština](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamština](../vi/README.md) > **Raději klonovat lokálně?** > -> Tento repozitář obsahuje více než 50 překladů do jazyků, což výrazně zvětšuje velikost ke stažení. Chcete-li klonovat bez překladů, použijte sparse checkout: +> Tento repozitář obsahuje více než 50 jazykových překladů, což výrazně zvětšuje velikost ke stažení. Pro klonování bez překladů použijte sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,202 +52,205 @@ Postupujte podle těchto kroků, abyste mohli začít používat tyto zdroje: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Toto vám poskytne vše potřebné pro dokončení kurzu s mnohem rychlejším stažením. +> To vám poskytne vše, co potřebujete ke splnění kurzu, s mnohem rychlejším stahováním. ## Obsah - Úvod - [Vítejte v rodině Phi](./md/01.Introduction/01/01.PhiFamily.md) - - [Nastavení prostředí](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Pochopení klíčových technologií](./md/01.Introduction/01/01.Understandingtech.md) + - [Nastavení vašeho prostředí](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Porozumění klíčovým technologiím](./md/01.Introduction/01/01.Understandingtech.md) - [Bezpečnost AI pro modely Phi](./md/01.Introduction/01/01.AISafety.md) - [Podpora hardwaru Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Modely Phi & dostupnost napříč platformami](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Použití Guidance-ai a Phi](./md/01.Introduction/01/01.Guidance.md) + - [Modely Phi a dostupnost na různých platformách](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Používání Guidance-ai a Phi](./md/01.Introduction/01/01.Guidance.md) - [Modely na GitHub Marketplace](https://github.com/marketplace/models) - [Katalog AI modelů Azure](https://ai.azure.com) -- Inference Phi v různých prostředích +- Inferenční Phi v různém prostředí - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [GitHub Modely](./md/01.Introduction/02/02.GitHubModel.md) + - [GitHub modely](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry Model Catalog](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Inference Phi Family +- Inferenční Phi rodiny - [Inference Phi na iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Inference Phi na Android](./md/01.Introduction/03/Android_Inference.md) - - [Inference Phi na Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Inference Phi na Androidu](./md/01.Introduction/03/Android_Inference.md) + - [Inference Phi na Jetsonu](./md/01.Introduction/03/Jetson_Inference.md) - [Inference Phi na AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Inference Phi s Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Inference Phi s Apple MLX Frameworkem](./md/01.Introduction/03/MLX_Inference.md) - [Inference Phi na lokálním serveru](./md/01.Introduction/03/Local_Server_Inference.md) - [Inference Phi na vzdáleném serveru pomocí AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Inference Phi s Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inference Phi—Vision lokálně](./md/01.Introduction/03/Vision_Inference.md) + - [Inference Phi s Rustem](./md/01.Introduction/03/Rust_Inference.md) + - [Inference Phi–Vision lokálně](./md/01.Introduction/03/Vision_Inference.md) - [Inference Phi s Kaito AKS, Azure kontejnery (oficiální podpora)](./md/01.Introduction/03/Kaito_Inference.md) -- [Kvantisování Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) - - [Kvantisování Phi-3.5 / 4 pomocí llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Kvantisování Phi-3.5 / 4 pomocí Generative AI extensions pro onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Kvantisování Phi-3.5 / 4 pomocí Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Kvantisování Phi-3.5 / 4 pomocí Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) +- [Kvantifikace Phi rodiny](./md/01.Introduction/04/QuantifyingPhi.md) + - [Kvantizace Phi-3.5 / 4 pomocí llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Kvantizace Phi-3.5 / 4 pomocí Generative AI rozšíření pro onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Kvantizace Phi-3.5 / 4 pomocí Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Kvantizace Phi-3.5 / 4 pomocí Apple MLX Frameworku](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Hodnocení Phi - - [Odpovědná AI](./md/01.Introduction/05/ResponsibleAI.md) + - [Zodpovědná AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry pro hodnocení](./md/01.Introduction/05/AIFoundry.md) - [Použití Promptflow pro hodnocení](./md/01.Introduction/05/Promptflow.md) - RAG s Azure AI Search - [Jak používat Phi-4-mini a Phi-4-multimodal (RAG) s Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud lokální hybridní RAG s SQLite FTS5 a phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Ukázky vývoje aplikací Phi - - Textové a chatové aplikace - - Ukázky Phi-4 - - [📓] [Chat s modelem Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Chat s místním modelem Phi-4 ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Chat .NET konzolová aplikace s Phi-4 ONNX pomocí Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Ukázky Phi-3 / 3.5 - - [Lokální chatbot v prohlížeči pomocí Phi3, ONNX Runtime Web a WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) +- Příklady vývoje aplikací Phi + - Textové a chat aplikace + - Phi-4 příklady + - [📓] [Chat s Phi-4-mini ONNX modelem](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Chat s Phi-4 lokálním ONNX modelem v .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Console aplikace v .NET pro chat s Phi-4 ONNX pomocí Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Vzorky Phi-3 / 3.5 + - [Místní chatbot v prohlížeči používající Phi3, ONNX Runtime Web a WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Více modelů - Interaktivní Phi-3-mini a OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [Multi Model - Interaktivní Phi-3-mini a OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Vytváření wrapperu a použití Phi-3 s MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [Optimalizace modelu - Jak optimalizovat model Phi-3-min pro ONNX Runtime Web s Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [WinUI3 aplikace s Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Ukázka aplikace WinUI3 s více modely poháněnou AI](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Doladění a integrace vlastních modelů Phi-3 s Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Doladění a integrace vlastních modelů Phi-3 s Prompt flow v Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Vyhodnocení doladěného modelu Phi-3 / Phi-3.5 v Microsoft Foundry zaměřené na zásady odpovědné AI Microsoftu](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Ukázka jazykového předpovídání Phi-3.5-mini-instruct (čínština/angličtina)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + -[WinUI3 Multi Model AI Powered Notes App Sample](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Doladění a integrace vlastních modelů Phi-3 pomocí Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Doladění a integrace vlastních modelů Phi-3 pomocí Prompt flow v Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Vyhodnocení doladěného modelu Phi-3 / Phi-3.5 v Microsoft Foundry se zaměřením na zásady odpovědné AI Microsoftu](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Vzorek jazykové predikce Phi-3.5-mini-instruct (čínština/angličtina)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Použití Windows GPU k vytvoření řešení Prompt flow s Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Použití Microsoft Phi-3.5 tflite k vytvoření Android aplikace](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Ukázka Q&A .NET s použitím lokálního ONNX modelu Phi-3 pomocí Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Konzolová chat .NET aplikace se Semantic Kernel a Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Q&A .NET příklad používající lokální ONNX model Phi-3 pomocí Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Konzolová chatovací .NET aplikace s Semantic Kernel a Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Ukázky kódu SDK Azure AI Inference založené na kódu - - Ukázky Phi-4 + - Kódové ukázky Azure AI Inference SDK + - Vzorky Phi-4 - [📓] [Generování kódu projektu pomocí Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Ukázky Phi-3 / 3.5 - - [Vytvořte si vlastní Visual Studio Code GitHub Copilot Chat s Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Vytvořte si svého vlastního Visual Studio Code Chat Copilot Agenta s Phi-3.5 pomocí GitHub modelů](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - Vzorky Phi-3 / 3.5 + - [Vytvořte si vlastní Visual Studio Code GitHub Copilot Chat s Microsoft Phi-3 rodinou](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Vytvořte si vlastního chatovacího agenta Visual Studio Code Copilot s Phi-3.5 pomocí GitHub modelů](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Ukázky pokročilého uvažování - - Ukázky Phi-4 - - [📓] [Ukázky Phi-4-mini-reasoning nebo Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Doladění Phi-4-mini-reasoning s Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Doladění Phi-4-mini-reasoning s Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - Vzorky Phi-4 + - [📓] [Vzorky Phi-4-mini-reasoning nebo Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Doladění Phi-4-mini-reasoning pomocí Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Doladění Phi-4-mini-reasoning pomocí Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-reasoning s GitHub modely](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-reasoning s Microsoft Foundry modely](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Ukázky + - Demo - [Phi-4-mini demo hostované na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demo hostované na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Ukázky vize - - Ukázky Phi-4 - - [📓] [Použití Phi-4-multimodal pro čtení obrázků a generování kódu](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Ukázky Phi-3 / 3.5 - - [📓][Phi-3-vision - Převedení textu z obrázku na text](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-4-multimodal demo hostované na Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Ukázky Vision + - Vzorky Phi-4 + - [📓] [Použití Phi-4-multimodal ke čtení obrázků a generování kódu](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Vzorky Phi-3 / 3.5 + - [📓][Phi-3-vision-Image text to text](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 recyklace](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Asistent vizuálního jazyka - s Phi3-Vision a OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [DEMO: Phi-3 Recycling](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - vizuální jazykový asistent - s Phi3-Vision a OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision ukázka s více rámci nebo více obrázky](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision lokální ONNX model pomocí Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menu založené Phi-3 Vision lokální ONNX model pomocí Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Phi-3.5 Vision vzorek více snímků nebo více obrázků](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision místní ONNX model používající Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Menu založený Phi-3 Vision místní ONNX model používající Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - Ukázky Reasoning-Vision - - Phi-4-Reasoning-Vision-15B + - Phi-4-Reasoning-Vision-15B - [📓] [Použití Phi-4-Reasoning-Vision-15B k detekci přecházení mimo přechod](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Použití Phi-4-Reasoning-Vision-15B na matematiku](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Použití Phi-4-Reasoning-Vision-15B k matematice](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - [📓] [Použití Phi-4-Reasoning-Vision-15B k detekci UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Ukázky matematiky - - Ukázky Phi-4-Mini-Flash-Reasoning-Instruct [Demo matematiky s Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Vzorky Phi-4-Mini-Flash-Reasoning-Instruct [Ukázka matematiky s Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - Ukázky audia - - Ukázky Phi-4 - - [📓] [Extrahování přepisů audia pomocí Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Ukázka audia Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Ukázka překladu řeči Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET konzolová aplikace s využitím Phi-4-multimodal k analýze audio souboru a generování přepisu](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - Ukázky zvuku + - Vzorky Phi-4 + - [📓] [Extrahování přepisů zvuku pomocí Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Vzorek zvuku Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Vzorek překladu řeči Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET konzolová aplikace používající Phi-4-multimodal ke zpracování zvukového souboru a generování přepisu](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - Ukázky MOE - - Ukázky Phi-3 / 3.5 - - [📓] [Ukázka Phi-3.5 Mixture of Experts Models (MoEs) pro sociální média](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Vytváření pipeline s Retrieval-Augmented Generation (RAG) s NVIDIA NIM Phi-3 MOE, Azure AI Search a LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - Vzorky Phi-3 / 3.5 + - [📓] [Vzorek sociálních médií s Phi-3.5 Mixture of Experts Models (MoEs)](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Vytváření Retrieval-Augmented Generation (RAG) pipeline s NVIDIA NIM Phi-3 MOE, Azure AI Search a LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - Ukázky volání funkcí - - Ukázky Phi-4 🆕 + - Vzorky Phi-4 🆕 - [📓] [Použití volání funkcí s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - [📓] [Použití volání funkcí k vytvoření multi-agentů s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Použití volání funkcí s Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [Použití volání funkcí s ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Ukázky míchání multimodálních dat - - Ukázky Phi-4 🆕 + - Ukázky multimodálního míchání + - Vzorky Phi-4 🆕 - [📓] [Použití Phi-4-multimodal jako technologického novináře](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [.NET konzolová aplikace používající Phi-4-multimodal k analýze obrázků](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Doladění Phi ukázek +- Doladění Phi vzorků - [Scénáře doladění](./md/03.FineTuning/FineTuning_Scenarios.md) - [Doladění vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Doladění: Nechte Phi-3 stát se průmyslovým expertem](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Doladění Phi-3 pomocí AI Toolkit pro VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Doladění Phi-3 pomocí Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Doladění Phi-3 pomocí Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Doladění Phi-3 pomocí QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Doladění Phi-3 pomocí Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Doladění Phi-3 pomocí Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Doladění Nechte Phi-3 stát se průmyslovým expertem](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Doladění Phi-3 s AI Toolkit pro VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Doladění Phi-3 s Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Doladění Phi-3 s Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Doladění Phi-3 s QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Doladění Phi-3 s Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Doladění Phi-3 s Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - [Doladění s Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Doladění s praktickou laboratoří Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) + - [Doladění s Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - [Doladění Phi-3-vision s Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Doladění Phi-3 s Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Doladění Phi-3-vision (oficiální podpora)](./md/03.FineTuning/FineTuning_Vision.md) - - [Ladění Phi-3 s Kaito AKS, Azure kontejnery (oficiální podpora)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Ladění Phi-3 a 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + + - [Jemné doladění Phi-3 s Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Jemné doladění Phi-3-vision (oficiální podpora)](./md/03.FineTuning/FineTuning_Vision.md) + - [Jemné doladění Phi-3 s Kaito AKS, Azure kontejnery (oficiální podpora)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Jemné doladění Phi-3 a 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Praktická laboratoř - - [Prozkoumávání nejmodernějších modelů: LLM, SLM, lokální vývoj a další](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Odemknutí potenciálu NLP: Ladění s Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Objevování nejmodernějších modelů: LLM, SLM, lokální vývoj a další](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Odemknutí potenciálu NLP: Jemné doladění s Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Akademické výzkumné práce a publikace +- Akademické výzkumné články a publikace - [Textbooks Are All You Need II: technická zpráva phi-1.5](https://arxiv.org/abs/2309.05463) - - [Technická zpráva Phi-3: Vysoce schopný jazykový model lokálně na vašem telefonu](https://arxiv.org/abs/2404.14219) + - [Technická zpráva Phi-3: Vysoce schopný jazykový model přímo ve vašem telefonu](https://arxiv.org/abs/2404.14219) - [Technická zpráva Phi-4](https://arxiv.org/abs/2412.08905) - - [Technická zpráva Phi-4-Mini: Kompaktní, ale výkonné multimodální jazykové modely pomocí směsi LoRA](https://arxiv.org/abs/2503.01743) + - [Technická zpráva Phi-4-Mini: Kompaktní, ale výkonné multimodální jazykové modely pomocí Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - [Optimalizace malých jazykových modelů pro volání funkcí ve vozidle](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Ladění PHI-3 pro odpovídání na otázky s více možnostmi: Metodologie, výsledky a výzvy](https://arxiv.org/abs/2501.01588) + - [(WhyPHI) Jemné doladění PHI-3 pro odpovídání na otázky s více možnostmi: Metodologie, výsledky a výzvy](https://arxiv.org/abs/2501.01588) - [Technická zpráva Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Technická zpráva Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Používání modelů Phi +## Používání Phi modelů ### Phi na Microsoft Foundry -Můžete se naučit, jak používat Microsoft Phi a jak stavět end-to-end řešení na různých hardwarových zařízeních. Pro vyzkoušení Phi začněte hraním s modely a přizpůsobováním Phi pro vaše scénáře pomocí [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Více se dozvíte v návodu Začínáme s [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md). +Můžete se naučit, jak používat Microsoft Phi a jak stavět E2E řešení na různých hardwarech. Chcete-li Phi sami vyzkoušet, začněte hraním s modely a přizpůsobováním Phi svým scénářům pomocí [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Více se dozvíte v Začínáme s [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Hřiště** -Každý model má své vlastní hřiště pro testování modelu na [Azure AI Playground](https://aka.ms/try-phi3). +**Hřiště (Playground)** +Každý model má své vlastní hřiště pro testování modelu [Azure AI Playground](https://aka.ms/try-phi3). ### Phi na GitHub modelech -Můžete se naučit, jak používat Microsoft Phi a jak stavět end-to-end řešení na různých hardwarových zařízeních. Pro vyzkoušení Phi začněte hraním s modelem a přizpůsobováním Phi pro vaše scénáře pomocí [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Více se dozvíte v návodu Začínáme s [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md). +Můžete se naučit, jak používat Microsoft Phi a jak stavět E2E řešení na různých hardwarech. Chcete-li Phi sami vyzkoušet, začněte hrát s modelem a přizpůsobovat Phi svým scénářům pomocí [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Více se dozvíte v Začínáme s [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Hřiště** +**Hřiště (Playground)** Každý model má vyhrazené [hřiště pro testování modelu](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi na Hugging Face -Model také najdete na [Hugging Face](https://huggingface.co/microsoft). +Model lze také najít na [Hugging Face](https://huggingface.co/microsoft) -**Hřiště** - [Hřiště Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Hřiště (Playground)** + [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) ## 🎒 Další kurzy -Náš tým připravuje i další kurzy! Podívejte se na: +Náš tým produkuje další kurzy! Podívejte se: ### LangChain @@ -256,15 +259,15 @@ Náš tým připravuje i další kurzy! Podívejte se na: [![LangChain pro začátečníky](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agenti +### Azure / Edge / MCP / Agents [![AZD pro začátečníky](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI pro začátečníky](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP pro začátečníky](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agenti pro začátečníky](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents pro začátečníky](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Série Generativní AI +### Séria Generativní AI [![Generativní AI pro začátečníky](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Generativní AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) [![Generativní AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) @@ -274,7 +277,7 @@ Náš tým připravuje i další kurzy! Podívejte se na: ### Základní učení [![Strojové učení pro začátečníky](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Datová věda pro začátečníky](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science pro začátečníky](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI pro začátečníky](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![Kybernetická bezpečnost pro začátečníky](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Webový vývoj pro začátečníky](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) @@ -283,31 +286,34 @@ Náš tým připravuje i další kurzy! Podívejte se na: --- -### Série Copilot +### Séria Copilot [![Copilot pro AI párové programování](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot pro C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Odpovědná AI +## Odpovědná AI + +Microsoft se zavazuje pomáhat našim zákazníkům používat naše AI produkty odpovědně, sdílet naše poznatky a budovat partnerství založená na důvěře prostřednictvím nástrojů jako jsou Transparency Notes a Impact Assessments. Mnoho z těchto zdrojů najdete na [https://aka.ms/RAI](https://aka.ms/RAI). +Přístup Microsoftu k odpovědné AI je založen na našich AI principech: spravedlnost, spolehlivost a bezpečnost, soukromí a zabezpečení, inkluzivita, transparentnost a odpovědnost. + +Velké modely pro přirozený jazyk, obrázky a řeč - jako ty použité v tomto příkladu - se mohou potenciálně chovat nespravedlivě, nespolehlivě nebo urážlivě, což může způsobit škodu. Prosím nahlédněte do [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), abyste byli informováni o rizicích a omezeních. -Microsoft se zavazuje pomáhat našim zákazníkům používat naše AI produkty odpovědně, sdílet naše poznatky a budovat partnerství založená na důvěře pomocí nástrojů jako jsou Poznámky o transparentnosti a Hodnocení dopadů. Mnoho z těchto zdrojů naleznete na [https://aka.ms/RAI](https://aka.ms/RAI). -Přístup Microsoftu k odpovědné AI je založen na našich principech AI: spravedlnost, spolehlivost a bezpečnost, ochrana soukromí a bezpečnost, inkluzivita, transparentnost a odpovědnost. -Velké jazykové, obrazové a hlasové modely na rozsáhlém měřítku - jako jsou ty použité v tomto příkladu - mohou potenciálně vykazovat chování, které je nespravedlivé, nespolehlivé nebo urážlivé, což může vést ke škodám. Prosím prostudujte [Poznámku o transparentnosti služby Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), abyste byli informováni o rizicích a omezeních. -Doporučený přístup k zmírnění těchto rizik je zahrnout do vaší architektury bezpečnostní systém, který dokáže detekovat a zabránit škodlivému chování. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) poskytuje nezávislou vrstvu ochrany, schopnou detekovat škodlivý obsah generovaný uživateli i AI v aplikacích a službách. Azure AI Content Safety zahrnuje textové a obrazové API, které vám umožní detekovat škodlivý materiál. V rámci Microsoft Foundry služba Content Safety umožňuje prohlížet, zkoumat a vyzkoušet si ukázkový kód pro detekci škodlivého obsahu napříč různými modalitami. Následující [dokumentace rychlého startu](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) vás provede procesem odesílání požadavků na službu. +Doporučeným přístupem k zmírnění těchto rizik je zahrnout do vaší architektury bezpečnostní systém, který dokáže detekovat a zabránit škodlivému chování. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) poskytuje nezávislou vrstvu ochrany, která dokáže detekovat škodlivý obsah generovaný uživateli i AI v aplikacích a službách. Azure AI Content Safety zahrnuje API pro text a obrázky, které vám umožňují detekovat škodlivý materiál. V Microsoft Foundry služba Content Safety umožňuje prohlížet, zkoumat a vyzkoušet ukázkové kódy pro detekci škodlivého obsahu napříč různými modalitami. Následující [dokumentace quickstart](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) vás provede vytvářením požadavků na službu. -Dalším aspektem, který je třeba vzít v úvahu, je celkový výkon aplikace. V případě multimodálních a multimodelových aplikací chápeme výkon tak, že systém funguje podle očekávání vás i vašich uživatelů, včetně nevytváření škodlivých výstupů. Je důležité zhodnotit výkon vaší celkové aplikace pomocí [hodnotitelů výkonnosti, kvality, rizika a bezpečnosti](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Také máte možnost vytvářet a hodnotit pomocí [vlastních hodnotitelů](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Dalším aspektem, který je třeba vzít v úvahu, je celkový výkon aplikace. U multimodálních a multimodelových aplikací považujeme za výkon, že systém funguje podle očekávání vás a vašich uživatelů, včetně nevytváření škodlivých výstupů. Je důležité posoudit výkon vaší celkové aplikace pomocí [hodnotitelů Výkon a Kvalita a Riziko a Bezpečnost](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Máte také možnost vytvářet a hodnotit pomocí [vlastních hodnotitelů](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Svou AI aplikaci můžete hodnotit ve vašem vývojovém prostředí pomocí [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Na základě testovacího datasetu nebo cíle jsou generace vaší generativní AI aplikace kvantitativně měřeny vestavěnými hodnotiteli nebo vámi zvolenými vlastními hodnotiteli. Chcete-li začít s Azure AI Evaluation SDK pro vyhodnocení vašeho systému, můžete se řídit [průvodcem rychlého startu](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Po spuštění hodnocení můžete [vizualizovat výsledky v Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Můžete hodnotit vaši AI aplikaci ve vašem vývojovém prostředí pomocí [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Na základě testovacího datasetu nebo cíle jsou generace vaší generativní AI aplikace kvantitativně měřeny pomocí vestavěných hodnotitelů nebo vlastních hodnotitelů podle vašeho výběru. Pro začátek s Azure AI Evaluation SDK, abyste mohli hodnotit váš systém, můžete sledovat [průvodce rychlým startem](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Po spuštění hodnocení můžete [vizualizovat výsledky v Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Ochranné známky -Tento projekt může obsahovat ochranné známky nebo loga projektů, produktů či služeb. Autorizované použití ochranných známek nebo log Microsoftu podléhá a musí dodržovat [Pravidla použití ochranných známek a značek Microsoftu](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). Použití ochranných známek nebo log Microsoftu v upravených verzích tohoto projektu nesmí způsobit záměnu nebo naznačovat podporu Microsoftem. Jakékoli použití ochranných známek nebo log třetích stran podléhá pravidlům těchto třetích stran. +Tento projekt může obsahovat ochranné známky nebo loga projektů, produktů či služeb. Autorizované používání ochranných známek nebo log Microsoft podléhá a musí se řídit [Pokyny Microsoftu pro ochranné známky a značky](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Používání ochranných známek nebo log Microsoft v upravených verzích tohoto projektu nesmí způsobovat záměnu ani naznačovat sponzorství ze strany Microsoftu. Jakékoliv použití ochranných známek nebo log třetích stran podléhá zásadám těchto třetích stran. ## Získání pomoci -Pokud narazíte na potíže nebo máte dotazy ohledně vývoje AI aplikací, přidejte se k: +Pokud se zaseknete nebo máte jakékoliv dotazy týkající se tvorby AI aplikací, připojte se: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) @@ -318,6 +324,6 @@ Pokud máte zpětnou vazbu k produktu nebo narazíte na chyby při vývoji, nav --- -**Prohlášení o vyloučení odpovědnosti**: -Tento dokument byl přeložen pomocí AI překladatelské služby [Co-op Translator](https://github.com/Azure/co-op-translator). I když se snažíme o přesnost, mějte prosím na paměti, že automatizované překlady mohou obsahovat chyby nebo nepřesnosti. Původní dokument v jeho rodném jazyce by měl být považován za autoritativní zdroj. Pro kritické informace se doporučuje profesionální lidský překlad. Nejsme odpovědní za jakékoliv nedorozumění nebo špatné výklady vzniklé z použití tohoto překladu. +**Prohlášení o omezení odpovědnosti**: +Tento dokument byl přeložen pomocí AI překladatelské služby [Co-op Translator](https://github.com/Azure/co-op-translator). Přestože usilujeme o co největší přesnost, mějte prosím na paměti, že automatizované překlady mohou obsahovat chyby nebo nepřesnosti. Originální dokument v jeho mateřském jazyce by měl být považován za autoritativní zdroj. Pro kritické informace se doporučuje profesionální lidský překlad. Nejsme odpovědní za jakékoli nedorozumění nebo nesprávné interpretace vzniklé použitím tohoto překladu. \ No newline at end of file diff --git a/translations/cs/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/cs/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..3427e14ef --- /dev/null +++ b/translations/cs/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybridní RAG s Microsoft phi-4-mini a SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Autor:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Zaměření:** Malé jazykové modely (SLM), SQLite FTS5 BM25, husté embeddings, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivace: Dilema vyhledávání klíčových slov v lokálních SLM\n", + "Standardní RAG architektury, které spoléhají pouze na husté vektorové embeddings, často selhávají při vyhledávání přesných číselných tokenů (např. `2,340,000 TL`, kódy smluv, čísla účtů). \n", + "Naopak řídké lexikální vyhledávání (BM25) postrádá semantické synonyma a přeformulované otázky.\n", + "\n", + "Tento návod ukazuje, jak implementovat **vysokorychlostní hybridní vyhledávací engine v paměti**, který kombinuje:\n", + "1. **Husté vektory** (kosinová podobnost)\n", + "2. **Řídké lexikální vyhledávání** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Generování podložených citací (`[1]`, `[2]`)** s Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Duální schéma SQLite (husté vektory + virtuální tabulka FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Vzorkování příjmu a měření výkonu spuštění\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Formulace založených promptů pro Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Prohlášení o omezení odpovědnosti**:\nTento dokument byl přeložen pomocí AI překladatelské služby [Co-op Translator](https://github.com/Azure/co-op-translator). Přestože usilujeme o co největší přesnost, mějte prosím na paměti, že automatizované překlady mohou obsahovat chyby nebo nepřesnosti. Originální dokument v jeho mateřském jazyce by měl být považován za autoritativní zdroj. Pro kritické informace se doporučuje profesionální lidský překlad. Nejsme odpovědní za jakékoli nedorozumění nebo nesprávné interpretace vzniklé použitím tohoto překladu.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/da/.co-op-translator.json b/translations/da/.co-op-translator.json index fdfe8e00f..11f1cdab7 100644 --- a/translations/da/.co-op-translator.json +++ b/translations/da/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "da" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:05:09+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:13:31+00:00", "source_file": "README.md", "language_code": "da" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "da" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:19:08+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "da" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:11+00:00", diff --git a/translations/da/README.md b/translations/da/README.md index 86edee17a..66d5d7525 100644 --- a/translations/da/README.md +++ b/translations/da/README.md @@ -1,42 +1,42 @@ -# Phi Cookbook: Praktiske Eksempler med Microsofts Phi-modeller +# Phi Cookbook: Praktiske Eksempler med Microsofts Phi-Modeller -[![Open and use the samples in GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Open in Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Åbn og brug eksemplerne i GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Åbn i Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Åbn&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub contributors](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub bidragydere](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub issues](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub pull-requests](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Velkomne](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub watchers](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub overvågere](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub stars](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub stjerner](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi er en række open source AI-modeller udviklet af Microsoft. +Phi er en serie af open source AI-modeller udviklet af Microsoft. -Phi er i øjeblikket den mest kraftfulde og omkostningseffektive småsprogmodel (SLM) med meget gode benchmarks inden for flere sprog, ræsonnement, tekst/chat-generering, kodning, billeder, lyd og andre scenarier. +Phi er i øjeblikket den mest kraftfulde og omkostningseffektive lille sprogmodel (SLM), med meget gode benchmarks inden for flersprog, ræsonnement, tekst/chat generering, kodning, billeder, lyd og andre scenarier. -Du kan implementere Phi i skyen eller på edge-enheder, og du kan nemt bygge generative AI-applikationer med begrænset regnekraft. +Du kan implementere Phi til skyen eller til edge-enheder, og du kan nemt bygge generative AI-applikationer med begrænset computerkraft. Følg disse trin for at komme i gang med at bruge disse ressourcer: -1. **Fork det repository**: Klik [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Klon repositoryet**: `git clone https://github.com/microsoft/PhiCookBook.git` +1. **Fork Repositoryet**: Klik [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Klon Repositoryet**: `git clone https://github.com/microsoft/PhiCookBook.git` 3. [**Deltag i Microsoft AI Discord-fællesskabet og mød eksperter og andre udviklere**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/da/cover.eb18d1b9605d754b.webp) -### 🌐 Understøttelse af flere sprog +### 🌐 Flersprogsunderstøttelse -#### Understøttet via GitHub Action (Automatiseret & Altid Opdateret) +#### Understøttet via GitHub Action (Automatiseret og Altid Opdateret) -[Arabisk](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarsk](../bg/README.md) | [Burmesisk (Myanmar)](../my/README.md) | [Kinesisk (Forenklet)](../zh-CN/README.md) | [Kinesisk (Traditionel, Hong Kong)](../zh-HK/README.md) | [Kinesisk (Traditionel, Macau)](../zh-MO/README.md) | [Kinesisk (Traditionel, Taiwan)](../zh-TW/README.md) | [Kroatisk](../hr/README.md) | [Tjekkisk](../cs/README.md) | [Dansk](./README.md) | [Hollandsk](../nl/README.md) | [Estisk](../et/README.md) | [Finsk](../fi/README.md) | [Fransk](../fr/README.md) | [Tysk](../de/README.md) | [Græsk](../el/README.md) | [Hebraisk](../he/README.md) | [Hindi](../hi/README.md) | [Ungarsk](../hu/README.md) | [Indonesisk](../id/README.md) | [Italiensk](../it/README.md) | [Japansk](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreansk](../ko/README.md) | [Litauisk](../lt/README.md) | [Malayisk](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigeriansk Pidgin](../pcm/README.md) | [Norsk](../no/README.md) | [Persisk (Farsi)](../fa/README.md) | [Polsk](../pl/README.md) | [Portugisisk (Brasilien)](../pt-BR/README.md) | [Portugisisk (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumænsk](../ro/README.md) | [Russisk](../ru/README.md) | [Serbisk (Kyrillisk)](../sr/README.md) | [Slovakisk](../sk/README.md) | [Slovensk](../sl/README.md) | [Spansk](../es/README.md) | [Swahili](../sw/README.md) | [Svensk](../sv/README.md) | [Tagalog (Filippinsk)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Tyrkisk](../tr/README.md) | [Ukrainsk](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamesisk](../vi/README.md) +[Arabisk](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarsk](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Kinesisk (Forenklet)](../zh-CN/README.md) | [Kinesisk (Traditionel, Hong Kong)](../zh-HK/README.md) | [Kinesisk (Traditionel, Macau)](../zh-MO/README.md) | [Kinesisk (Traditionel, Taiwan)](../zh-TW/README.md) | [Kroatisk](../hr/README.md) | [Tjekkisk](../cs/README.md) | [Dansk](./README.md) | [Hollandsk](../nl/README.md) | [Estisk](../et/README.md) | [Finsk](../fi/README.md) | [Fransk](../fr/README.md) | [Tysk](../de/README.md) | [Græsk](../el/README.md) | [Hebraisk](../he/README.md) | [Hindi](../hi/README.md) | [Ungarsk](../hu/README.md) | [Indonesisk](../id/README.md) | [Italiensk](../it/README.md) | [Japansk](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreansk](../ko/README.md) | [Litauisk](../lt/README.md) | [Malayisk](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalesisk](../ne/README.md) | [Nigeriansk Pidgin](../pcm/README.md) | [Norsk](../no/README.md) | [Persisk (Farsi)](../fa/README.md) | [Polsk](../pl/README.md) | [Portugisisk (Brasilien)](../pt-BR/README.md) | [Portugisisk (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumænsk](../ro/README.md) | [Russisk](../ru/README.md) | [Serbisk (Cyrillisk)](../sr/README.md) | [Slovakisk](../sk/README.md) | [Slovensk](../sl/README.md) | [Spansk](../es/README.md) | [Swahili](../sw/README.md) | [Svensk](../sv/README.md) | [Tagalog (Filippinsk)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Tyrkisk](../tr/README.md) | [Ukrainsk](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamesisk](../vi/README.md) > **Foretrækker du at klone lokalt?** > -> Dette repository inkluderer 50+ sprogoversættelser, hvilket væsentligt øger downloadstørrelsen. For at klone uden oversættelser, brug sparse checkout: +> Dette repository indeholder 50+ sprogoversættelser, hvilket betydeligt øger downloadstørrelsen. For at klone uden oversættelser, brug sparsom udtjekning: > > **Bash / macOS / Linux:** > ```bash @@ -60,141 +60,143 @@ Følg disse trin for at komme i gang med at bruge disse ressourcer: - Introduktion - [Velkommen til Phi-familien](./md/01.Introduction/01/01.PhiFamily.md) - [Opsætning af dit miljø](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Forståelse af nøgle teknologier](./md/01.Introduction/01/01.Understandingtech.md) + - [Forståelse af nøgleteknologier](./md/01.Introduction/01/01.Understandingtech.md) - [AI-sikkerhed for Phi-modeller](./md/01.Introduction/01/01.AISafety.md) - - [Phi hardwareunderstøttelse](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Phi hardware support](./md/01.Introduction/01/01.Hardwaresupport.md) - [Phi modeller & tilgængelighed på tværs af platforme](./md/01.Introduction/01/01.Edgeandcloud.md) - [Brug af Guidance-ai og Phi](./md/01.Introduction/01/01.Guidance.md) - [GitHub Marketplace Modeller](https://github.com/marketplace/models) - - [Azure AI Modelkatalog](https://ai.azure.com) + - [Azure AI modelkatalog](https://ai.azure.com) -- Inferens af Phi i forskellige miljøer +- Inference Phi i forskellige miljøer - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [GitHub Modeller](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry Modelkatalog](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [GitHub modeller](./md/01.Introduction/02/02.GitHubModel.md) + - [Microsoft Foundry modelkatalog](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Inferens Phi Familie - - [Inferens Phi på iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Inferens Phi på Android](./md/01.Introduction/03/Android_Inference.md) - - [Inferens Phi på Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Inferens Phi på AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Inferens Phi med Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Inferens Phi på Lokal Server](./md/01.Introduction/03/Local_Server_Inference.md) - - [Inferens Phi på Fjernserver ved brug af AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Inferens Phi med Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inferens Phi--Vision lokalt](./md/01.Introduction/03/Vision_Inference.md) - - [Inferens Phi med Kaito AKS, Azure Containers (officiel support)](./md/01.Introduction/03/Kaito_Inference.md) -- [Kvantisering af Phi-familien](./md/01.Introduction/04/QuantifyingPhi.md) +- Inference Phi Family + - [Inference Phi i iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Inference Phi i Android](./md/01.Introduction/03/Android_Inference.md) + - [Inference Phi i Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Inference Phi i AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Inference Phi med Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Inference Phi i lokal server](./md/01.Introduction/03/Local_Server_Inference.md) + - [Inference Phi i fjernserver ved hjælp af AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Inference Phi med Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Inference Phi--Vision lokalt](./md/01.Introduction/03/Vision_Inference.md) + - [Inference Phi med Kaito AKS, Azure Containers (officiel support)](./md/01.Introduction/03/Kaito_Inference.md) +- [Kvantisering af Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) - [Kvantisering af Phi-3.5 / 4 ved brug af llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Kvantisering af Phi-3.5 / 4 ved brug af Generative AI extensioner til onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Kvantisering af Phi-3.5 / 4 ved brug af Generative AI udvidelser for onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [Kvantisering af Phi-3.5 / 4 ved brug af Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - [Kvantisering af Phi-3.5 / 4 ved brug af Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- Evaluering af Phi +- Evaluering Phi - [Ansvarlig AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry til evaluering](./md/01.Introduction/05/AIFoundry.md) - [Brug af Promptflow til evaluering](./md/01.Introduction/05/Promptflow.md) - RAG med Azure AI Search - - [Sådan bruger du Phi-4-mini og Phi-4-multimodal(RAG) med Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Hvordan man bruger Phi-4-mini og Phi-4-multimodal (RAG) med Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud lokal hybrid RAG med SQLite FTS5 og phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Phi applikationsudviklings-eksempler - - Tekst & Chat Applikationer - - Phi-4 Eksempler - - [📓] [Chat med Phi-4-mini ONNX Model](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Chat med Phi-4 lokal ONNX Model .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - Tekst- & chat-applikationer + - Phi-4 eksempler + - [📓] [Chat med Phi-4-mini ONNX model](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Chat med Phi-4 lokal ONNX model .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - [Chat .NET Console App med Phi-4 ONNX ved brug af Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 Eksempler - - [Lokal Chatbot i browseren ved brug af Phi3, ONNX Runtime Web og WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Lokal Chatbot i browseren ved hjælp af Phi3, ONNX Runtime Web og WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [Multi Model - Interaktiv Phi-3-mini og OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Opbygning af en wrapper og brug af Phi-3 med MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Modeloptimering - Hvordan man optimerer Phi-3-minimodel til ONNX Runtime Web med Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Modeloptimering - Hvordan man optimerer Phi-3-minimodellen for ONNX Runtime Web med Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [WinUI3 App med Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Multi Model AI-Drevet Notes App Eksempel](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Finjuster og integrer brugerdefinerede Phi-3 modeller med Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Finjuster og integrer brugerdefinerede Phi-3 modeller med Prompt flow i Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Evaluer den finjusterede Phi-3 / Phi-3.5 model i Microsoft Foundry med fokus på Microsofts principper for ansvarlig AI](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct sprogforudsigelseseksempel (kinesisk/engelsk)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + -[WinUI3 Multi Model AI Drevet Noter App Eksempel](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Finjuster og integrer tilpassede Phi-3 modeller med Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Finjuster og integrer tilpassede Phi-3 modeller med Prompt flow i Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Evaluer den finjusterede Phi-3 / Phi-3.5 Model i Microsoft Foundry med fokus på Microsofts principper for ansvarlig AI](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct sprogforudsigelse eksempel (kinesisk/engelsk)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Brug af Windows GPU til at oprette Prompt flow løsning med Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Brug af Microsoft Phi-3.5 tflite til at oprette Android app](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET eksempel ved brug af lokal ONNX Phi-3 model med Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Konsolchat .NET app med Semantic Kernel og Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Brug af Windows GPU til at skabe Prompt flow løsning med Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Brug af Microsoft Phi-3.5 tflite til at lave Android app](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Q&A .NET Eksempel med lokal ONNX Phi-3 model ved brug af Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Console chat .NET app med Semantic Kernel og Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK kodebaserede eksempler + - Azure AI Inference SDK Kodebaserede Eksempler - Phi-4 Eksempler - [📓] [Generer projektkode ved hjælp af Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 Eksempler - - [Byg din egen Visual Studio Code GitHub Copilot Chat med Microsoft Phi-3-familien](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Opret din egen Visual Studio Code Chat Copilot Agent med Phi-3.5 fra GitHub modeller](./md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Byg din egen Visual Studio Code GitHub Copilot Chat med Microsoft Phi-3 Familie](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Opret din egen Visual Studio Code Chat Copilot Agent med Phi-3.5 ved brug af GitHub Modeller](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - Avancerede ræsonnementseksempler + - Avancerede Resonerings Eksempler - Phi-4 Eksempler - - [📓] [Phi-4-mini-reasoning eller Phi-4-reasoning Eksempler](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Finjustering af Phi-4-mini-reasoning med Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Finjustering af Phi-4-mini-reasoning med Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning med GitHub modeller](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning med Microsoft Foundry modeller](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demonstrationer - - [Phi-4-mini demos hostet på Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demos hostet på Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [📓] [Phi-4-mini-resonering eller Phi-4-resonerings Eksempler](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Finjustering af Phi-4-mini-resonering med Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Finjustering af Phi-4-mini-resonering med Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-resonering med GitHub Modeller](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-resonering med Microsoft Foundry Modeller](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Demoer + - [Phi-4-mini demoer hostet på Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal demoer hostet på Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - Vision Eksempler - Phi-4 Eksempler - [📓] [Brug Phi-4-multimodal til at læse billeder og generere kode](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 Eksempler - [📓][Phi-3-vision-Billede tekst til tekst](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP Indlejring](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [DEMO: Phi-3 Genbrug](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Visuel sprogassistent - med Phi3-Vision og OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3-vision - Visuel sproghjælper - med Phi3-Vision og OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - [📓][Phi-3.5 Vision multi-frame eller multi-billede eksempel](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - [Phi-3 Vision Lokal ONNX Model ved brug af Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menu baseret Phi-3 Vision Lokal ONNX Model ved brug af Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [Menu-baseret Phi-3 Vision Lokal ONNX Model ved brug af Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - Reasoning-Vision Eksempler - - Phi-4-Reasoning-Vision-15B - - [📓] [Brug af Phi-4-Reasoning-Vision-15B til at opdage krydsning udenfor fodgængerfeltet](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Brug af Phi-4-Reasoning-Vision-15B til matematik](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Brug af Phi-4-Reasoning-Vision-15B til at opdage UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - Resonerings-Vision Eksempler + - Phi-4-Resonerings-Vision-15B + - [📓] [Brug af Phi-4-Resonerings-Vision-15B til at opdage ulovlig vejgang](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Brug af Phi-4-Resonerings-Vision-15B til matematik](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Brug af Phi-4-Resonerings-Vision-15B til UI detektion](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Matematik Eksempler - Phi-4-Mini-Flash-Reasoning-Instruct Eksempler [Matematik Demo med Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - Audio Eksempler - Phi-4 Eksempler - - [📓] [Uddragning af lydudskrifter ved brug af Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal lyd eksempel](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal taleoversættelse eksempel](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET konsolapplikation ved brug af Phi-4-multimodal lyd til at analysere en lydfil og generere udskrift](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [📓] [Udtrækning af lydtranskriptioner ved hjælp af Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal Lyd Eksempel](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal Taleoversættelses Eksempel](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET konsolapplikation ved brug af Phi-4-multimodal Lyd til at analysere en lydfil og generere transkript](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE Eksempler - Phi-3 / 3.5 Eksempler - - [📓] [Phi-3.5 Mixture of Experts Models (MoEs) Social Media Eksempel](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Opbygning af Retrieval-Augmented Generation (RAG) pipeline med NVIDIA NIM Phi-3 MOE, Azure AI Search og LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 Mixture of Experts Modeller (MoEs) Sociale Medier Eksempel](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Opbygning af en Retrieval-Augmented Generation (RAG) Pipeline med NVIDIA NIM Phi-3 MOE, Azure AI Search og LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Funktionsopkald Eksempler + - Funktion-Kald Eksempler - Phi-4 Eksempler 🆕 - - [📓] [Brug af funktionsopkald med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Brug af funktionsopkald til at skabe multi-agenter med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Brug af funktionsopkald med Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Brug af funktionsopkald med ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Multimodal Blandings Eksempler + - [📓] [Brug af Funktion-Kald med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Brug af Funktion-Kald til at skabe multi-agenter med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Brug af Funktion-Kald med Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Brug af Funktion-Kald med ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Multimodal Mixning Eksempler - Phi-4 Eksempler 🆕 - - [📓] [Brug af Phi-4-multimodal som teknologijournalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET konsolapplikation ved brug af Phi-4-multimodal til billedanalyse](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [📓] [Brug af Phi-4-multimodal som teknologi journalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET konsolapplikation ved brug af Phi-4-multimodal til at analysere billeder](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - Finjustering af Phi Eksempler - [Finjusteringsscenarier](./md/03.FineTuning/FineTuning_Scenarios.md) - [Finjustering vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - [Finjustering Lad Phi-3 blive en brancheekspert](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Finjustering af Phi-3 med AI Toolkit for VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Finjustering af Phi-3 med AI Toolkit til VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [Finjustering af Phi-3 med Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - [Finjustering af Phi-3 med Lora](./md/03.FineTuning/FineTuning_Lora.md) - [Finjustering af Phi-3 med QLora](./md/03.FineTuning/FineTuning_Qlora.md) @@ -203,6 +205,7 @@ Følg disse trin for at komme i gang med at bruge disse ressourcer: - [Finjustering med Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Finjustering med Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - [Finjustering af Phi-3-vision med Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Finjustering af Phi-3 med Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - [Finjustering af Phi-3-vision (officiel support)](./md/03.FineTuning/FineTuning_Vision.md) - [Finjustering af Phi-3 med Kaito AKS, Azure Containers (officiel support)](./md/03.FineTuning/FineTuning_Kaito.md) @@ -210,115 +213,117 @@ Følg disse trin for at komme i gang med at bruge disse ressourcer: - Hands on Lab - [Udforskning af banebrydende modeller: LLMs, SLMs, lokal udvikling og mere](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Frigørelse af NLP Potentiale: Finjustering med Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Udnyttelse af NLP Potentiale: Finjustering med Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Akademiske Forskningsartikler og Publikationer +- Akademiske forskningsartikler og publikationer - [Textbooks Are All You Need II: phi-1.5 teknisk rapport](https://arxiv.org/abs/2309.05463) - - [Phi-3 teknisk rapport: En stærkt kapabel sprogsmodel lokalt på din telefon](https://arxiv.org/abs/2404.14219) + - [Phi-3 teknisk rapport: En højt kapabel sprogmodel lokalt på din telefon](https://arxiv.org/abs/2404.14219) - [Phi-4 teknisk rapport](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini teknisk rapport: Kompakte men kraftfulde multimodale sprogsmodeller via Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Optimering af små sprogsmodeller til funktionsopkald i køretøjer](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Finjustering af PHI-3 til Multiple-Choice spørgsmål: Metodologi, resultater og udfordringer](https://arxiv.org/abs/2501.01588) + - [Phi-4-Mini teknisk rapport: Kompakte men kraftfulde multimodale sprogmodeller via mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Optimering af små sprogmodeller til funktionsopkald i køretøjer](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Finjustering af PHI-3 til multiple-choice spørgsmål: Metodologi, resultater og udfordringer](https://arxiv.org/abs/2501.01588) - [Phi-4-reasoning teknisk rapport](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Phi-4-mini-reasoning teknisk rapport](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Brug af Phi Modeller +## Brug af Phi modeller ### Phi på Microsoft Foundry -Du kan lære, hvordan du bruger Microsoft Phi, og hvordan du bygger E2E-løsninger på dine forskellige hardwareenheder. For at prøve Phi selv kan du starte med at lege med modellerne og tilpasse Phi til dine scenarier ved hjælp af [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Du kan lære mere i Kom godt i gang med [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Du kan lære, hvordan du bruger Microsoft Phi og bygger E2E-løsninger på dine forskellige hardwareenheder. For at opleve Phi selv, start med at lege med modellerne og tilpas Phi til dine scenarier ved hjælp af [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) Du kan lære mere om at komme i gang med [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Legeplads** +**Playground** Hver model har en dedikeret legeplads til at teste modellen [Azure AI Playground](https://aka.ms/try-phi3). -### Phi på GitHub Modeller +### Phi på GitHub modeller -Du kan lære, hvordan du bruger Microsoft Phi, og hvordan du bygger E2E-løsninger på dine forskellige hardwareenheder. For at prøve Phi selv kan du starte med at lege med modellen og tilpasse Phi til dine scenarier ved hjælp af [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Du kan lære mere i Kom godt i gang med [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Du kan lære, hvordan du bruger Microsoft Phi og bygger E2E-løsninger på dine forskellige hardwareenheder. For at opleve Phi selv, start med at lege med modellen og tilpas Phi til dine scenarier ved hjælp af [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) Du kan lære mere om at komme i gang med [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Legeplads** -Hver model har en dedikeret [legeplads til at teste modellen](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Playground** +Hver model har en dedikeret [legeplads for at teste modellen](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi på Hugging Face Du kan også finde modellen på [Hugging Face](https://huggingface.co/microsoft) -**Legeplads** -[Hugging Chat legeplads](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Playground** + [Hugging Chat legeplads](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Andre Kurser + ## 🎒 Andre kurser -Vores team producerer andre kurser! Se her: +Vores team producerer andre kurser! Tjek dem ud: -### LangChain -[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain for Beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +### LangChain +[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain for Beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agenter -[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agenter +[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Generativ AI Serie -[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### Generativ AI-serie +[![Generativ AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generativ AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generativ AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generativ AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Kerneuddannelse -[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) + +### Kerne-læring +[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersikkerhed for begyndere](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Webudvikling for begyndere](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR-udvikling for begyndere](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Copilot Serie -[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + +### Copilot-serie +[![Copilot til AI Paret Programmering](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot til C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Ansvarlig AI +## Ansvarlig AI -Microsoft er forpligtet til at hjælpe vores kunder med at bruge vores AI-produkter ansvarligt, dele vores læringer og opbygge tillidsbaserede partnerskaber gennem værktøjer som Transparency Notes og Impact Assessments. Mange af disse ressourcer kan findes på [https://aka.ms/RAI](https://aka.ms/RAI). +Microsoft er forpligtet til at hjælpe vores kunder med at bruge vores AI-produkter ansvarligt, dele vores erfaringer og opbygge tillidsbaserede partnerskaber gennem værktøjer som Transparency Notes og Impact Assessments. Mange af disse ressourcer kan findes på [https://aka.ms/RAI](https://aka.ms/RAI). Microsofts tilgang til ansvarlig AI er forankret i vores AI-principper om retfærdighed, pålidelighed og sikkerhed, privatliv og sikkerhed, inklusivitet, gennemsigtighed og ansvarlighed. -Store naturlige sprog-, billede- og tale-modeller - som dem, der bruges i denne prøve - kan potentielt opføre sig på måder, der er uretfærdige, upålidelige eller stødende, og derved forårsage skade. Konsulter venligst [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) for at blive informeret om risici og begrænsninger. -Den anbefalede tilgang til at afbøde disse risici er at inkludere et sikkerhedssystem i din arkitektur, som kan registrere og forhindre skadelig adfærd. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) tilbyder et uafhængigt beskyttelseslag, der kan opdage skadeligt bruger-genereret og AI-genereret indhold i applikationer og tjenester. Azure AI Content Safety inkluderer tekst- og billed-API'er, der gør det muligt at opdage materiale, som er skadeligt. Inden for Microsoft Foundry tillader Content Safety-tjenesten dig at se, udforske og prøve eksempelkode til at opdage skadeligt indhold på tværs af forskellige modaliteter. Følgende [quickstart dokumentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) guider dig gennem at lave forespørgsler til tjenesten. +Store sprog-, billede- og tale-modeller - som dem brugt i dette eksempel - kan potentielt opføre sig på måder, der er uretfærdige, upålidelige eller stødende, og dermed forårsage skade. Konsulter venligst [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) for at blive informeret om risici og begrænsninger. + + +Den anbefalede tilgang til at mindske disse risici er at inkludere et sikkerhedssystem i din arkitektur, der kan opdage og forhindre skadelig adfærd. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) leverer et uafhængigt beskyttelseslag, der er i stand til at opdage skadeligt brugergenereret og AI-genereret indhold i applikationer og tjenester. Azure AI Content Safety inkluderer tekst- og billed-API'er, der giver dig mulighed for at opdage materiale, der er skadeligt. Inden for Microsoft Foundry giver Content Safety-tjenesten dig mulighed for at se, udforske og prøve eksempelkode til at opdage skadeligt indhold på tværs af forskellige modaliteter. Den følgende [quickstart-dokumentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) guider dig gennem at foretage forespørgsler til tjenesten. -En anden faktor, der skal tages i betragtning, er den overordnede applikationsydelse. Med multimodale og multimodel-applikationer betragter vi ydeevne som, at systemet fungerer som du og dine brugere forventer, inklusive ikke at generere skadelige output. Det er vigtigt at vurdere ydeevnen af din samlede applikation ved hjælp af [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Du har også mulighed for at oprette og evaluere med [brugerdefinerede evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +En anden faktor at tage højde for er den samlede applikationsydelse. Med multimodale og multimodel-applikationer betragter vi ydelse som, at systemet fungerer, som du og dine brugere forventer, herunder ikke at generere skadelige output. Det er vigtigt at vurdere ydelsen af din samlede applikation ved hjælp af [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Du har også mulighed for at oprette og evaluere med [brugerdefinerede evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Du kan evaluere din AI-applikation i dit udviklingsmiljø ved at bruge [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Givet enten et testdatasæt eller et mål, bliver dine genererende AI-applikationsgenereringer kvantitativt målt med indbyggede evaluators eller brugerdefinerede evaluators efter eget valg. For at komme i gang med azure ai evaluation sdk til at evaluere dit system, kan du følge [quickstart-guiden](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Når du har udført en evalueringskørsel, kan du [visualisere resultaterne i Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Du kan evaluere din AI-applikation i dit udviklingsmiljø ved hjælp af [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Ved enten et testdatasæt eller et mål bliver dine generative AI-applikationsgenerationer kvantitativt målt med indbyggede evaluatorer eller brugerdefinerede evaluatorer efter eget valg. For at komme i gang med azure ai evaluation sdk til at evaluere dit system kan du følge [quickstart-guiden](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Når du har udført en evaluering, kan du [visualisere resultaterne i Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Varemærker -Dette projekt kan indeholde varemærker eller logoer for projekter, produkter eller tjenester. Autoriseret brug af Microsofts varemærker eller logoer er underlagt og skal følge [Microsofts retningslinjer for varemærker og mærker](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Brug af Microsofts varemærker eller logoer i modificerede versioner af dette projekt må ikke skabe forvirring eller antyde Microsoft-sponsorat. Enhver brug af tredjeparts varemærker eller logoer er underlagt disse tredjeparts retningslinjer. +Dette projekt kan indeholde varemærker eller logoer for projekter, produkter eller tjenester. Autoriseret brug af Microsoft-varemærker eller logoer er underlagt og skal følge [Microsofts varemærke- og brandretningslinjer](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Brug af Microsoft-varemærker eller logoer i ændrede versioner af dette projekt må ikke skabe forvirring eller antyde Microsoft-sponsorat. Enhver brug af tredjeparts varemærker eller logoer er underlagt disse tredjeparts politikker. ## Få hjælp -Hvis du sidder fast eller har spørgsmål om at bygge AI-apps, så deltag i: +Hvis du sidder fast eller har spørgsmål om at bygge AI-apps, deltag i: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Hvis du har produktfeedback eller oplever fejl under udvikling, besøg: +Hvis du har produktfeedback eller fejl under udviklingen, besøg: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Ansvarsfraskrivelse**: -Dette dokument er oversat ved hjælp af AI-oversættelsestjenesten [Co-op Translator](https://github.com/Azure/co-op-translator). Selvom vi bestræber os på nøjagtighed, bedes du være opmærksom på, at automatiserede oversættelser kan indeholde fejl eller unøjagtigheder. Det oprindelige dokument på dets modersmål skal betragtes som den autoritative kilde. For kritisk information anbefales professionel menneskelig oversættelse. Vi påtager os intet ansvar for eventuelle misforståelser eller fejltolkninger som følge af brugen af denne oversættelse. +**Ansvarsfraskrivelse**: +Dette dokument er blevet oversat ved hjælp af AI-oversættelsestjenesten [Co-op Translator](https://github.com/Azure/co-op-translator). Selvom vi bestræber os på nøjagtighed, skal du være opmærksom på, at automatiserede oversættelser kan indeholde fejl eller unøjagtigheder. Det originale dokument på dets oprindelige sprog bør betragtes som den autoritative kilde. For kritisk information anbefales professionel menneskelig oversættelse. Vi påtager os intet ansvar for misforståelser eller fejltolkninger, der opstår som følge af brugen af denne oversættelse. \ No newline at end of file diff --git a/translations/da/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/da/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..312da6e7c --- /dev/null +++ b/translations/da/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrid RAG med Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Forfatter:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Fokus:** Små Sprogmodeller (SLM'er), SQLite FTS5 BM25, Tætte Indlejringer, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivation: Dilemmaet med Søgeordsgengivelse i Lokale SLM'er\n", + "Standard RAG-arkitekturer, der udelukkende bygger på tætte vektorindlejringer, fejler ofte i at hente præcise numeriske koder (f.eks. `2.340.000 TL`, kontraktkoder, kontonumre). \n", + "Omvendt går sparsomme leksikale søgninger (BM25) glip af semantiske synonymer og omformulerede spørgsmål.\n", + "\n", + "Denne opskrift viser, hvordan man implementerer en **highspeed, in-memory Hybrid Retrieval-motor**, der kombinerer:\n", + "1. **Tætte Vektorer** (Cosinus Lighed)\n", + "2. **Sparsomme Leksikale Søgefunktioner** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Jordforankret Citatgenerering (`[1]`, `[2]`)** med Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Dobbelt SQLite-skema (tætte vektorer + virtuel FTS5 BM25-tabel)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Eksempel Indtagning & Udførelsesbenchmark\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Grounded Prompt Formulering for Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Ansvarsfraskrivelse**:\nDette dokument er blevet oversat ved hjælp af AI-oversættelsestjenesten [Co-op Translator](https://github.com/Azure/co-op-translator). Selvom vi bestræber os på nøjagtighed, skal du være opmærksom på, at automatiserede oversættelser kan indeholde fejl eller unøjagtigheder. Det originale dokument på dets oprindelige sprog bør betragtes som den autoritative kilde. For kritisk information anbefales professionel menneskelig oversættelse. Vi påtager os intet ansvar for misforståelser eller fejltolkninger, der opstår som følge af brugen af denne oversættelse.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/de/.co-op-translator.json b/translations/de/.co-op-translator.json index b214d6b9c..46bfa0472 100644 --- a/translations/de/.co-op-translator.json +++ b/translations/de/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "de" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:29:46+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:40:59+00:00", "source_file": "README.md", "language_code": "de" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "de" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:47:10+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "de" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:29:27+00:00", diff --git a/translations/de/README.md b/translations/de/README.md index cf3108ad8..f4b02bc81 100644 --- a/translations/de/README.md +++ b/translations/de/README.md @@ -1,12 +1,12 @@ -# Phi Kochbuch: Praktische Beispiele mit Microsofts Phi-Modellen +# Phi Cookbook: Praktische Beispiele mit Microsofts Phi-Modellen [![Öffnen und Verwenden der Beispiele in GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![In Dev Containers öffnen](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Im Dev Containers öffnen](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub-Beitragende](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub-Issues](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub Pull-Requests](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Willkommen](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![PRs willkommen](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![GitHub-Beobachter](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub-Forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) @@ -16,13 +16,13 @@ Phi ist eine Reihe von Open-Source-KI-Modellen, die von Microsoft entwickelt wurden. -Phi ist derzeit das leistungsstärkste und kosteneffizienteste kleine Sprachmodell (SLM) mit sehr guten Benchmarks in mehrsprachigen, logischen, Text-/Chat-Generierungs-, Codier-, Bild-, Audio- und weiteren Szenarien. +Phi ist derzeit das leistungsstärkste und kosteneffizienteste kleine Sprachmodell (SLM) mit sehr guten Benchmarks in mehrsprachigen, logischen, Text/Chat-Generierungs-, Codierungs-, Bild-, Audio- und anderen Szenarien. -Sie können Phi in der Cloud oder auf Edge-Geräten bereitstellen und generative KI-Anwendungen mit begrenzter Rechenleistung einfach erstellen. +Sie können Phi in der Cloud oder auf Edge-Geräten einsetzen und einfach generative KI-Anwendungen mit begrenzter Rechenleistung erstellen. -Folgen Sie diesen Schritten, um mit der Nutzung dieser Ressourcen zu beginnen: -1. **Forken Sie das Repository**: Klicken Sie [![GitHub-Forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Klonen Sie das Repository**: `git clone https://github.com/microsoft/PhiCookBook.git` +Folgen Sie diesen Schritten, um loszulegen: +1. **Repository forken**: Klicken Sie [![GitHub-Forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Repository klonen**: `git clone https://github.com/microsoft/PhiCookBook.git` 3. [**Treten Sie der Microsoft AI Discord Community bei und treffen Sie Experten und andere Entwickler**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/de/cover.eb18d1b9605d754b.webp) @@ -32,11 +32,11 @@ Folgen Sie diesen Schritten, um mit der Nutzung dieser Ressourcen zu beginnen: #### Unterstützt über GitHub Action (Automatisiert & Immer Aktuell) -[Arabisch](../ar/README.md) | [Bengalisch](../bn/README.md) | [Bulgarisch](../bg/README.md) | [Birmanisch (Myanmar)](../my/README.md) | [Chinesisch (vereinfacht)](../zh-CN/README.md) | [Chinesisch (traditionell, Hongkong)](../zh-HK/README.md) | [Chinesisch (traditionell, Macau)](../zh-MO/README.md) | [Chinesisch (traditionell, Taiwan)](../zh-TW/README.md) | [Kroatisch](../hr/README.md) | [Tschechisch](../cs/README.md) | [Dänisch](../da/README.md) | [Niederländisch](../nl/README.md) | [Estnisch](../et/README.md) | [Finnisch](../fi/README.md) | [Französisch](../fr/README.md) | [Deutsch](./README.md) | [Griechisch](../el/README.md) | [Hebräisch](../he/README.md) | [Hindi](../hi/README.md) | [Ungarisch](../hu/README.md) | [Indonesisch](../id/README.md) | [Italienisch](../it/README.md) | [Japanisch](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreanisch](../ko/README.md) | [Litauisch](../lt/README.md) | [Malaiisch](../ms/README.md) | [Malaiischalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalesisch](../ne/README.md) | [Nigerianisches Pidgin](../pcm/README.md) | [Norwegisch](../no/README.md) | [Persisch (Farsi)](../fa/README.md) | [Polnisch](../pl/README.md) | [Portugiesisch (Brasilien)](../pt-BR/README.md) | [Portugiesisch (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumänisch](../ro/README.md) | [Russisch](../ru/README.md) | [Serbisch (Kyrillisch)](../sr/README.md) | [Slowakisch](../sk/README.md) | [Slowenisch](../sl/README.md) | [Spanisch](../es/README.md) | [Suaheli](../sw/README.md) | [Schwedisch](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thailändisch](../th/README.md) | [Türkisch](../tr/README.md) | [Ukrainisch](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamesisch](../vi/README.md) +[Arabisch](../ar/README.md) | [Bengalisch](../bn/README.md) | [Bulgarisch](../bg/README.md) | [Birmanisch (Myanmar)](../my/README.md) | [Chinesisch (Vereinfacht)](../zh-CN/README.md) | [Chinesisch (Traditionell, Hongkong)](../zh-HK/README.md) | [Chinesisch (Traditionell, Macau)](../zh-MO/README.md) | [Chinesisch (Traditionell, Taiwan)](../zh-TW/README.md) | [Kroatisch](../hr/README.md) | [Tschechisch](../cs/README.md) | [Dänisch](../da/README.md) | [Niederländisch](../nl/README.md) | [Estnisch](../et/README.md) | [Finnisch](../fi/README.md) | [Französisch](../fr/README.md) | [Deutsch](./README.md) | [Griechisch](../el/README.md) | [Hebräisch](../he/README.md) | [Hindi](../hi/README.md) | [Ungarisch](../hu/README.md) | [Indonesisch](../id/README.md) | [Italienisch](../it/README.md) | [Japanisch](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreanisch](../ko/README.md) | [Litauisch](../lt/README.md) | [Malaiisch](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalesisch](../ne/README.md) | [Nigerianisches Pidgin](../pcm/README.md) | [Norwegisch](../no/README.md) | [Persisch (Farsi)](../fa/README.md) | [Polnisch](../pl/README.md) | [Portugiesisch (Brasilien)](../pt-BR/README.md) | [Portugiesisch (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumänisch](../ro/README.md) | [Russisch](../ru/README.md) | [Serbisch (Kyrillisch)](../sr/README.md) | [Slowakisch](../sk/README.md) | [Slowenisch](../sl/README.md) | [Spanisch](../es/README.md) | [Suaheli](../sw/README.md) | [Schwedisch](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thailändisch](../th/README.md) | [Türkisch](../tr/README.md) | [Ukrainisch](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamesisch](../vi/README.md) -> **Bevorzugen Sie das lokale Klonen?** +> **Bevorzugen Sie ein lokales Klonen?** > -> Dieses Repository enthält über 50 Sprachübersetzungen, was die Download-Größe erheblich vergrößert. Um ohne Übersetzungen zu klonen, verwenden Sie sparse checkout: +> Dieses Repository enthält über 50 Sprachübersetzungen, was die Downloadgröße erheblich erhöht. Um ohne Übersetzungen zu klonen, verwenden Sie Sparse Checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,7 +52,7 @@ Folgen Sie diesen Schritten, um mit der Nutzung dieser Ressourcen zu beginnen: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> So erhalten Sie alles, was Sie benötigen, um den Kurs mit einem viel schnelleren Download abzuschließen. +> Dies gibt Ihnen alles, was Sie benötigen, um den Kurs mit einem viel schnelleren Download abzuschließen. ## Inhaltsverzeichnis @@ -60,132 +60,134 @@ Folgen Sie diesen Schritten, um mit der Nutzung dieser Ressourcen zu beginnen: - Einführung - [Willkommen in der Phi-Familie](./md/01.Introduction/01/01.PhiFamily.md) - [Einrichten Ihrer Umgebung](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Verstehen der Schlüsseltechnologien](./md/01.Introduction/01/01.Understandingtech.md) + - [Verstehen von Schlüsseltechnologien](./md/01.Introduction/01/01.Understandingtech.md) - [KI-Sicherheit für Phi-Modelle](./md/01.Introduction/01/01.AISafety.md) - [Phi Hardware-Unterstützung](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi-Modelle & Verfügbarkeit auf Plattformen](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Phi-Modelle & Verfügbarkeit über Plattformen](./md/01.Introduction/01/01.Edgeandcloud.md) - [Verwendung von Guidance-ai und Phi](./md/01.Introduction/01/01.Guidance.md) - [GitHub Marketplace Modelle](https://github.com/marketplace/models) - - [Azure AI Modell-Katalog](https://ai.azure.com) + - [Azure AI Modellkatalog](https://ai.azure.com) -- Phi Inferenz in verschiedenen Umgebungen - - [Hugging Face](./md/01.Introduction/02/01.HF.md) - - [GitHub Modelle](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry Modell-Katalog](./md/01.Introduction/02/03.AzureAIFoundry.md) +- Ausführung von Phi in verschiedenen Umgebungen + - [Hugging face](./md/01.Introduction/02/01.HF.md) + - [GitHub-Modelle](./md/01.Introduction/02/02.GitHubModel.md) + - [Microsoft Foundry Modellkatalog](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Lokal](./md/01.Introduction/02/07.FoundryLocal.md) -- Phi Familien-Inferenz - - [Phi Inferenz auf iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Phi Inferenz auf Android](./md/01.Introduction/03/Android_Inference.md) - - [Phi Inferenz auf Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Phi Inferenz auf AI-PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Phi Inferenz mit Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Phi Inferenz auf lokalem Server](./md/01.Introduction/03/Local_Server_Inference.md) - - [Phi Inferenz auf Remote-Server mit AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Phi Inferenz mit Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Phi--Vision Inferenz lokal](./md/01.Introduction/03/Vision_Inference.md) - - [Phi Inferenz mit Kaito AKS, Azure Containern (offizielle Unterstützung)](./md/01.Introduction/03/Kaito_Inference.md) +- Ausführung der Phi-Familie + - [Ausführung von Phi unter iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Ausführung von Phi unter Android](./md/01.Introduction/03/Android_Inference.md) + - [Ausführung von Phi unter Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Ausführung von Phi unter AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Ausführung von Phi mit Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Ausführung von Phi im lokalen Server](./md/01.Introduction/03/Local_Server_Inference.md) + - [Ausführung von Phi im Remote-Server mit AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Ausführung von Phi mit Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Ausführung von Phi--Vision lokal](./md/01.Introduction/03/Vision_Inference.md) + - [Ausführung von Phi mit Kaito AKS, Azure Containers (offizielle Unterstützung)](./md/01.Introduction/03/Kaito_Inference.md) - [Quantifizierung der Phi-Familie](./md/01.Introduction/04/QuantifyingPhi.md) - - [Quantifizierung Phi-3.5 / 4 mit llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Quantifizierung Phi-3.5 / 4 mit generativen AI-Erweiterungen für onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Quantifizierung Phi-3.5 / 4 mit Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Quantifizierung Phi-3.5 / 4 mit Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Evaluation Phi - - [Verantwortliche KI](./md/01.Introduction/05/ResponsibleAI.md) - - [Microsoft Foundry für Evaluation](./md/01.Introduction/05/AIFoundry.md) - - [Verwendung von Promptflow für Evaluation](./md/01.Introduction/05/Promptflow.md) + - [Quantisierung von Phi-3.5 / 4 mit llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Quantisierung von Phi-3.5 / 4 mit Generative AI Erweiterungen für onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Quantisierung von Phi-3.5 / 4 mit Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Quantisierung von Phi-3.5 / 4 mit Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- Bewertung von Phi + - [Verantwortungsvolle KI](./md/01.Introduction/05/ResponsibleAI.md) + - [Microsoft Foundry für Bewertung](./md/01.Introduction/05/AIFoundry.md) + - [Verwendung von Promptflow für Bewertung](./md/01.Introduction/05/Promptflow.md) - RAG mit Azure AI Search - - [Wie man Phi-4-mini und Phi-4-multimodal(RAG) mit Azure AI Search verwendet](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Wie man Phi-4-mini und Phi-4-multimodal (RAG) mit Azure AI Search verwendet](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud lokales hybrides RAG mit SQLite FTS5 und phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Phi Anwendungsentwicklung Beispiele - Text- & Chat-Anwendungen - Phi-4 Beispiele - [📓] [Chat mit Phi-4-mini ONNX Modell](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Chat mit lokalem Phi-4 ONNX Modell .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Chat mit Phi-4 lokalem ONNX Modell .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - [Chat .NET Konsolen-App mit Phi-4 ONNX unter Verwendung von Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 Beispiele - - [Lokaler Chatbot im Browser unter Verwendung von Phi3, ONNX Runtime Web und WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Lokaler Chatbot im Browser mit Phi3, ONNX Runtime Web und WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Modell - Interaktives Phi-3-mini und OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [Multi Model - Interaktives Phi-3-mini und OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Erstellen eines Wrappers und Verwendung von Phi-3 mit MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Modelloptimierung - Wie man das Phi-3-mini-Modell für ONNX Runtime Web mit Olive optimiert](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Modelloptimierung - Wie man das Phi-3-min Modell für ONNX Runtime Web mit Olive optimiert](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [WinUI3 App mit Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Multi Model KI-gestützte Notizen App Beispiel](https://github.com/microsoft/ai-powered-notes-winui3-sample) + -[WinUI3 Multi Model KI-betriebene Notizen App Beispiel](https://github.com/microsoft/ai-powered-notes-winui3-sample) - [Feinabstimmung und Integration benutzerdefinierter Phi-3-Modelle mit Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - [Feinabstimmung und Integration benutzerdefinierter Phi-3-Modelle mit Prompt flow in Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Bewertung des feinabgestimmten Phi-3 / Phi-3.5 Modells in Microsoft Foundry mit Fokus auf die Responsible AI Prinzipien von Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct Sprachvorhersagebeispiel (Chinesisch/Englisch)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Bewertung des feinabgestimmten Phi-3 / Phi-3.5 Modells in Microsoft Foundry mit Fokus auf Microsofts Prinzipien für verantwortungsvolle KI](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct Sprachvorhersage Beispiel (Chinesisch/Englisch)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Verwendung von Windows GPU zur Erstellung einer Prompt flow Lösung mit Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Verwendung von Microsoft Phi-3.5 tflite zur Erstellung einer Android App](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Verwendung der Windows GPU zur Erstellung einer Prompt flow-Lösung mit Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Verwendung von Microsoft Phi-3.5 tflite zur Erstellung einer Android-App](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - [Q&A .NET Beispiel mit lokalem ONNX Phi-3 Modell unter Verwendung von Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - [Konsolen-Chat .NET App mit Semantic Kernel und Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK Code Beispiele + - Azure AI Inference SDK Codebasierte Beispiele - Phi-4 Beispiele - - [📓] [Projekterstellung mit Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [📓] [Generiere Projektcode mit Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 Beispiele - - [Erstellen Sie Ihren eigenen Visual Studio Code GitHub Copilot Chat mit Microsoft Phi-3 Familie](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Erstellen Sie Ihren eigenen Visual Studio Code Chat Copilot Agent mit Phi-3.5 anhand von GitHub Modellen](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Erstelle deinen eigenen Visual Studio Code GitHub Copilot Chat mit der Microsoft Phi-3 Familie](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Erstelle deinen eigenen Visual Studio Code Chat Copilot Agent mit Phi-3.5 anhand von GitHub Modellen](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - Erweiterte Reasoning Beispiele + - Fortgeschrittene Reasoning Beispiele - Phi-4 Beispiele - [📓] [Phi-4-mini-reasoning oder Phi-4-reasoning Beispiele](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Feinabstimmung von Phi-4-mini-reasoning mit Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Feinabstimmung von Phi-4-mini-reasoning mit Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Feinabstimmung Phi-4-mini-reasoning mit Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Feinabstimmung Phi-4-mini-reasoning mit Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-reasoning mit GitHub Modellen](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-reasoning mit Microsoft Foundry Modellen](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demos - [Phi-4-mini Demos gehostet auf Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal Demos gehostet auf Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal Demos gehostet auf Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - Vision Beispiele - Phi-4 Beispiele - - [📓] [Verwendung von Phi-4-multimodal zum Lesen von Bildern und zur Codeerzeugung](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [📓] [Verwende Phi-4-multimodal zum Lesen von Bildern und Generieren von Code](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 Beispiele - - [📓][Phi-3-vision-Bild Text zu Text](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-Image Text zu Text](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Einbettung](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [DEMO: Phi-3 Recycling](https://github.com/jennifermarsman/PhiRecycling/) - [Phi-3-vision - Visueller Sprachassistent - mit Phi3-Vision und OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision Multi-Frame oder Multi-Image Beispiel](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [📓][Phi-3.5 Vision Mehrfach-Frame oder Mehrfach-Bild Beispiel](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - [Phi-3 Vision Lokales ONNX Modell unter Verwendung von Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - [Menübasiertes Phi-3 Vision Lokales ONNX Modell unter Verwendung von Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - Reasoning-Vision Beispiele - Phi-4-Reasoning-Vision-15B - - [📓] [Verwendung von Phi-4-Reasoning-Vision-15B zur Erkennung von Jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Verwendung von Phi-4-Reasoning-Vision-15B zur Erkennung von Falschgehen](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [Verwendung von Phi-4-Reasoning-Vision-15B für Mathematik](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - [📓] [Verwendung von Phi-4-Reasoning-Vision-15B zur UI-Erkennung](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Mathematik Beispiele - - Phi-4-Mini-Flash-Reasoning-Instruct Beispiele [Mathematik Demo mit Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-Mini-Flash-Reasoning-Instruct Beispiele [Mathe Demo mit Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - Audio Beispiele - Phi-4 Beispiele - - [📓] [Extrahieren von Audiotranskripten mit Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Extrahieren von Audio-Transkripten mit Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Phi-4-multimodal Audio Beispiel](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal Sprachübersetzungsbeispiel](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET Konsolenanwendung mit Phi-4-multimodal Audio zur Analyse einer Audiodatei und Erzeugung eines Transkripts](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [📓] [Phi-4-multimodal Sprachübersetzungs-Beispiel](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET Konsolenanwendung mit Phi-4-multimodal Audio zur Analyse einer Audiodatei und zur Erstellung eines Transkripts](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE Beispiele - Phi-3 / 3.5 Beispiele - [📓] [Phi-3.5 Mixture of Experts Modelle (MoEs) Social Media Beispiel](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - [📓] [Erstellen einer Retrieval-Augmented Generation (RAG) Pipeline mit NVIDIA NIM Phi-3 MOE, Azure AI Search und LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Function Calling Beispiele + - Funktion Aufruf Beispiele - Phi-4 Beispiele 🆕 - [📓] [Verwendung von Function Calling mit Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - [📓] [Verwendung von Function Calling zur Erstellung von Multi-Agenten mit Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Verwendung von Function Calling mit Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [Verwendung von Function Calling mit ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Beispiel für multimodales Mischen + - Multimodale Mischungsbeispiele - Phi-4 Beispiele 🆕 - [📓] [Verwendung von Phi-4-multimodal als Technologiejournalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [.NET Konsolenanwendung mit Phi-4-multimodal zur Analyse von Bildern](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) @@ -193,118 +195,121 @@ Folgen Sie diesen Schritten, um mit der Nutzung dieser Ressourcen zu beginnen: - Feinabstimmung Phi Beispiele - [Feinabstimmungsszenarien](./md/03.FineTuning/FineTuning_Scenarios.md) - [Feinabstimmung vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Feinabstimmung Lassen Sie Phi-3 ein Industrieexperte werden](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Feinabstimmung von Phi-3 mit AI Toolkit für VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Feinabstimmung von Phi-3 mit Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Feinabstimmung von Phi-3 mit Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Feinabstimmung von Phi-3 mit QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Feinabstimmung von Phi-3 mit Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Feinabstimmung von Phi-3 mit Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Feinabstimmung: Lass Phi-3 zum Industrieexperten werden](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Feinabstimmung Phi-3 mit AI Toolkit für VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Feinabstimmung Phi-3 mit Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Feinabstimmung Phi-3 mit Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Feinabstimmung Phi-3 mit QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Feinabstimmung Phi-3 mit Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Feinabstimmung Phi-3 mit Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - [Feinabstimmung mit Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Feinabstimmung mit Microsoft Olive Hands-On Labor](./md/03.FineTuning/olive-lab/readme.md) - - [Feinabstimmung von Phi-3-vision mit Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Feinabstimmung von Phi-3 mit Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Feinabstimmung mit Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) + - [Feinabstimmung Phi-3-vision mit Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Feinabstimmung von Phi-3 mit dem Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - [Feinabstimmung von Phi-3-vision (offizielle Unterstützung)](./md/03.FineTuning/FineTuning_Vision.md) - - [Feinabstimmung Phi-3 mit Kaito AKS, Azure Containers (offizielle Unterstützung)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Feinabstimmung Phi-3 und 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + - [Feinabstimmung von Phi-3 mit Kaito AKS, Azure Container (offizielle Unterstützung)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Feinabstimmung von Phi-3 und 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) -- Hands-on-Lab - - [Erkundung hochmoderner Modelle: LLMs, SLMs, lokale Entwicklung und mehr](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Entfesseln des NLP-Potenzials: Feinabstimmung mit Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) +- Praktisches Labor + - [Erkundung modernster Modelle: LLMs, SLMs, lokale Entwicklung und mehr](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Das Potenzial von NLP erschließen: Feinabstimmung mit Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Wissenschaftliche Forschungsarbeiten und Publikationen - - [Lehrbücher sind alles, was Sie brauchen II: phi-1.5 technischer Bericht](https://arxiv.org/abs/2309.05463) - - [Phi-3 Technischer Bericht: Ein hochleistungsfähiges Sprachmodell lokal auf Ihrem Telefon](https://arxiv.org/abs/2404.14219) +- Wissenschaftliche Forschungsberichte und Veröffentlichungen + - [Textbooks Are All You Need II: phi-1.5 technischer Bericht](https://arxiv.org/abs/2309.05463) + - [Phi-3 Technischer Bericht: Ein leistungsfähiges Sprachmodell lokal auf Ihrem Telefon](https://arxiv.org/abs/2404.14219) - [Phi-4 Technischer Bericht](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini Technischer Bericht: Kompakte und dennoch leistungsstarke multimodale Sprachmodelle durch Mischung von LoRAs](https://arxiv.org/abs/2503.01743) - - [Optimierung kleiner Sprachmodelle für In-Fahrzeug-Funktionsaufrufe](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Feinabstimmung PHI-3 für Multiple-Choice-Fragen: Methodik, Ergebnisse und Herausforderungen](https://arxiv.org/abs/2501.01588) - - [Phi-4-reasoning Technischer Bericht](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini-reasoning Technischer Bericht](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Phi-4-Mini Technischer Bericht: Kompakte und dennoch leistungsstarke multimodale Sprachmodelle mittels Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Optimierung kleiner Sprachmodelle für In-Fahrzeug Funktionsaufrufe](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Feinabstimmung PHI-3 für Multiple-Choice-Fragebeantwortung: Methodik, Ergebnisse und Herausforderungen](https://arxiv.org/abs/2501.01588) + - [Phi-4-reasoning technischer Bericht](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-reasoning technischer Bericht](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Verwendung von Phi-Modellen ### Phi auf Microsoft Foundry -Sie können lernen, wie man Microsoft Phi verwendet und wie man E2E-Lösungen auf Ihren verschiedenen Hardwaregeräten erstellt. Um Phi selbst zu erleben, beginnen Sie damit, mit den Modellen zu experimentieren und Phi für Ihre Szenarien anzupassen, indem Sie den [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) nutzen. Mehr erfahren Sie unter Erste Schritte mit [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Sie können lernen, wie Sie Microsoft Phi verwenden und wie Sie E2E-Lösungen auf Ihren verschiedenen Hardwaregeräten erstellen. Um Phi selbst zu erleben, beginnen Sie damit, mit den Modellen zu experimentieren und Phi für Ihre Szenarien mit dem [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) anzupassen. Weitere Informationen finden Sie unter Erste Schritte mit [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Spielwiese** -Jedes Modell verfügt über eine eigene Spielwiese zum Testen des Modells [Azure AI Playground](https://aka.ms/try-phi3). +**Spielplatz** +Jedes Modell verfügt über einen eigenen Spielplatz, um das Modell zu testen, den [Azure AI Playground](https://aka.ms/try-phi3). -### Phi auf GitHub Models +### Phi auf GitHub Modelle -Sie können lernen, wie man Microsoft Phi verwendet und wie man E2E-Lösungen auf Ihren verschiedenen Hardwaregeräten erstellt. Um Phi selbst zu erleben, beginnen Sie damit, mit dem Modell zu experimentieren und Phi für Ihre Szenarien anzupassen, indem Sie den [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) nutzen. Mehr erfahren Sie unter Erste Schritte mit [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Sie können lernen, wie Sie Microsoft Phi verwenden und wie Sie E2E-Lösungen auf Ihren verschiedenen Hardwaregeräten erstellen. Um Phi selbst zu erleben, beginnen Sie damit, mit dem Modell zu experimentieren und Phi für Ihre Szenarien mit dem [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) anzupassen. Weitere Informationen finden Sie unter Erste Schritte mit [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Spielwiese** -Jedes Modell hat eine eigene [Spielwiese zum Testen des Modells](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Spielplatz** +Jedes Modell verfügt über einen eigenen [Spielplatz, um das Modell zu testen](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi auf Hugging Face -Sie können das Modell auch auf [Hugging Face](https://huggingface.co/microsoft) finden. +Sie können das Modell auch auf [Hugging Face](https://huggingface.co/microsoft) finden -**Spielwiese** - [Hugging Chat Spielwiese](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Spielplatz** + [Hugging Chat Spielplatz](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) - ## 🎒 Weitere Kurse + ## 🎒 Andere Kurse -Unser Team produziert weitere Kurse! Schauen Sie vorbei: +Unser Team bietet weitere Kurse an! Schauen Sie sich an: ### LangChain -[![LangChain4j für Anfänger](https://img.shields.io/badge/LangChain4j%20für%20Anfänger-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js für Anfänger](https://img.shields.io/badge/LangChain.js%20für%20Anfänger-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain für Anfänger](https://img.shields.io/badge/LangChain%20für%20Anfänger-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain4j für Anfänger](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js für Anfänger](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain für Anfänger](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- ### Azure / Edge / MCP / Agents -[![AZD für Anfänger](https://img.shields.io/badge/AZD%20für%20Anfänger-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI für Anfänger](https://img.shields.io/badge/Edge%20AI%20für%20Anfänger-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP für Anfänger](https://img.shields.io/badge/MCP%20für%20Anfänger-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents für Anfänger](https://img.shields.io/badge/AI%20Agents%20für%20Anfänger-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AZD für Anfänger](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI für Anfänger](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP für Anfänger](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![KI-Agenten für Anfänger](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Generative KI Serie -[![Generative KI für Anfänger](https://img.shields.io/badge/Generative%20KI%20für%20Anfänger-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative KI (.NET)](https://img.shields.io/badge/Generative%20KI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative KI (Java)](https://img.shields.io/badge/Generative%20KI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generative KI (JavaScript)](https://img.shields.io/badge/Generative%20KI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### Generative KI-Reihe +[![Generative KI für Anfänger](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative KI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative KI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generative KI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- ### Kernlernen -[![ML für Anfänger](https://img.shields.io/badge/ML%20für%20Anfänger-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Datenwissenschaft für Anfänger](https://img.shields.io/badge/Datenwissenschaft%20für%20Anfänger-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![KI für Anfänger](https://img.shields.io/badge/KI%20für%20Anfänger-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersicherheit für Anfänger](https://img.shields.io/badge/Cybersicherheit%20für%20Anfänger-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Webentwicklung für Anfänger](https://img.shields.io/badge/Web%20Dev%20für%20Anfänger-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT für Anfänger](https://img.shields.io/badge/IoT%20für%20Anfänger-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR-Entwicklung für Anfänger](https://img.shields.io/badge/XR%20Entwicklung%20für%20Anfänger-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![ML für Anfänger](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Datenwissenschaft für Anfänger](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![KI für Anfänger](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersicherheit für Anfänger](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Webentwicklung für Anfänger](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT für Anfänger](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR-Entwicklung für Anfänger](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Copilot-Serie -[![Copilot für KI-Paarprogrammierung](https://img.shields.io/badge/Copilot%20für%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot für C#/.NET](https://img.shields.io/badge/Copilot%20für%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +### Copilot-Reihe +[![Copilot für KI-unterstütztes Paarprogrammieren](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot für C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot Abenteuer](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Verantwortliche KI +## Verantwortungsvolle KI + +Microsoft engagiert sich dafür, unseren Kunden zu helfen, unsere KI-Produkte verantwortungsvoll zu nutzen, unsere Erkenntnisse zu teilen und vertrauensbasierte Partnerschaften durch Werkzeuge wie Transparenznotizen und Folgewirkungsbeurteilungen aufzubauen. Viele dieser Ressourcen finden Sie unter [https://aka.ms/RAI](https://aka.ms/RAI). +Der Ansatz von Microsoft zur verantwortungsvollen KI basiert auf unseren KI-Prinzipien von Fairness, Zuverlässigkeit und Sicherheit, Datenschutz und Sicherheit, Inklusivität, Transparenz und Verantwortlichkeit. + +Groß angelegte Sprach-, Bild- und Sprachmodelle – wie die in diesem Beispiel verwendeten – können sich potenziell unfair, unzuverlässig oder anstößig verhalten und so Schäden verursachen. Bitte konsultieren Sie die [Azure OpenAI-Dienst Transparenznote](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), um über Risiken und Einschränkungen informiert zu sein. -Microsoft engagiert sich dafür, unseren Kunden zu helfen, unsere KI-Produkte verantwortungsvoll zu nutzen, unsere Erkenntnisse zu teilen und vertrauensbasierte Partnerschaften durch Tools wie Transparenz-Notizen und Wirkungseinschätzungen aufzubauen. Viele dieser Ressourcen finden Sie unter [https://aka.ms/RAI](https://aka.ms/RAI). -Der Ansatz von Microsoft zu verantwortlicher KI basiert auf unseren KI-Prinzipien Fairness, Zuverlässigkeit und Sicherheit, Privatsphäre und Sicherheit, Inklusivität, Transparenz und Verantwortlichkeit. -Groß angelegte Sprach-, Bild- und Sprachmodelle – wie die in diesem Beispiel verwendeten – können sich potenziell unfair, unzuverlässig oder anstößig verhalten und dadurch Schäden verursachen. Bitte konsultieren Sie die [Transparenz-Note zum Azure OpenAI-Dienst](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), um über Risiken und Einschränkungen informiert zu sein. -Der empfohlene Ansatz zur Minderung dieser Risiken besteht darin, ein Sicherheitssystem in Ihre Architektur zu integrieren, das schädliches Verhalten erkennen und verhindern kann. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) bietet eine unabhängige Schutzeinheit, die schädliche von Nutzern generierte und KI-generierte Inhalte in Anwendungen und Diensten erkennen kann. Azure AI Content Safety umfasst Text- und Bild-APIs, mit denen Sie schädliches Material erkennen können. Innerhalb von Microsoft Foundry ermöglicht Ihnen der Content Safety-Dienst, Beispielcode zum Erkennen schädlicher Inhalte über verschiedene Modalitäten hinweg anzusehen, zu erkunden und auszuprobieren. Die folgende [Schnellstart-Dokumentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) führt Sie durch das Senden von Anfragen an den Dienst. +Der empfohlene Ansatz zur Minderung dieser Risiken besteht darin, ein Sicherheitssystem in Ihre Architektur zu integrieren, das schädliches Verhalten erkennen und verhindern kann. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) bietet eine unabhängige Schutzschicht, die in der Lage ist, schädliche von Nutzern generierte und KI-generierte Inhalte in Anwendungen und Diensten zu erkennen. Azure AI Content Safety umfasst Text- und Bild-APIs, mit denen Sie schädliches Material erkennen können. Innerhalb von Microsoft Foundry ermöglicht der Content Safety-Dienst das Anzeigen, Erkunden und Ausprobieren von Beispielcode zur Erkennung schädlicher Inhalte in verschiedenen Modalitäten. Die folgende [Quickstart-Dokumentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) führt Sie durch das Erstellen von Anfragen an den Dienst. -Ein weiterer Aspekt, der zu berücksichtigen ist, ist die Gesamtleistung der Anwendung. Bei multimodalen und multimodellen Anwendungen verstehen wir unter Leistung, dass das System so funktioniert, wie Sie und Ihre Benutzer es erwarten, einschließlich der Vermeidung von schädlichen Ausgaben. Es ist wichtig, die Leistung Ihrer gesamten Anwendung mit [Performance- und Qualitäts- sowie Risiko- und Sicherheitsevaluator](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) zu bewerten. Sie haben auch die Möglichkeit, mit [benutzerdefinierten Evaluatoren](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) zu erstellen und zu evaluieren. +Ein weiterer zu berücksichtigender Aspekt ist die Gesamtleistung der Anwendung. Bei multimodalen und multimodellen Anwendungen verstehen wir unter Leistung, dass das System so funktioniert, wie Sie und Ihre Benutzer es erwarten, einschließlich der Nicht-Erzeugung schädlicher Ausgaben. Es ist wichtig, die Leistung Ihrer Gesamtanwendung mit [Performance- und Qualitäts- sowie Risiko- und Sicherheitsbewertungswerkzeugen](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) zu bewerten. Sie haben auch die Möglichkeit, mit [benutzerdefinierten Bewertungswerkzeugen](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) zu erstellen und zu bewerten. -Sie können Ihre KI-Anwendung in Ihrer Entwicklungsumgebung mit dem [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) bewerten. Basierend auf einem Testdatensatz oder einem Ziel werden die Generierungen Ihrer generativen KI-Anwendung quantitativ mit eingebauten oder benutzerdefinierten Evaluatoren Ihrer Wahl gemessen. Um mit dem Azure AI Evaluation SDK zur Bewertung Ihres Systems zu beginnen, können Sie der [Schnellstartanleitung](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) folgen. Nachdem Sie eine Evaluierung durchgeführt haben, können Sie die [Ergebnisse in Microsoft Foundry visualisieren](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Sie können Ihre KI-Anwendung in Ihrer Entwicklungsumgebung mit dem [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) bewerten. Anhand eines Testdatensatzes oder eines Ziels werden Ihre generativen KI-Anwendungs-Generierungen quantitativ mit integrierten Bewertungswerkzeugen oder Ihren eigenen benutzerdefinierten Bewertungswerkzeugen gemessen. Um mit dem Azure AI Evaluation SDK zur Bewertung Ihres Systems zu beginnen, können Sie dem [Quickstart-Leitfaden](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) folgen. Sobald Sie eine Bewertung durchgeführt haben, können Sie die [Ergebnisse in Microsoft Foundry visualisieren](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Marken -Dieses Projekt kann Marken oder Logos für Projekte, Produkte oder Dienstleistungen enthalten. Die autorisierte Verwendung von Microsoft-Marken oder -Logos unterliegt den [Microsoft-Marken- und Markenrichtlinien](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) und muss diesen folgen. -Die Verwendung von Microsoft-Marken oder -Logos in modifizierten Versionen dieses Projekts darf keine Verwirrung stiften oder eine Microsoft-Unterstützung implizieren. Jegliche Verwendung von Marken oder Logos Dritter unterliegt den jeweiligen Richtlinien dieser Dritten. +Dieses Projekt kann Marken oder Logos für Projekte, Produkte oder Dienste enthalten. Die autorisierte Verwendung von Microsoft-Marken oder -Logos unterliegt den [Microsoft-Marken- & Markenrichtlinien](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) und muss diese befolgen. +Die Verwendung von Microsoft-Marken oder -Logos in modifizierten Versionen dieses Projekts darf nicht zu Verwirrung führen oder auf eine Microsoft-Unterstützung hinweisen. Jede Verwendung von Marken oder Logos Dritter unterliegt den Richtlinien dieser Dritten. ## Hilfe erhalten @@ -312,13 +317,13 @@ Wenn Sie nicht weiterkommen oder Fragen zum Erstellen von KI-Anwendungen haben, [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Wenn Sie Produktfeedback geben oder Fehler beim Erstellen feststellen, besuchen Sie: +Wenn Sie Produkt-Feedback oder Fehler beim Erstellen haben, besuchen Sie: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Haftungsausschluss**: -Dieses Dokument wurde mithilfe des KI-Übersetzungsdienstes [Co-op Translator](https://github.com/Azure/co-op-translator) übersetzt. Obwohl wir auf Genauigkeit achten, beachten Sie bitte, dass automatisierte Übersetzungen Fehler oder Ungenauigkeiten enthalten können. Das Originaldokument in seiner ursprünglichen Sprache ist als autoritative Quelle zu betrachten. Für wichtige Informationen wird eine professionelle menschliche Übersetzung empfohlen. Wir übernehmen keine Haftung für Missverständnisse oder Fehlinterpretationen, die aus der Verwendung dieser Übersetzung entstehen. +**Haftungsausschluss**: +Dieses Dokument wurde mit dem KI-Übersetzungsdienst [Co-op Translator](https://github.com/Azure/co-op-translator) übersetzt. Obwohl wir uns um Genauigkeit bemühen, beachten Sie bitte, dass automatisierte Übersetzungen Fehler oder Ungenauigkeiten enthalten können. Das Originaldokument in seiner Ursprungssprache gilt als maßgebliche Quelle. Bei kritischen Informationen wird eine professionelle menschliche Übersetzung empfohlen. Wir übernehmen keine Haftung für Missverständnisse oder Fehlinterpretationen, die aus der Verwendung dieser Übersetzung entstehen. \ No newline at end of file diff --git a/translations/de/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/de/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..980ce6b8c --- /dev/null +++ b/translations/de/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrides RAG mit Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Autor:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Fokus:** Kleine Sprachmodelle (SLMs), SQLite FTS5 BM25, Dichte Einbettungen, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivation: Das Keyword-Retrieval-Dilemma bei lokalen SLMs\n", + "Standard-RAG-Architekturen, die ausschließlich auf dichte Vektor-Einbettungen setzen, versagen oft dabei, exakte numerische Token (z. B. `2.340.000 TL`, Vertragscodes, Kontonummern) abzurufen. \n", + "Dagegen verpasst die spärliche lexikalische Suche (BM25) semantische Synonyme und paraphrasierte Fragen.\n", + "\n", + "Dieses Kochbuch zeigt, wie man eine **hochgeschwindigkeitsfähige, speicherinterne Hybrid-Retrieval-Engine** implementiert, die kombiniert:\n", + "1. **Dichte Vektoren** (Kosinusähnlichkeit)\n", + "2. **Spärliche lexikalische Suche** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Grounded Citation Generation (`[1]`, `[2]`)** mit Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Duales SQLite-Schema (Dichte Vektoren + Virtuelle FTS5 BM25 Tabelle)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Stichprobenaufnahme- und Ausführungsbenchmark\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Grounded Prompt-Formulierung für Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Haftungsausschluss**:\nDieses Dokument wurde mit dem KI-Übersetzungsdienst [Co-op Translator](https://github.com/Azure/co-op-translator) übersetzt. Obwohl wir uns um Genauigkeit bemühen, beachten Sie bitte, dass automatisierte Übersetzungen Fehler oder Ungenauigkeiten enthalten können. Das Originaldokument in seiner Ursprungssprache gilt als maßgebliche Quelle. Bei kritischen Informationen wird eine professionelle menschliche Übersetzung empfohlen. Wir übernehmen keine Haftung für Missverständnisse oder Fehlinterpretationen, die aus der Verwendung dieser Übersetzung entstehen.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/el/.co-op-translator.json b/translations/el/.co-op-translator.json index fed74bbb2..b7e38dcd2 100644 --- a/translations/el/.co-op-translator.json +++ b/translations/el/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "el" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:56:48+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:11:02+00:00", "source_file": "README.md", "language_code": "el" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "el" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:18:39+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "el" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:58+00:00", diff --git a/translations/el/README.md b/translations/el/README.md index e7c9859ab..f2fe568b6 100644 --- a/translations/el/README.md +++ b/translations/el/README.md @@ -1,12 +1,12 @@ -# Phi Cookbook: Παραδείγματα με Πρακτική Εφαρμογή με τα Μοντέλα Phi της Microsoft +# Phi Cookbook: Παραδείγματα Hands-On με τα Μοντέλα Phi της Microsoft [![Άνοιγμα και χρήση των δειγμάτων στο GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Άνοιγμα σε Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![Συνεισφέροντες GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![Θέματα GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Αιτήματα αλλαγών GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![Καλωσορίζουμε PRs](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Αιτήματα έλξης GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![Τα PR είναι ευπρόσδεκτα](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![Παρατηρητές GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![Forks GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) @@ -14,29 +14,29 @@ [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Το Phi είναι μια σειρά ανοικτών μοντέλων τεχνητής νοημοσύνης που αναπτύχθηκαν από τη Microsoft. +Το Phi είναι μια σειρά μοντέλων AI ανοιχτού κώδικα που αναπτύχθηκαν από τη Microsoft. -Το Phi είναι αυτή τη στιγμή το πιο ισχυρό και οικονομικά αποδοτικό μικρό μοντέλο γλώσσας (SLM), με πολύ καλά benchmarks σε πολυγλωσσικά, λογική, παραγωγή κειμένου/συνομιλίας, κωδικοποίηση, εικόνες, ήχο και άλλα σενάρια. +Το Phi είναι σήμερα το ισχυρότερο και πιο αποδοτικό μοντέλο μικρής γλώσσας (SLM), με πολύ καλά χαρακτηριστικά σε πολύγλωσση υποστήριξη, λογική, δημιουργία κειμένου/συνομιλίας, κωδικοποίηση, εικόνες, ήχο και άλλα σενάρια. -Μπορείτε να αναπτύξετε το Phi στο cloud ή σε συσκευές άκρης, και μπορείτε εύκολα να χτίσετε εφαρμογές γεννητικής AI με περιορισμένη υπολογιστική ισχύ. +Μπορείτε να αναπτύξετε το Phi στο cloud ή σε συσκευές άκρου, και μπορείτε εύκολα να δημιουργήσετε εφαρμογές γεννητικής AI με περιορισμένη υπολογιστική ισχύ. -Ακολουθήστε αυτά τα βήματα για να ξεκινήσετε με τη χρήση αυτού του πόρου: -1. **Fork το Αποθετήριο**: Κάντε κλικ [![Forks GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Κλωνοποιήστε το Αποθετήριο**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Εγγραφείτε στην Κοινότητα Microsoft AI στο Discord και συναντήστε ειδικούς και άλλους προγραμματιστές**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +Ακολουθήστε αυτά τα βήματα για να ξεκινήσετε με αυτούς τους πόρους: +1. **Fork το Αποθετήριο**: Κάντε κλικ [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Κλωνοποίηση του Αποθετηρίου**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Εγγραφείτε στην Κοινότητα Discord της Microsoft AI και συναντήστε ειδικούς και άλλους προγραμματιστές**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/el/cover.eb18d1b9605d754b.webp) -### 🌐 Υποστήριξη Πολλών Γλωσσών +### 🌐 Υποστήριξη Πολυγλωσσικού Περιβάλλοντος -#### Υποστηρίζεται μέσω GitHub Action (Αυτοματοποιημένο και Πάντα Ενημερωμένο) +#### Υποστηρίζεται μέσω GitHub Action (Αυτοματοποιημένο & Πάντα Ενημερωμένο) -[Αραβικά](../ar/README.md) | [Βεγγαλικά](../bn/README.md) | [Βουλγαρικά](../bg/README.md) | [Βιρμανικά (Μιανμάρ)](../my/README.md) | [Κινέζικα (Απλοποιημένα)](../zh-CN/README.md) | [Κινέζικα (Παραδοσιακά, Χονγκ Κονγκ)](../zh-HK/README.md) | [Κινέζικα (Παραδοσιακά, Μακάο)](../zh-MO/README.md) | [Κινέζικα (Παραδοσιακά, Ταϊβάν)](../zh-TW/README.md) | [Κροατικά](../hr/README.md) | [Τσέχικα](../cs/README.md) | [Δανέζικα](../da/README.md) | [Ολλανδικά](../nl/README.md) | [Εσθονικά](../et/README.md) | [Φινλανδικά](../fi/README.md) | [Γαλλικά](../fr/README.md) | [Γερμανικά](../de/README.md) | [Ελληνικά](./README.md) | [Εβραϊκά](../he/README.md) | [Χίντι](../hi/README.md) | [Ουγγρικά](../hu/README.md) | [Ινδονησιακά](../id/README.md) | [Ιταλικά](../it/README.md) | [Ιαπωνικά](../ja/README.md) | [Κανάντα](../kn/README.md) | [Χμερ](../km/README.md) | [Κορεατικά](../ko/README.md) | [Λιθουανικά](../lt/README.md) | [Μαλάι](../ms/README.md) | [Μαλαγιάλαμ](../ml/README.md) | [Μαραθικά](../mr/README.md) | [Νεπαλικά](../ne/README.md) | [Νιγηριανή Πίνγκιν](../pcm/README.md) | [Νορβηγικά](../no/README.md) | [Περσικά (Φαρσί)](../fa/README.md) | [Πολωνικά](../pl/README.md) | [Πορτογαλικά (Βραζιλία)](../pt-BR/README.md) | [Πορτογαλικά (Πορτογαλία)](../pt-PT/README.md) | [Πουντζάμπι (Gurmukhi)](../pa/README.md) | [Ρουμανικά](../ro/README.md) | [Ρωσικά](../ru/README.md) | [Σερβικά (Κυριλλικά)](../sr/README.md) | [Σλοβακικά](../sk/README.md) | [Σλοβενικά](../sl/README.md) | [Ισπανικά](../es/README.md) | [Σουαχίλι](../sw/README.md) | [Σουηδικά](../sv/README.md) | [Ταγκαλόγκ (Φιλιππινέζικα)](../tl/README.md) | [Ταμίλ](../ta/README.md) | [Τελούγκου](../te/README.md) | [Ταϊλανδικά](../th/README.md) | [Τουρκικά](../tr/README.md) | [Ουκρανικά](../uk/README.md) | [Ουρντού](../ur/README.md) | [Βιετναμέζικα](../vi/README.md) +[Αραβικά](../ar/README.md) | [Βεγγαλικά](../bn/README.md) | [Βουλγαρικά](../bg/README.md) | [Βιρμανικά (Μιανμάρ)](../my/README.md) | [Κινέζικα (Απλοποιημένα)](../zh-CN/README.md) | [Κινέζικα (Παραδοσιακά, Χονγκ Κονγκ)](../zh-HK/README.md) | [Κινέζικα (Παραδοσιακά, Μακάο)](../zh-MO/README.md) | [Κινέζικα (Παραδοσιακά, Ταϊβάν)](../zh-TW/README.md) | [Κροατικά](../hr/README.md) | [Τσέχικα](../cs/README.md) | [Δανέζικα](../da/README.md) | [Ολλανδικά](../nl/README.md) | [Εσθονικά](../et/README.md) | [Φινλανδικά](../fi/README.md) | [Γαλλικά](../fr/README.md) | [Γερμανικά](../de/README.md) | [Ελληνικά](./README.md) | [Εβραϊκά](../he/README.md) | [Χίντι](../hi/README.md) | [Ουγγρικά](../hu/README.md) | [Ινδονησιακά](../id/README.md) | [Ιταλικά](../it/README.md) | [Ιαπωνικά](../ja/README.md) | [Κανάντα](../kn/README.md) | [Χμερ](../km/README.md) | [Κορεατικά](../ko/README.md) | [Λιθουανικά](../lt/README.md) | [Μαλαϊκά](../ms/README.md) | [Μαλαγιαλάμ](../ml/README.md) | [Μαράθι](../mr/README.md) | [Νεπαλικά](../ne/README.md) | [Νιγηριανή Πιντζίν](../pcm/README.md) | [Νορβηγικά](../no/README.md) | [Περσικά (Φαρσί)](../fa/README.md) | [Πολωνικά](../pl/README.md) | [Πορτογαλικά (Βραζιλία)](../pt-BR/README.md) | [Πορτογαλικά (Πορτογαλία)](../pt-PT/README.md) | [Πουντζάμπι (Γκουρμούκι)](../pa/README.md) | [Ρουμανικά](../ro/README.md) | [Ρωσικά](../ru/README.md) | [Σερβικά (Κυριλλικά)](../sr/README.md) | [Σλοβακικά](../sk/README.md) | [Σλοβενικά](../sl/README.md) | [Ισπανικά](../es/README.md) | [Σουαχίλι](../sw/README.md) | [Σουηδικά](../sv/README.md) | [Ταγκάλογκ (Φιλιππίνες)](../tl/README.md) | [Ταμίλ](../ta/README.md) | [Τελγού](../te/README.md) | [Ταϊλανδικά](../th/README.md) | [Τουρκικά](../tr/README.md) | [Ουκρανικά](../uk/README.md) | [Ουρντού](../ur/README.md) | [Βιετναμέζικα](../vi/README.md) -> **Προτιμάτε να κλωνοποιήσετε τοπικά;** +> **Προτιμάτε να Κλωνοποιήσετε Τοπικά;** > -> Αυτό το αποθετήριο περιλαμβάνει πάνω από 50 μεταφράσεις γλωσσών που αυξάνουν σημαντικά το μέγεθος λήψης. Για να κλωνοποιήσετε χωρίς μεταφράσεις, χρησιμοποιήστε sparse checkout: +> Αυτό το αποθετήριο περιλαμβάνει περισσότερες από 50 γλωσσικές μεταφράσεις που αυξάνουν σημαντικά το μέγεθος του download. Για να κλωνοποιήσετε χωρίς τις μεταφράσεις, χρησιμοποιήστε το sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,47 +52,47 @@ > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Αυτό σας δίνει όλα όσα χρειάζεστε για να ολοκληρώσετε το μάθημα με πολύ πιο γρήγορη λήψη. +> Αυτό σας δίνει ό,τι χρειάζεστε για να ολοκληρώσετε το μάθημα με πολύ ταχύτερο download. -## Περιεχόμενα +## Πίνακας Περιεχομένων - Εισαγωγή - - [Καλωσορίσατε στην Οικογένεια Phi](./md/01.Introduction/01/01.PhiFamily.md) - - [Εγκατάσταση περιβάλλοντος](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Κατανόηση των Κύριων Τεχνολογιών](./md/01.Introduction/01/01.Understandingtech.md) - - [Ασφάλεια Τεχνητής Νοημοσύνης για τα Μοντέλα Phi](./md/01.Introduction/01/01.AISafety.md) + - [Καλώς ήρθατε στην Οικογένεια Phi](./md/01.Introduction/01/01.PhiFamily.md) + - [Ρύθμιση του περιβάλλοντος σας](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Κατανόηση βασικών τεχνολογιών](./md/01.Introduction/01/01.Understandingtech.md) + - [Ασφάλεια AI για τα Μοντέλα Phi](./md/01.Introduction/01/01.AISafety.md) - [Υποστήριξη Υλικού Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Μοντέλα Phi & Διαθεσιμότητα σε πλατφόρμες](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Χρήση Guidance-ai και Phi](./md/01.Introduction/01/01.Guidance.md) + - [Μοντέλα Phi & Διαθεσιμότητα σε Πλατφόρμες](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Χρήση του Guidance-ai και Phi](./md/01.Introduction/01/01.Guidance.md) - [Μοντέλα GitHub Marketplace](https://github.com/marketplace/models) - [Κατάλογος Μοντέλων Azure AI](https://ai.azure.com) -- Εκτέλεση Phi σε διαφορετικά περιβάλλοντα +- Εκτέλεση Phi σε διαφορετικό περιβάλλον - [Hugging face](./md/01.Introduction/02/01.HF.md) - [Μοντέλα GitHub](./md/01.Introduction/02/02.GitHubModel.md) - [Κατάλογος Μοντέλων Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [Εργαλειοθήκη AI VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) + - [Τοπικό Foundry](./md/01.Introduction/02/07.FoundryLocal.md) -- Εκτέλεση Φι στην Οικογένεια Phi +- Εκτέλεση Phi Family - [Εκτέλεση Phi σε iOS](./md/01.Introduction/03/iOS_Inference.md) - [Εκτέλεση Phi σε Android](./md/01.Introduction/03/Android_Inference.md) - [Εκτέλεση Phi σε Jetson](./md/01.Introduction/03/Jetson_Inference.md) - [Εκτέλεση Phi σε AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Εκτέλεση Phi με το Πλαίσιο Apple MLX](./md/01.Introduction/03/MLX_Inference.md) - - [Εκτέλεση Phi σε τοπικό διακομιστή](./md/01.Introduction/03/Local_Server_Inference.md) - - [Εκτέλεση Phi σε απομακρυσμένο διακομιστή χρησιμοποιώντας το AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Εκτέλεση Phi με το Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Εκτέλεση Phi σε Τοπικό Διακομιστή](./md/01.Introduction/03/Local_Server_Inference.md) + - [Εκτέλεση Phi σε Απομακρυσμένο Διακομιστή χρησιμοποιώντας την Εργαλειοθήκη AI](./md/01.Introduction/03/Remote_Interence.md) - [Εκτέλεση Phi με Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Εκτέλεση Phi--Vision τοπικά](./md/01.Introduction/03/Vision_Inference.md) + - [Εκτέλεση Phi - Όραμα τοπικά](./md/01.Introduction/03/Vision_Inference.md) - [Εκτέλεση Phi με Kaito AKS, Azure Containers (επίσημη υποστήριξη)](./md/01.Introduction/03/Kaito_Inference.md) -- [Ποσοτικοποίηση της Οικογένειας Phi](./md/01.Introduction/04/QuantifyingPhi.md) +- [Ποσοτικοποίηση Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) - [Ποσοτικοποίηση Phi-3.5 / 4 με χρήση llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - [Ποσοτικοποίηση Phi-3.5 / 4 με χρήση επεκτάσεων Generative AI για onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [Ποσοτικοποίηση Phi-3.5 / 4 με χρήση Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Ποσοτικοποίηση Phi-3.5 / 4 με χρήση του Πλαισίου Apple MLX](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Ποσοτικοποίηση Phi-3.5 / 4 με χρήση Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Αξιολόγηση Phi - [Υπεύθυνη AI](./md/01.Introduction/05/ResponsibleAI.md) @@ -100,225 +100,230 @@ - [Χρήση Promptflow για Αξιολόγηση](./md/01.Introduction/05/Promptflow.md) - RAG με Azure AI Search - - [Πώς να χρησιμοποιήσετε το Phi-4-mini και το Phi-4-multimodal (RAG) με το Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Πώς να χρησιμοποιήσετε το Phi-4-mini και Phi-4-multimodal(RAG) με Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Τοπικό Υβριδικό RAG χωρίς Cloud με SQLite FTS5 και phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Παραδείγματα ανάπτυξης εφαρμογών Phi - Εφαρμογές Κειμένου & Συνομιλίας - - Δείγματα Phi-4 - - [📓] [Συνομιλία με το Μοντέλο Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - Παραδείγματα Phi-4 + - [📓] [Συνομιλία με το μοντέλο Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Συνομιλία με το τοπικό μοντέλο Phi-4 ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Συνομιλία εφαρμογής κονσόλας .NET με Phi-4 ONNX χρησιμοποιώντας Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Δείγματα Phi-3 / 3.5 - - [Τοπικός chatbot στον περιηγητή χρησιμοποιώντας Phi3, ONNX Runtime Web και WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Συνομιλία εφαρμογής κονσόλας .NET με Phi-4 ONNX χρησιμοποιώντας Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Delta-3 / 3.5 Δείγματα + - [Τοπικό Chatbot στο πρόγραμμα περιήγησης χρησιμοποιώντας Phi3, ONNX Runtime Web και WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [Πολλαπλό Μοντέλο - Διαδραστικό Phi-3-mini και OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Δημιουργία wrapper και χρήση του Phi-3 με MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [MLFlow - Δημιουργία wrapper και χρήση Phi-3 με MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [Βελτιστοποίηση Μοντέλου - Πώς να βελτιστοποιήσετε το μοντέλο Phi-3-min για ONNX Runtime Web με Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 Εφαρμογή με Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Δείγμα εφαρμογής Σημειώσεων με πολλαπλά μοντέλα WinUI3 με τεχνητή νοημοσύνη](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Λεπτομερής εκπαίδευση και ενσωμάτωση προσαρμοσμένων μοντέλων Phi-3 με Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Λεπτομερής εκπαίδευση και ενσωμάτωση προσαρμοσμένων μοντέλων Phi-3 με Prompt flow στο Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Αξιολόγηση του λεπτομερώς εκπαιδευμένου μοντέλου Phi-3 / Phi-3.5 στο Microsoft Foundry με έμφαση στις Αρχές Υπεύθυνης AI της Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Δείγμα πρόβλεψης γλώσσας Phi-3.5-mini-instruct (Κινέζικα/Αγγλικά)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Εφαρμογή WinUI3 με Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[Δείγμα εφαρμογής WinUI3 πολλαπλού μοντέλου με AI Powered Σημειώσεις](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Βελτιστοποίηση και Ενσωμάτωση προσαρμοσμένων μοντέλων Phi-3 με Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Βελτιστοποίηση και Ενσωμάτωση προσαρμοσμένων μοντέλων Phi-3 με Prompt flow στο Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Αξιολόγηση του Βελτιστοποιημένου μοντέλου Phi-3 / Phi-3.5 στο Microsoft Foundry εστιάζοντας στις Αρχές Υπεύθυνης AI της Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Delta-3.5-mini-instruct παράδειγμα πρόβλεψης γλώσσας (Κινέζικα/Αγγλικά)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Χρήση Windows GPU για δημιουργία λύσης Prompt flow με Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Χρήση Microsoft Phi-3.5 tflite για δημιουργία εφαρμογής Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET Παράδειγμα χρήσης τοπικού μοντέλου ONNX Phi-3 με το Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Παράδειγμα Q&A .NET με τοπικό μοντέλο ONNX Phi-3 χρησιμοποιώντας το Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - [Κονσόλα συνομιλίας .NET εφαρμογή με Semantic Kernel και Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Δείγματα Κώδικα Βασισμένα σε Azure AI Inference SDK - - Δείγματα Phi-4 - - [📓] [Δημιουργήστε κώδικα έργου χρησιμοποιώντας Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Δείγματα Phi-3 / 3.5 - - [Δημιουργήστε το δικό σας Visual Studio Code GitHub Copilot Chat με την οικογένεια Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Δημιουργήστε τον δικό σας Visual Studio Code Chat Copilot Agent με Phi-3.5 από μοντέλα GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - Δείγματα Προχωρημένης Λογικής - - Δείγματα Phi-4 - - [📓] [Δείγματα Phi-4-mini-reasoning ή Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Λεπτομερής εκπαίδευση Phi-4-mini-reasoning με Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Λεπτομερής εκπαίδευση Phi-4-mini-reasoning με Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning με μοντέλα GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning με μοντέλα Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Παραδείγματα Κώδικα Azure AI Inference SDK + - Delta-4 Δείγματα + - [📓] [Δημιουργία κώδικα έργου χρησιμοποιώντας Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Delta-3 / 3.5 Δείγματα + - [Δημιουργήστε τον δικό σας Visual Studio Code GitHub Copilot Chat με την οικογένεια Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Δημιουργήστε τον δικό σας Visual Studio Code Chat Copilot Agent με Phi-3.5 από Μοντέλα GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - Προηγμένα Παραδείγματα Λογικής + - Delta-4 Δείγματα + - [📓] [Phi-4-mini-λογική ή Phi-4-λογική Παραδείγματα](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Βελτιστοποίηση Phi-4-mini-λογικής με Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Βελτιστοποίηση Phi-4-mini-λογικής με Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-λογική με Μοντέλα GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-λογική με Microsoft Foundry Μοντέλα](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Επιδείξεις - - [Phi-4-mini επιδείξεις φιλοξενούμενες στο Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal επιδείξεις φιλοξενούμενες στο Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Δείγματα Όρασης - - Δείγματα Phi-4 + - [Phi-4-mini επιδείξεις που φιλοξενούνται στο Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal επιδείξεις που φιλοξενούνται στο Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Παραδείγματα Όρασης + - Delta-4 Δείγματα - [📓] [Χρήση Phi-4-multimodal για ανάγνωση εικόνων και δημιουργία κώδικα](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Δείγματα Phi-3 / 3.5 - - [📓][Phi-3-όραση - μετατροπή κειμένου εικόνας σε κείμενο](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-όραση-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-όραση CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - Delta-3 / 3.5 Δείγματα + - [📓][Phi-3-vision-Μετατροπή κειμένου εικόνας σε κείμενο](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vision CLIP Ενσωμάτωση](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [ΕΠΙΔΕΙΞΗ: Phi-3 Ανακύκλωση](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-όραση - Βοηθός οπτικής γλώσσας - με Phi3-Vision και OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Όραση Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Όραση OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Όραση παράδειγμα πολλαπλών καρέ ή πολλαπλών εικόνων](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Όραση Τοπικό μοντέλο ONNX χρησιμοποιώντας το Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Μενού βασισμένο τοπικό μοντέλο ONNX Phi-3 Όραση χρησιμοποιώντας το Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Δείγματα Λογικής-Όρασης - - Phi-4-Λογική-Όραση-15B - - [📓] [Χρήση Phi-4-Λογικής-Όρασης-15B για ανίχνευση παράνομης διάσχισης δρόμου](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Χρήση Phi-4-Λογικής-Όρασης-15B στα μαθηματικά](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [Phi-3-vision - Βοηθός Οπτικής Γλώσσας - με Phi3-Vision και OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision πολλαπλών καρέ ή πολλαπλής εικόνας παράδειγμα](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision Τοπικό Μοντέλο ONNX χρησιμοποιώντας το Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Μενού βασισμένο στο Phi-3 Vision Τοπικό Μοντέλο ONNX χρησιμοποιώντας το Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Παραδείγματα Λογικής-Όρασης + - Phi-4-Λογική-Όρασης-15B + - [📓] [Χρήση Phi-4-Λογικής-Όρασης-15B για ανίχνευση jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Χρήση Phi-4-Λογικής-Όρασης-15B για μαθηματικά](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - [📓] [Χρήση Phi-4-Λογικής-Όρασης-15B για ανίχνευση UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - Δείγματα Μαθηματικών - - Φύλλο εργασίας Φλας Λογικής Phi-4-Mini-Instruct [Επίδειξη Μαθηματικών με Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Μαθηματικά Παραδείγματα + - Delta-4-Mini-Flash-Λογική-Εντολές Παραδείγματα [Επίδειξη Μαθηματικών με Delta-4-Mini-Flash-Λογική-Εντολές](./md/02.Application/09.Math/MathDemo.ipynb) - - Δείγματα Ήχου - - Δείγματα Phi-4 + - Παραδείγματα Ήχου + - Delta-4 Δείγματα - [📓] [Εξαγωγή μεταγραφών ήχου χρησιμοποιώντας Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Δείγμα ήχου Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Δείγμα μετάφρασης ομιλίας Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET κονσολική εφαρμογή που χρησιμοποιεί Phi-4-multimodal Audio για ανάλυση αρχείου ήχου και δημιουργία μεταγραφής](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - Δείγματα MOE - - Δείγματα Phi-3 / 3.5 - - [📓] [Μείγμα Μοντέλων Ειδικών (MoEs) Phi-3.5 Δείγμα Κοινωνικών Μέσων](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Δημιουργία Pipeline Παραγωγής Ενισχυμένης Ανάκτησης (RAG) με NVIDIA NIM Phi-3 MOE, Azure AI Search και LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-4-multimodal Δείγμα Ήχου](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal Δείγμα Μετάφρασης Ομιλίας](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET εφαρμογή κονσόλας χρησιμοποιώντας Phi-4-multimodal ήχο για ανάλυση αρχείου ήχου και δημιουργία μεταγραφής](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE Παραδείγματα + - Delta-3 / 3.5 Δείγματα + - [📓] [Delta-3.5 Μείγμα Ειδικών Μοντέλων (MoEs) Δείγμα Κοινωνικών Δικτύων](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Δημιουργία Retrieval-Augmented Generation (RAG) Pipeline με NVIDIA NIM Phi-3 MOE, Azure AI Search, και LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Δείγματα Κλήσεων Λειτουργιών - - Δείγματα Phi-4 🆕 - - [📓] [Χρήση Κλήσεων Λειτουργιών με Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Χρήση Κλήσεων Λειτουργιών για δημιουργία πολλαπλών πρακτόρων με Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Χρήση Κλήσεων Λειτουργιών με Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Χρήση Κλήσεων Λειτουργιών με ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Δείγματα Μικτής Πολυμορφίας - - Δείγματα Phi-4 🆕 - - [📓] [Χρήση Phi-4-multimodal ως τεχνολογικός δημοσιογράφος](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET κονσολική εφαρμογή που χρησιμοποιεί Phi-4-multimodal για ανάλυση εικόνων](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Λεπτομερής Εκπαίδευση Δειγμάτων Phi - - [Σενάρια Λεπτομερούς Εκπαίδευσης](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Λεπτομερής Εκπαίδευση έναντι RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Αφήστε το Phi-3 να γίνει ειδικός της βιομηχανίας με λεπτομερή εκπαίδευση](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Λεπτομερής Εκπαίδευση Phi-3 με το AI Toolkit για VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Λεπτομερής Εκπαίδευση Phi-3 με την Υπηρεσία Azure Machine Learning](./md/03.FineTuning/Introduce_AzureML.md) - - [Λεπτομερής Εκπαίδευση Phi-3 με Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Λεπτομερής Εκπαίδευση Phi-3 με QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Λεπτομερής Εκπαίδευση Phi-3 με Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Λεπτομερής Εκπαίδευση Phi-3 με Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Λεπτομερής Εκπαίδευση με Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Πρακτικό εργαστήριο λεπτομερούς εκπαίδευσης με Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [Λεπτομερής Εκπαίδευση Phi-3-vision με Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Λεπτομερής Εκπαίδευση Phi-3 με Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Λεπτομερής Εκπαίδευση Phi-3-vision (επίσημη υποστήριξη)](./md/03.FineTuning/FineTuning_Vision.md) - - [Βελτιστοποίηση Phi-3 με Kaito AKS, Azure Containers(επίσημη Υποστήριξη)](./md/03.FineTuning/FineTuning_Kaito.md) + - Παραδείγματα Κλήσης Συναρτήσεων + - Delta-4 Δείγματα 🆕 + - [📓] [Χρήση Κλήσης Συναρτήσεων με Delta-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Χρήση Κλήσης Συναρτήσεων για δημιουργία multi-agents με Delta-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Χρήση Κλήσης Συναρτήσεων με Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Χρήση Κλήσης Συναρτήσεων με ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Παραδείγματα Αναμείξεων Multimodal + - Delta-4 Δείγματα 🆕 + - [📓] [Χρήση Delta-4-multimodal ως Τεχνολογικού δημοσιογράφου](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET εφαρμογή κονσόλας χρησιμοποιώντας Delta-4-multimodal για ανάλυση εικόνων](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Βελτιστοποίηση Delta Παραδειγμάτων + - [Σενάρια Βελτιστοποίησης](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Βελτιστοποίηση έναντι RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Άφησε το Delta-3 να γίνει ειδικός της βιομηχανίας](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Βελτιστοποίηση Delta-3 με AI Toolkit για VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Βελτιστοποίηση Delta-3 με Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Βελτιστοποίηση Delta-3 με Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Βελτιστοποίηση Delta-3 με QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Βελτιστοποίηση Delta-3 με Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Βελτιστοποίηση Delta-3 με Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Βελτιστοποίηση με Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Βελτιστοποίηση με Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) + - [Βελτιστοποίηση Delta-3-όραση με Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Βελτιστοποίηση Phi-3 με το πλαίσιο Apple MLX](./md/03.FineTuning/FineTuning_MLX.md) + - [Βελτιστοποίηση Phi-3-vision (επίσημη υποστήριξη)](./md/03.FineTuning/FineTuning_Vision.md) + - [Βελτιστοποίηση Phi-3 με Kaito AKS, Azure Containers (επίσημη υποστήριξη)](./md/03.FineTuning/FineTuning_Kaito.md) - [Βελτιστοποίηση Phi-3 και 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) -- Εργαστηριακή Εξάσκηση - - [Εξερεύνηση προηγμένων μοντέλων: LLMs, SLMs, τοπική ανάπτυξη και άλλα](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Δενδροκοπία Δυνατοτήτων NLP: Βελτιστοποίηση με Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) +- Πρακτικό Εργαστήριο + - [Εξερεύνηση μοντέλων αιχμής: LLMs, SLMs, τοπική ανάπτυξη και πολλά άλλα](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Απελευθέρωση δυνατοτήτων NLP: Βελτιστοποίηση με Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - Ακαδημαϊκά Ερευνητικά Άρθρα και Δημοσιεύσεις - - [Τα Βιβλία Ειναι Όλα Όσα Χρειάζεστε ΙΙ: τεχνική αναφορά phi-1.5](https://arxiv.org/abs/2309.05463) - - [Τεχνική Αναφορά Phi-3: Ένα Ισχυρό Μοντέλο Γλώσσας Τοπικά στο Τηλέφωνό Σας](https://arxiv.org/abs/2404.14219) - - [Τεχνική Αναφορά Phi-4](https://arxiv.org/abs/2412.08905) - - [Τεχνική Αναφορά Phi-4-Mini: Συμπαγή αλλά Ισχυρά Πολυμορφικά Μοντέλα Γλώσσας μέσω Μείγματος LoRAs](https://arxiv.org/abs/2503.01743) - - [Βελτιστοποίηση Μικρών Μοντέλων Γλώσσας για Κλήσεις Λειτουργιών Εντός Οχήματος](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Βελτιστοποίηση PHI-3 για Απαντήσεις Πολλαπλής Επιλογής: Μεθοδολογία, Αποτελέσματα και Προκλήσεις](https://arxiv.org/abs/2501.01588) - - [Τεχνική Αναφορά Συλλογισμού Phi-4](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Τεχνική Αναφορά Συλλογισμού Phi-4-mini](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Textbooks Are All You Need II: phi-1.5 τεχνική αναφορά](https://arxiv.org/abs/2309.05463) + - [Phi-3 Τεχνική Αναφορά: Πολύ ικανό γλωσσικό μοντέλο τοπικά στην συσκευή σας](https://arxiv.org/abs/2404.14219) + - [Phi-4 Τεχνική Αναφορά](https://arxiv.org/abs/2412.08905) + - [Phi-4-Mini Τεχνική Αναφορά: Συμπαγή αλλά ισχυρά πολυμορφικά γλωσσικά μοντέλα μέσω Μείγματος-LoRA](https://arxiv.org/abs/2503.01743) + - [Βελτιστοποίηση μικρών γλωσσικών μοντέλων για κλήση λειτουργιών εντός οχήματος](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Βελτιστοποίηση PHI-3 για απάντηση σε ερωτήσεις πολλαπλής επιλογής: Μεθοδολογία, Αποτελέσματα και Προκλήσεις](https://arxiv.org/abs/2501.01588) + - [Phi-4-reasoning Τεχνική Αναφορά](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-reasoning Τεχνική Αναφορά](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Χρήση Μοντέλων Phi ### Phi στο Microsoft Foundry -Μπορείτε να μάθετε πώς να χρησιμοποιείτε το Microsoft Phi και πώς να δημιουργείτε ολοκληρωμένες λύσεις στα διάφορα υλικά σας. Για να ζήσετε την εμπειρία με το Phi, ξεκινήστε παίζοντας με τα μοντέλα και προσαρμόζοντας το Phi για τα σενάριά σας χρησιμοποιώντας τον [Κατάλογο Μοντέλων Azure AI Microsoft Foundry](https://aka.ms/phi3-azure-ai). Μπορείτε να μάθετε περισσότερα στο Ξεκινώντας με το [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Μπορείτε να μάθετε πώς να χρησιμοποιείτε το Microsoft Phi και πώς να δημιουργείτε ολοκληρωμένες λύσεις στις διάφορες συσκευές σας. Για να δοκιμάσετε το Phi μόνοι σας, ξεκινήστε παίζοντας με τα μοντέλα και προσαρμόζοντας το Phi για τα σενάριά σας χρησιμοποιώντας τον [Κατάλογο Μοντέλων Azure AI του Microsoft Foundry](https://aka.ms/phi3-azure-ai). Μπορείτε να μάθετε περισσότερα στην Εισαγωγή με το [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Παιχνιδότοπος** -Κάθε μοντέλο διαθέτει έναν αφιερωμένο παιχνιδότοπο για να δοκιμάσετε το μοντέλο [Azure AI Playground](https://aka.ms/try-phi3). +**Παιδική Χαρά** +Κάθε μοντέλο έχει μία αφιερωμένη παιδική χαρά για να δοκιμάσετε το μοντέλο [Azure AI Playground](https://aka.ms/try-phi3). -### Phi στα Μοντέλα GitHub +### Phi στα μοντέλα GitHub -Μπορείτε να μάθετε πώς να χρησιμοποιείτε το Microsoft Phi και πώς να δημιουργείτε ολοκληρωμένες λύσεις στα διάφορα υλικά σας. Για να ζήσετε την εμπειρία με το Phi, ξεκινήστε παίζοντας με το μοντέλο και προσαρμόζοντας το Phi για τα σενάριά σας χρησιμοποιώντας τον [Κατάλογο Μοντέλων GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Μπορείτε να μάθετε περισσότερα στο Ξεκινώντας με τον [Κατάλογο Μοντέλων GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) +Μπορείτε να μάθετε πώς να χρησιμοποιείτε το Microsoft Phi και πώς να δημιουργείτε ολοκληρωμένες λύσεις στις διάφορες συσκευές σας. Για να δοκιμάσετε το Phi μόνοι σας, ξεκινήστε παίζοντας με το μοντέλο και προσαρμόζοντας το Phi για τα σενάριά σας χρησιμοποιώντας τον [Κατάλογο Μοντέλων GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Μπορείτε να μάθετε περισσότερα στην Εισαγωγή με τον [Κατάλογο Μοντέλων GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Παιχνιδότοπος** -Κάθε μοντέλο έχει έναν αφιερωμένο [παιχνιδότοπο για να δοκιμάσετε το μοντέλο](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Παιδική Χαρά** +Κάθε μοντέλο έχει μία αφιερωμένη [παιδική χαρά για να δοκιμάσετε το μοντέλο](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi στο Hugging Face Μπορείτε επίσης να βρείτε το μοντέλο στο [Hugging Face](https://huggingface.co/microsoft) -**Παιχνιδότοπος** -[Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Παιδική Χαρά** + [Παιδική χαρά Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) ## 🎒 Άλλα Μαθήματα -Η ομάδα μας παράγει και άλλα μαθήματα! Δείτε: +Η ομάδα μας παράγει και άλλα μαθήματα! Ρίξτε μια ματιά: -### LangChain -[![LangChain4j για Αρχάριους](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js για Αρχάριους](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain για Αρχάριους](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +### LangChain +[![LangChain4j για Αρχάριους](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js για Αρχάριους](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain για Αρχάριους](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agents -[![AZD για Αρχάριους](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI για Αρχάριους](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP για Αρχάριους](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents για Αρχάριους](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agents +[![AZD για Αρχάριους](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI για Αρχάριους](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP για Αρχάριους](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents για Αρχάριους](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Σειρά Γενετικής Τεχνητής Νοημοσύνης -[![Γενετική Τεχνητή Νοημοσύνη για Αρχάριους](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Γενετική Τεχνητή Νοημοσύνη (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Γενετική Τεχνητή Νοημοσύνη (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Γενετική Τεχνητή Νοημοσύνη (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### Σειρά Γενετικής Τεχνητής Νοημοσύνης +[![Γενετική Τεχνητή Νοημοσύνη για Αρχάριους](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Γενετική Τεχνητή Νοημοσύνη (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Γενετική Τεχνητή Νοημοσύνη (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Γενετική Τεχνητή Νοημοσύνη (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Πυρήνας Μάθησης -[![Μηχανική Μάθηση για Αρχάριους](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Επιστήμη Δεδομένων για Αρχάριους](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![Τεχνητή Νοημοσύνη για Αρχάριους](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Κυβερνοασφάλεια για Αρχάριους](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Ανάπτυξη Web για Αρχάριους](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT για Αρχάριους](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![Ανάπτυξη XR για Αρχάριους](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) + +### Βασική Εκμάθηση +[![Μηχανική Μάθηση για Αρχάριους](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Επιστήμη Δεδομένων για Αρχάριους](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![Τεχνητή Νοημοσύνη για Αρχάριους](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Κυβερνοασφάλεια για Αρχάριους](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Ανάπτυξη Web για Αρχάριους](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT για Αρχάριους](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR Ανάπτυξη για Αρχάριους](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Σειρά Copilot -[![Copilot για Συνεργατικό Προγραμματισμό AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot για C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Περιπέτεια Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + +### Σειρά Copilot +[![Copilot για AI Ζευγαρωτό Προγραμματισμό](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot για C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Περιπέτεια Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Υπεύθυνη Τεχνητή Νοημοσύνη +## Υπεύθυνη Τεχνητή Νοημοσύνη + +Η Microsoft δεσμεύεται να βοηθά τους πελάτες μας να χρησιμοποιούν τα προϊόντα τεχνητής νοημοσύνης με υπευθυνότητα, να μοιράζεται τα συμπεράσματά μας και να δημιουργεί σχέσεις εμπιστοσύνης μέσω εργαλείων όπως οι Σημειώσεις Διαφάνειας και οι Αξιολογήσεις Επιπτώσεων. Πολλοί από αυτούς τους πόρους είναι διαθέσιμοι στο [https://aka.ms/RAI](https://aka.ms/RAI). +Η προσέγγιση της Microsoft για την υπεύθυνη τεχνητή νοημοσύνη βασίζεται στις αρχές τεχνητής νοημοσύνης μας για δικαιοσύνη, αξιοπιστία και ασφάλεια, ιδιωτικότητα και ασφάλεια, συμπερίληψη, διαφάνεια και λογοδοσία. + +Τα μεγάλης κλίμακας μοντέλα φυσικής γλώσσας, εικόνας και ομιλίας - όπως αυτά που χρησιμοποιούνται σε αυτό το παράδειγμα - ενδέχεται να συμπεριφέρονται με τρόπους που είναι άδικοι, αναξιόπιστοι ή προσβλητικοί, προκαλώντας έτσι ζημίες. Παρακαλούμε ανατρέξτε στη [Σημείωση Διαφάνειας της υπηρεσίας Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) για να ενημερωθείτε για τους κινδύνους και τους περιορισμούς. -Η Microsoft δεσμεύεται να βοηθάει τους πελάτες της να χρησιμοποιούν τα προϊόντα ΤΝ υπεύθυνα, μοιράζοντας τα μαθήματα που έχουμε μάθει και χτίζοντας συνεργασίες βασισμένες στην εμπιστοσύνη μέσω εργαλείων όπως Σημειώσεις Διαφάνειας και Αξιολογήσεις Επιπτώσεων. Πολλοί από αυτούς τους πόρους είναι διαθέσιμοι στο [https://aka.ms/RAI](https://aka.ms/RAI). -Η προσέγγιση της Microsoft για την υπεύθυνη ΤΝ βασίζεται στις αρχές μας για την ΤΝ που είναι η δικαιοσύνη, η αξιοπιστία και ασφάλεια, η προστασία της ιδιωτικότητας και η ασφάλεια, η συμπερίληψη, η διαφάνεια και η λογοδοσία. -Τα μεγάλα μοντέλα φυσικής γλώσσας, εικόνας και ομιλίας - όπως αυτά που χρησιμοποιούνται σε αυτό το δείγμα - μπορούν να συμπεριφερθούν με τρόπους που είναι άδικοι, αναξιόπιστοι ή προσβλητικοί, κάτι που μπορεί να προκαλέσει βλάβες. Παρακαλούμε συμβουλευτείτε τη [Σημείωση Διαφάνειας της υπηρεσίας Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) για να ενημερωθείτε σχετικά με τους κινδύνους και περιορισμούς. -Η συνιστώμενη προσέγγιση για την αντιμετώπιση αυτών των κινδύνων είναι η ένταξη ενός συστήματος ασφαλείας στην αρχιτεκτονική σας που μπορεί να ανιχνεύει και να αποτρέπει επιβλαβείς συμπεριφορές. Το [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) παρέχει ένα ανεξάρτητο επίπεδο προστασίας, ικανό να ανιχνεύει επιβλαβές περιεχόμενο που δημιουργείται από χρήστες και AI σε εφαρμογές και υπηρεσίες. Το Azure AI Content Safety περιλαμβάνει API για κείμενο και εικόνα που σας επιτρέπουν να εντοπίζετε υλικό που είναι επιβλαβές. Μέσα στο Microsoft Foundry, η υπηρεσία Content Safety σας επιτρέπει να βλέπετε, να εξερευνάτε και να δοκιμάζετε δείγματα κώδικα για την ανίχνευση επιβλαβούς περιεχομένου σε διάφορες μορφές. Η ακόλουθη [τεκμηρίωση γρήγορης εκκίνησης](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) σας καθοδηγεί στη δημιουργία αιτημάτων προς την υπηρεσία. +Η συνιστώμενη προσέγγιση για τη μείωση αυτών των κινδύνων είναι η ενσωμάτωση ενός συστήματος ασφάλειας στην αρχιτεκτονική σας που μπορεί να εντοπίζει και να αποτρέπει επιβλαβή συμπεριφορά. Το [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) παρέχει ένα ανεξάρτητο επίπεδο προστασίας, ικανό να εντοπίζει επιβλαβές περιεχόμενο που δημιουργείται από χρήστες και AI σε εφαρμογές και υπηρεσίες. Το Azure AI Content Safety περιλαμβάνει APIs κειμένου και εικόνων που σας επιτρέπουν να εντοπίζετε υλικό που είναι επιβλαβές. Μέσα στο Microsoft Foundry, η υπηρεσία Content Safety σας επιτρέπει να προβάλετε, να εξερευνήσετε και να δοκιμάσετε δείγματα κώδικα για τον εντοπισμό επιβλαβούς περιεχομένου σε διαφορετικές μορφές. Η ακόλουθη [τεκμηρίωση γρήγορης εκκίνησης](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) σας καθοδηγεί μέσω της αποστολής αιτημάτων στην υπηρεσία. -Ένας ακόμη παράγοντας που πρέπει να ληφθεί υπόψη είναι η συνολική απόδοση της εφαρμογής. Σε εφαρμογές πολυτροπικές και πολυμοντέλων, θεωρούμε ότι η απόδοση σημαίνει ότι το σύστημα λειτουργεί όπως εσείς και οι χρήστες σας περιμένετε, συμπεριλαμβανομένου να μην παράγει επιβλαβή αποτελέσματα. Είναι σημαντικό να αξιολογήσετε την απόδοση της συνολικής εφαρμογής σας χρησιμοποιώντας τους [αξιολογητές Απόδοσης, Ποιότητας, Κινδύνου και Ασφάλειας](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Επιπλέον, έχετε τη δυνατότητα να δημιουργήσετε και να αξιολογήσετε με [προσαρμοσμένους αξιολογητές](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Ένα άλλο σημείο που πρέπει να λάβετε υπόψη είναι η συνολική απόδοση της εφαρμογής. Σε πολυμορφικές εφαρμογές και εφαρμογές με πολλαπλά μοντέλα, η απόδοση θεωρείται ότι σημαίνει ότι το σύστημα λειτουργεί όπως εσείς και οι χρήστες σας περιμένετε, συμπεριλαμβανομένου του να μην δημιουργεί επιβλαβή αποτελέσματα. Είναι σημαντικό να αξιολογήσετε την απόδοση της συνολικής εφαρμογής σας χρησιμοποιώντας [εκτιμητές Απόδοσης και Ποιότητας και Κινδύνου και Ασφάλειας](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Έχετε επίσης τη δυνατότητα να δημιουργήσετε και να αξιολογήσετε με [προσαρμοσμένους εκτιμητές](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Μπορείτε να αξιολογήσετε την AI εφαρμογή σας στο περιβάλλον ανάπτυξής σας χρησιμοποιώντας το [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Είτε με ένα σύνολο δεδομένων δοκιμής είτε με έναν στόχο, οι γεννήτριες της εφαρμογής τεχνητής νοημοσύνης σας μετρώνται ποσοτικά με ενσωματωμένους ή προσαρμοσμένους αξιολογητές της επιλογής σας. Για να ξεκινήσετε με το azure ai evaluation sdk και να αξιολογήσετε το σύστημά σας, μπορείτε να ακολουθήσετε τον [οδηγό γρήγορης εκκίνησης](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Αφού εκτελέσετε έναν κύκλο αξιολόγησης, μπορείτε να [οπτικοποιήσετε τα αποτελέσματα στο Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Μπορείτε να αξιολογήσετε την AI εφαρμογή σας στο περιβάλλον ανάπτυξής σας χρησιμοποιώντας το [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Δίνοντας είτε ένα σετ δοκιμαστικών δεδομένων είτε έναν στόχο, οι γεννήσεις της γενετικής AI εφαρμογής σας μετρώνται ποσοτικά με ενσωματωμένους εκτιμητές ή προσαρμοσμένους εκτιμητές της επιλογής σας. Για να ξεκινήσετε με το azure ai evaluation sdk για την αξιολόγηση του συστήματός σας, μπορείτε να ακολουθήσετε τον [οδηγό γρήγορης εκκίνησης](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Μόλις εκτελέσετε μια αξιολόγηση, μπορείτε να [οπτικοποιήσετε τα αποτελέσματα στο Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Εμπορικά Σήματα -Αυτό το έργο ενδέχεται να περιέχει εμπορικά σήματα ή λογότυπα για έργα, προϊόντα ή υπηρεσίες. Η εξουσιοδοτημένη χρήση των εμπορικών σημάτων ή λογότυπων της Microsoft υπόκειται στους και πρέπει να ακολουθεί τις [Οδηγίες για Εμπορικά Σήματα & Μάρκες της Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Η χρήση εμπορικών σημάτων ή λογότυπων της Microsoft σε τροποποιημένες εκδόσεις αυτού του έργου δεν πρέπει να προκαλεί σύγχυση ή να υπονοεί χορηγία από τη Microsoft. Οποιαδήποτε χρήση εμπορικών σημάτων ή λογότυπων τρίτων υπόκειται στις πολιτικές αυτών των τρίτων. +Αυτό το έργο μπορεί να περιέχει εμπορικά σήματα ή λογότυπα για έργα, προϊόντα ή υπηρεσίες. Η εξουσιοδοτημένη χρήση των εμπορικών σημάτων ή λογοτύπων της Microsoft υπόκειται και πρέπει να ακολουθεί τις [Οδηγίες Εμπορικών Σημάτων & Επωνυμίας της Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Η χρήση των εμπορικών σημάτων ή λογοτύπων της Microsoft σε τροποποιημένες εκδόσεις αυτού του έργου δεν πρέπει να προκαλεί σύγχυση ή να υποδηλώνει χορηγία από τη Microsoft. Οποιαδήποτε χρήση εμπορικών σημάτων ή λογοτύπων τρίτων υπόκειται στις πολιτικές αυτών των τρίτων. ## Λήψη Βοήθειας -Εάν κολλήσετε ή έχετε ερωτήσεις σχετικά με την κατασκευή εφαρμογών AI, ενώστε: +Εάν κολλήσετε ή έχετε ερωτήσεις σχετικά με την κατασκευή εφαρμογών AI, συμμετέχετε: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Εάν έχετε σχόλια προϊόντος ή λάθη κατά την κατασκευή, επισκεφτείτε: +Εάν έχετε σχόλια προϊόντος ή λάθη κατά την κατασκευή, επισκεφθείτε: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Αποποίηση Ευθυνών**: -Αυτό το έγγραφο έχει μεταφραστεί χρησιμοποιώντας την υπηρεσία μετάφρασης με ΤΝ [Co-op Translator](https://github.com/Azure/co-op-translator). Αν και προσπαθούμε για ακρίβεια, παρακαλούμε να έχετε υπόψη ότι οι αυτοματοποιημένες μεταφράσεις ενδέχεται να περιέχουν λάθη ή ανακρίβειες. Το πρωτότυπο έγγραφο στη μητρική του γλώσσα πρέπει να θεωρείται η αξιόπιστη πηγή. Για κρίσιμες πληροφορίες, συνιστάται επαγγελματική ανθρώπινη μετάφραση. Δεν φέρουμε ευθύνη για οποιεσδήποτε παρερμηνείες ή παρανοήσεις προκύψουν από τη χρήση αυτής της μετάφρασης. +**Αποποίηση ευθυνών**: +Αυτό το έγγραφο έχει μεταφραστεί χρησιμοποιώντας την υπηρεσία μετάφρασης με τεχνητή νοημοσύνη [Co-op Translator](https://github.com/Azure/co-op-translator). Ενώ επιδιώκουμε την ακρίβεια, παρακαλούμε να έχετε υπόψη ότι οι αυτοματοποιημένες μεταφράσεις ενδέχεται να περιέχουν λάθη ή ανακρίβειες. Το πρωτότυπο έγγραφο στη μητρική του γλώσσα πρέπει να θεωρείται η αυθεντική πηγή. Για κρίσιμες πληροφορίες, συνιστάται επαγγελματική ανθρώπινη μετάφραση. Δεν φέρουμε ευθύνη για τυχόν παρεξηγήσεις ή λανθασμένες ερμηνείες που προκύπτουν από τη χρήση αυτής της μετάφρασης. \ No newline at end of file diff --git a/translations/el/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/el/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..a5c454ffe --- /dev/null +++ b/translations/el/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Υβριδικό RAG με Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Συντάκτης:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Εστίαση:** Μικρά Γλωσσικά Μοντέλα (SLMs), SQLite FTS5 BM25, Πυκνά Ενσωματώματα, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Κίνητρο: Το Δίλημμα Επαναφοράς Λέξεων-κλειδιών στα Τοπικά SLMs\n", + "Τα τυπικά αρχιτεκτονικά RAG που βασίζονται αποκλειστικά σε πυκνά ενσωματώματα διανυσμάτων συχνά αποτυγχάνουν να ανακτήσουν ακριβείς αριθμητικούς κωδικούς (π.χ., `2,340,000 TL`, κωδικοί συμβολαίων, αριθμοί λογαριασμών). \n", + "Αντιθέτως, η αραιή λεξική αναζήτηση (BM25) χάνει σημασιολογικά συνώνυμα και παραφρασμένες ερωτήσεις.\n", + "\n", + "Αυτός ο οδηγός δείχνει πώς να υλοποιήσετε μια **υψηλής ταχύτητας, εντός μνήμης υβριδική μηχανή ανάκτησης** που συνδυάζει:\n", + "1. **Πυκνά Διανύσματα** (Ομοιότητα συνημίτονου)\n", + "2. **Αραιή Λεξική Αναζήτηση** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Παραγωγή Τεκμηριωμένων Παραπομπών (`[1]`, `[2]`)** με Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Διπλό Σχήμα SQLite (Συμπαγείς Διανύσματα + Εικονικός Πίνακας FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Δείγμα Κατανάλωσης & Δοκιμασία Εκτέλεσης\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Διατύπωση Βασισμένη σε Συμφραζόμενα για το Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Αποποίηση ευθυνών**:\nΑυτό το έγγραφο έχει μεταφραστεί χρησιμοποιώντας την υπηρεσία μετάφρασης με τεχνητή νοημοσύνη [Co-op Translator](https://github.com/Azure/co-op-translator). Ενώ επιδιώκουμε την ακρίβεια, παρακαλούμε να έχετε υπόψη ότι οι αυτοματοποιημένες μεταφράσεις ενδέχεται να περιέχουν λάθη ή ανακρίβειες. Το πρωτότυπο έγγραφο στη μητρική του γλώσσα πρέπει να θεωρείται η αυθεντική πηγή. Για κρίσιμες πληροφορίες, συνιστάται επαγγελματική ανθρώπινη μετάφραση. Δεν φέρουμε ευθύνη για τυχόν παρεξηγήσεις ή λανθασμένες ερμηνείες που προκύπτουν από τη χρήση αυτής της μετάφρασης.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/en/.co-op-translator.json b/translations/en/.co-op-translator.json index 0ba892d41..c9944970d 100644 --- a/translations/en/.co-op-translator.json +++ b/translations/en/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "en" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T07:31:26+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T16:02:38+00:00", "source_file": "README.md", "language_code": "en" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "en" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T16:11:51+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "en" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:29:15+00:00", diff --git a/translations/en/README.md b/translations/en/README.md index 79bc2f3c1..281e5131a 100644 --- a/translations/en/README.md +++ b/translations/en/README.md @@ -14,13 +14,13 @@ [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi is a series of open source AI models developed by Microsoft. +Phi is a series of open source AI models developed by Microsoft. -Phi is currently the most powerful and cost-effective small language model (SLM), with very good benchmarks in multi-language, reasoning, text/chat generation, coding, images, audio and other scenarios. +Phi is currently the most powerful and cost-effective small language model (SLM), with very good benchmarks in multi-language, reasoning, text/chat generation,coding, images, audio and other scenarios. You can deploy Phi to the cloud or to edge devices, and you can easily build generative AI applications with limited computing power. -Follow these steps to get started using these resource: +Follow these steps to get started using these resource : 1. **Fork the Repository**: Click [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Clone the Repository**: `git clone https://github.com/microsoft/PhiCookBook.git` 3. [**Join The Microsoft AI Discord Community and meet experts and fellow developers**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) @@ -101,6 +101,7 @@ Follow these steps to get started using these resource: - RAG with Azure AI Search - [How to use Phi-4-mini and Phi-4-multimodal(RAG) with Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud Local Hybrid RAG with SQLite FTS5 and phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Phi application development samples - Text & Chat Applications @@ -108,6 +109,7 @@ Follow these steps to get started using these resource: - [📓] [Chat With Phi-4-mini ONNX Model](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Chat with Phi-4 local ONNX Model .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - [Chat .NET Console App with Phi-4 ONNX using Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 Samples - [Local Chatbot in the browser using Phi3, ONNX Runtime Web and WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) @@ -203,6 +205,7 @@ Follow these steps to get started using these resource: - [Fine-tuning with Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Fine-tuning with Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - [Fine-tuning Phi-3-vision with Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Fine-tuning Phi-3 with Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - [Fine-tuning Phi-3-vision (official support)](./md/03.FineTuning/FineTuning_Vision.md) - [Fine-Tuning Phi-3 with Kaito AKS , Azure Containers(official Support)](./md/03.FineTuning/FineTuning_Kaito.md) @@ -295,6 +298,8 @@ Microsoft is committed to helping our customers use our AI products responsibly, Microsoft’s approach to responsible AI is grounded in our AI principles of fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability. Large-scale natural language, image, and speech models - like the ones used in this sample - can potentially behave in ways that are unfair, unreliable, or offensive, in turn causing harms. Please consult the [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) to be informed about risks and limitations. + + The recommended approach to mitigating these risks is to include a safety system in your architecture that can detect and prevent harmful behavior. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) provides an independent layer of protection, able to detect harmful user-generated and AI-generated content in applications and services. Azure AI Content Safety includes text and image APIs that allow you to detect material that is harmful. Within Microsoft Foundry, the Content Safety service allows you to view, explore and try out sample code for detecting harmful content across different modalities. The following [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) guides you through making requests to the service. Another aspect to take into account is the overall application performance. With multi-modal and multi-models applications, we consider performance to mean that the system performs as you and your users expect, including not generating harmful outputs. It's important to assess the performance of your overall application using [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). You also have the ability to create and evaluate with [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). diff --git a/translations/en/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/en/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..cfe9106bc --- /dev/null +++ b/translations/en/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrid RAG with Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Author:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Focus:** Small Language Models (SLMs), SQLite FTS5 BM25, Dense Embeddings, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivation: The Keyword Recall Dilemma in Local SLMs\n", + "Standard RAG architectures relying purely on dense vector embeddings often fail to retrieve exact numerical tokens (e.g., `2,340,000 TL`, contract codes, account numbers). \n", + "Conversely, sparse lexical search (BM25) misses semantic synonyms and paraphrased questions.\n", + "\n", + "This cookbook demonstrates how to implement a **high-speed, in-memory Hybrid Retrieval engine** combining:\n", + "1. **Dense Vectors** (Cosine Similarity)\n", + "2. **Sparse Lexical Search** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Grounded Citation Generation (`[1]`, `[2]`)** with Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Dual SQLite Schema (Dense Vectors + Virtual FTS5 BM25 Table)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Sample Ingestion & Execution Benchmark\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Grounded Prompt Formulation for Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Disclaimer**:\nThis document has been translated using AI translation service [Co-op Translator](https://github.com/Azure/co-op-translator). While we strive for accuracy, please be aware that automated translations may contain errors or inaccuracies. The original document in its native language should be considered the authoritative source. For critical information, professional human translation is recommended. We are not liable for any misunderstandings or misinterpretations arising from the use of this translation.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/es/.co-op-translator.json b/translations/es/.co-op-translator.json index 0e9ad3766..60cd7cf25 100644 --- a/translations/es/.co-op-translator.json +++ b/translations/es/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "es" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T07:36:59+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T16:11:27+00:00", "source_file": "README.md", "language_code": "es" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "es" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T16:12:27+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "es" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:29:23+00:00", diff --git a/translations/es/README.md b/translations/es/README.md index 224f92537..dafec33f6 100644 --- a/translations/es/README.md +++ b/translations/es/README.md @@ -1,42 +1,42 @@ -# Phi Cookbook: Ejemplos Prácticos con los Modelos Phi de Microsoft +# Recetario Phi: Ejemplos prácticos con los modelos Phi de Microsoft [![Abrir y usar los ejemplos en GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Abrir en Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Abrir en Contenedores Dev](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![Colaboradores de GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![Problemas en GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Solicitudes de extracción en GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![Solicitudes de extracción bienvenidas](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Colaboradores en GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![Incidencias en GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![Pull requests en GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![Pull requests Bienvenidas](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![Seguidores en GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![Bifurcaciones en GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![Estrellas en GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) -[![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) +[![Discord de Microsoft Foundry](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi es una serie de modelos de inteligencia artificial de código abierto desarrollados por Microsoft. +Phi es una serie de modelos de IA de código abierto desarrollados por Microsoft. -Phi es actualmente el modelo de lenguaje pequeño (SLM) más potente y rentable, con muy buenos resultados en múltiples idiomas, razonamiento, generación de texto/chat, codificación, imágenes, audio y otros escenarios. +Phi es actualmente el modelo de lenguaje pequeño (SLM) más potente y rentable, con muy buenos resultados en múltiples idiomas, razonamiento, generación de texto/chat, codificación, imágenes, audio y otros escenarios. -Puedes desplegar Phi en la nube o en dispositivos perimetrales, y puedes construir fácilmente aplicaciones de IA generativa con potencia computacional limitada. +Puedes implementar Phi en la nube o en dispositivos de borde, y puedes construir fácilmente aplicaciones de IA generativa con potencia informática limitada. -Sigue estos pasos para empezar a usar estos recursos: -1. **Haz un fork del repositorio**: Haz clic en [![Bifurcaciones en GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Clona el repositorio**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Únete a la comunidad de Discord de Microsoft AI y conoce a expertos y otros desarrolladores**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +Sigue estos pasos para comenzar a usar estos recursos: +1. **Bifurca el Repositorio**: Haz clic en [![Bifurcaciones en GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Clona el Repositorio**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Únete a la Comunidad de Microsoft AI en Discord y conoce a expertos y otros desarrolladores**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/es/cover.eb18d1b9605d754b.webp) +![portada](../../translated_images/es/cover.eb18d1b9605d754b.webp) ### 🌐 Soporte Multilingüe -#### Soportado mediante GitHub Action (Automatizado y Siempre Actualizado) +#### Soportado a través de GitHub Action (Automatizado y Siempre Actualizado) -[Árabe](../ar/README.md) | [Bengalí](../bn/README.md) | [Búlgaro](../bg/README.md) | [Birmano (Myanmar)](../my/README.md) | [Chino (Simplificado)](../zh-CN/README.md) | [Chino (Tradicional, Hong Kong)](../zh-HK/README.md) | [Chino (Tradicional, Macao)](../zh-MO/README.md) | [Chino (Tradicional, Taiwán)](../zh-TW/README.md) | [Croata](../hr/README.md) | [Checo](../cs/README.md) | [Danés](../da/README.md) | [Holandés](../nl/README.md) | [Estonio](../et/README.md) | [Finlandés](../fi/README.md) | [Francés](../fr/README.md) | [Alemán](../de/README.md) | [Griego](../el/README.md) | [Hebreo](../he/README.md) | [Hindi](../hi/README.md) | [Húngaro](../hu/README.md) | [Indonesio](../id/README.md) | [Italiano](../it/README.md) | [Japonés](../ja/README.md) | [Kannada](../kn/README.md) | [Jemer](../km/README.md) | [Coreano](../ko/README.md) | [Lituano](../lt/README.md) | [Malayo](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalí](../ne/README.md) | [Pidgin Nigeriano](../pcm/README.md) | [Noruego](../no/README.md) | [Persa (Farsi)](../fa/README.md) | [Polaco](../pl/README.md) | [Portugués (Brasil)](../pt-BR/README.md) | [Portugués (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumano](../ro/README.md) | [Ruso](../ru/README.md) | [Serbio (Cirílico)](../sr/README.md) | [Eslovaco](../sk/README.md) | [Esloveno](../sl/README.md) | [Español](./README.md) | [Swahili](../sw/README.md) | [Sueco](../sv/README.md) | [Tagalo (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Tailandés](../th/README.md) | [Turco](../tr/README.md) | [Ucraniano](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamita](../vi/README.md) +[Árabe](../ar/README.md) | [Bengalí](../bn/README.md) | [Búlgaro](../bg/README.md) | [Birmano (Myanmar)](../my/README.md) | [Chino (Simplificado)](../zh-CN/README.md) | [Chino (Tradicional, Hong Kong)](../zh-HK/README.md) | [Chino (Tradicional, Macao)](../zh-MO/README.md) | [Chino (Tradicional, Taiwán)](../zh-TW/README.md) | [Croata](../hr/README.md) | [Checo](../cs/README.md) | [Danés](../da/README.md) | [Neerlandés](../nl/README.md) | [Estonio](../et/README.md) | [Finés](../fi/README.md) | [Francés](../fr/README.md) | [Alemán](../de/README.md) | [Griego](../el/README.md) | [Hebreo](../he/README.md) | [Hindi](../hi/README.md) | [Húngaro](../hu/README.md) | [Indonesio](../id/README.md) | [Italiano](../it/README.md) | [Japonés](../ja/README.md) | [Kannada](../kn/README.md) | [Jemer](../km/README.md) | [Coreano](../ko/README.md) | [Lituano](../lt/README.md) | [Malayo](../ms/README.md) | [Malabar](../ml/README.md) | [Maratí](../mr/README.md) | [Nepalí](../ne/README.md) | [Pidgin Nigeriano](../pcm/README.md) | [Noruego](../no/README.md) | [Persa (Farsi)](../fa/README.md) | [Polaco](../pl/README.md) | [Portugués (Brasil)](../pt-BR/README.md) | [Portugués (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumano](../ro/README.md) | [Ruso](../ru/README.md) | [Serbio (Cirílico)](../sr/README.md) | [Eslovaco](../sk/README.md) | [Esloveno](../sl/README.md) | [Español](./README.md) | [Swahili](../sw/README.md) | [Sueco](../sv/README.md) | [Tagalo (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Tailandés](../th/README.md) | [Turco](../tr/README.md) | [Ucraniano](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamita](../vi/README.md) > **¿Prefieres clonar localmente?** > -> Este repositorio incluye más de 50 traducciones a distintos idiomas, lo que incrementa significativamente el tamaño de descarga. Para clonar sin las traducciones, usa sparse checkout: +> Este repositorio incluye traducciones a más de 50 idiomas, lo cual incrementa significativamente el tamaño de la descarga. Para clonar sin las traducciones, usa checkout disperso: > > **Bash / macOS / Linux:** > ```bash @@ -58,142 +58,144 @@ Sigue estos pasos para empezar a usar estos recursos: ## Tabla de Contenidos - Introducción - - [Bienvenido a la familia Phi](./md/01.Introduction/01/01.PhiFamily.md) + - [Bienvenido a la Familia Phi](./md/01.Introduction/01/01.PhiFamily.md) - [Configurando tu entorno](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Comprendiendo tecnologías clave](./md/01.Introduction/01/01.Understandingtech.md) - - [Seguridad en IA para modelos Phi](./md/01.Introduction/01/01.AISafety.md) - - [Soporte de hardware Phi](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Entendiendo Tecnologías Clave](./md/01.Introduction/01/01.Understandingtech.md) + - [Seguridad en IA para los modelos Phi](./md/01.Introduction/01/01.AISafety.md) + - [Soporte de hardware para Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - [Modelos Phi y disponibilidad en distintas plataformas](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Usando Guidance-ai y Phi](./md/01.Introduction/01/01.Guidance.md) - - [Modelos en GitHub Marketplace](https://github.com/marketplace/models) - - [Catálogo de modelos Azure AI](https://ai.azure.com) + - [Uso de Guidance-ai y Phi](./md/01.Introduction/01/01.Guidance.md) + - [Modelos del Marketplace de GitHub](https://github.com/marketplace/models) + - [Catálogo de modelos de IA en Azure](https://ai.azure.com) - Inferencia Phi en diferentes entornos - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [Modelos de GitHub](./md/01.Introduction/02/02.GitHubModel.md) - - [Catálogo de modelos Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Modelos en GitHub](./md/01.Introduction/02/02.GitHubModel.md) + - [Catálogo de modelos de Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [Kit de Herramientas IA para VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Familia de inferencia Phi +- Inferencia en la Familia Phi - [Inferencia Phi en iOS](./md/01.Introduction/03/iOS_Inference.md) - [Inferencia Phi en Android](./md/01.Introduction/03/Android_Inference.md) - [Inferencia Phi en Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Inferencia Phi en AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Inferencia Phi con Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Inferencia Phi en PC de IA](./md/01.Introduction/03/AIPC_Inference.md) + - [Inferencia Phi con el Framework MLX de Apple](./md/01.Introduction/03/MLX_Inference.md) - [Inferencia Phi en servidor local](./md/01.Introduction/03/Local_Server_Inference.md) - [Inferencia Phi en servidor remoto usando AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - [Inferencia Phi con Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inferencia Phi—Visión en local](./md/01.Introduction/03/Vision_Inference.md) - - [Inferencia Phi con Kaito AKS, contenedores Azure (soporte oficial)](./md/01.Introduction/03/Kaito_Inference.md) -- [Cuantificación de la familia Phi](./md/01.Introduction/04/QuantifyingPhi.md) + - [Inferencia Phi--Visión local](./md/01.Introduction/03/Vision_Inference.md) + - [Inferencia Phi con Kaito AKS, contenedores de Azure (soporte oficial)](./md/01.Introduction/03/Kaito_Inference.md) +- [Cuantificación de la Familia Phi](./md/01.Introduction/04/QuantifyingPhi.md) - [Cuantificación de Phi-3.5 / 4 usando llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Cuantificación de Phi-3.5 / 4 usando extensiones de IA generativa para onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Cuantificación de Phi-3.5 / 4 usando extensiones IA generativa para onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [Cuantificación de Phi-3.5 / 4 usando Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Cuantificación de Phi-3.5 / 4 usando Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Cuantificación de Phi-3.5 / 4 usando el Framework MLX de Apple](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- Evaluación Phi +- Evaluación de Phi - [IA Responsable](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry para Evaluación](./md/01.Introduction/05/AIFoundry.md) - [Uso de Promptflow para Evaluación](./md/01.Introduction/05/Promptflow.md) - RAG con Azure AI Search - - [Cómo usar Phi-4-mini y Phi-4-multimodal(RAG) con Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Cómo usar Phi-4-mini y Phi-4-multimodal (RAG) con Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG híbrido local sin nube con SQLite FTS5 y phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Ejemplos de desarrollo de aplicaciones Phi - - Aplicaciones de texto y chat - - Phi-4 Ejemplos - - [📓] [Chat con modelo ONNX Phi-4-mini](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Chat con modelo ONNX Phi-4 local .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Chat app .NET consola con Phi-4 ONNX usando Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Aplicaciones de Texto y Chat + - Ejemplos Phi-4 + - [📓] [Chat con el modelo Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Chat con modelo ONNX local Phi-4 en .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [App Console Chat .NET con Phi-4 ONNX usando Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 Ejemplos - [Chatbot local en el navegador usando Phi3, ONNX Runtime Web y WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [Chat OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Modelo múltiple - Phi-3-mini interactivo y OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Construyendo un envoltorio y usando Phi-3 con MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Optimización de modelos - Cómo optimizar el modelo Phi-3-min para ONNX Runtime Web con Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [Aplicación WinUI3 con Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - - [Aplicación de notas con múltiples modelos impulsada por IA con WinUI3 - Ejemplo](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Ajuste fino e integración de modelos Phi-3 personalizados con Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Ajuste fino e integración de modelos Phi-3 personalizados con Prompt flow en Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Evaluar el modelo Phi-3 / Phi-3.5 afinado en Microsoft Foundry centrado en los principios de IA responsable de Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Ejemplo de predicción de idioma Phi-3.5-mini-instruct (Chino/Inglés)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Chatbot RAG Phi-3.5-Instruct WebGPU](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Chat de OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Multi Modelo - Phi-3-mini interactivo y OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Creando un wrapper y usando Phi-3 con MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Optimización de modelo - Cómo optimizar el modelo Phi-3-min para ONNX Runtime Web con Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [App WinUI3 con Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[Ejemplo de aplicación WinUI3 de notas impulsada por IA Multimodelo](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Ajuste fino e integración de modelos personalizados Phi-3 con Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Ajuste fino e integración de modelos personalizados Phi-3 con Prompt flow en Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Evaluar el modelo ajustado Phi-3 / Phi-3.5 en Microsoft Foundry enfocado en los principios de IA responsable de Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Ejemplo de predicción de lenguaje Phi-3.5-mini-instruct (Chino/Inglés)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Chatbot RAG WebGPU Phi-3.5-Instruct](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Uso de GPU de Windows para crear una solución Prompt flow con Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Uso de Microsoft Phi-3.5 tflite para crear una aplicación Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Uso de Microsoft Phi-3.5 tflite para crear una app Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - [Ejemplo Q&A .NET usando modelo local ONNX Phi-3 con Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Aplicación de chat de consola .NET con Semantic Kernel y Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [App chat consola .NET con Semantic Kernel y Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Ejemplos basados en código del SDK Azure AI Inference + - Ejemplos basados en código del SDK Azure AI Inference - Ejemplos Phi-4 - - [📓] [Generar código de proyecto usando Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [📓] [Generar código del proyecto usando Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Ejemplos Phi-3 / 3.5 - - [Construye tu propio chat GitHub Copilot para Visual Studio Code con la familia Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Crea tu propio agente de chat Copilot para Visual Studio Code con Phi-3.5 usando modelos de GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Construye tu propio Visual Studio Code GitHub Copilot Chat con Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Crea tu propio agente Visual Studio Code Chat Copilot con Phi-3.5 mediante modelos GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Ejemplos de razonamiento avanzado - Ejemplos Phi-4 - - [📓] [Ejemplos de Phi-4-mini-reasoning o Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Ajuste fino de Phi-4-mini-reasoning con Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Ajuste fino de Phi-4-mini-reasoning con Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning con modelos de GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning con modelos de Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demostraciones + - [📓] [Ejemplos Phi-4-mini-reasoning o Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Ajuste fino Phi-4-mini-reasoning con Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Ajuste fino Phi-4-mini-reasoning con Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-reasoning con modelos GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-reasoning con modelos Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Demos - [Demos Phi-4-mini alojados en Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Demos Phi-4-multimodal alojados en Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [Demos Phi-4-multimodal alojados en Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - Ejemplos de Visión - Ejemplos Phi-4 - - [📓] [Usa Phi-4-multimodal para leer imágenes y generar código](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [📓] [Usar Phi-4-multimodal para leer imágenes y generar código](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Ejemplos Phi-3 / 3.5 - - [📓][Phi-3-visión Imagen texto a texto](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-visión ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-visión CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-Texto de imagen a texto](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [DEMO: Reciclaje Phi-3](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-visión - Asistente de lenguaje visual - con Phi3-Vision y OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Visión Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Visión OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Ejemplo Phi-3.5 Visión multi-frame o multi-imagen](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Modelo local Phi-3 Visión ONNX usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Modelo local basado en menú Phi-3 Visión ONNX usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Ejemplos Razonamiento-Visión + - [Phi-3-vision - Asistente de lenguaje visual - con Phi3-Vision y OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Ejemplo Phi-3.5 Vision multi-frame o multi-imagen](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Modelo local ONNX Phi-3 Vision usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Modelo local ONNX Phi-3 Vision basado en menú usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Ejemplos de Razonamiento-Visión - Phi-4-Razonamiento-Visión-15B - - [📓] [Usando Phi-4-Razonamiento-Visión-15B para detectar cruce indebido](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Usando Phi-4-Razonamiento-Visión-15B para matemáticas](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Usando Phi-4-Razonamiento-Visión-15B para detectar interfaces de usuario](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [Uso de Phi-4-Razonamiento-Visión-15B para detectar cruce temerario](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Uso de Phi-4-Razonamiento-Visión-15B para matemáticas](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Uso de Phi-4-Razonamiento-Visión-15B para detectar UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - Ejemplos de matemáticas - - Ejemplos Phi-4-Mini-Flash-Reasoning-Instruct [Demo de matemáticas con Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Ejemplos de Matemáticas + - Ejemplos Phi-4-Mini-Flash-Razonamiento-Instruct [Demo de Matemáticas con Phi-4-Mini-Flash-Razonamiento-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - Ejemplos de audio + - Ejemplos de Audio - Ejemplos Phi-4 - [📓] [Extracción de transcripciones de audio usando Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Ejemplo de audio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Ejemplo de Audio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Ejemplo de traducción de voz Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [Aplicación de consola .NET usando audio Phi-4-multimodal para analizar archivo de audio y generar transcripción](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [Aplicación consola .NET usando Phi-4-multimodal Audio para analizar un archivo y generar transcripción](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - Ejemplos MOE - Ejemplos Phi-3 / 3.5 - - [📓] [Modelo de mezcla de expertos (MoEs) Phi-3.5 - Ejemplo en redes sociales](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Construyendo un pipeline de generación aumentada por recuperación (RAG) con NVIDIA NIM Phi-3 MOE, Azure AI Search y LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Modelos Phi-3.5 Mixture of Experts (MoEs) Ejemplo en redes sociales](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Construcción de un pipeline Retrieval-Augmented Generation (RAG) con NVIDIA NIM Phi-3 MOE, Azure AI Search y LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Ejemplos de llamadas a funciones + - Ejemplos de llamada a función - Ejemplos Phi-4 🆕 - - [📓] [Uso de llamadas a funciones con Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Uso de llamadas a funciones para crear multi-agentes con Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Uso de llamadas a funciones con Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Uso de llamadas a funciones con ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - [📓] [Uso de llamada a función con Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Uso de llamada a función para crear multi-agentes con Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Uso de llamada a función con Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Uso de llamada a función con ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - Ejemplos de mezcla multimodal - Ejemplos Phi-4 🆕 - [📓] [Uso de Phi-4-multimodal como periodista tecnológico](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [Aplicación de consola .NET usando Phi-4-multimodal para analizar imágenes](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [Aplicación consola .NET usando Phi-4-multimodal para analizar imágenes](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - Ajuste fino de ejemplos Phi - [Escenarios de ajuste fino](./md/03.FineTuning/FineTuning_Scenarios.md) - [Ajuste fino vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Permitir que Phi-3 se convierta en un experto industrial](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Ajuste fino: Deja que Phi-3 se convierta en un experto de la industria](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - [Ajuste fino de Phi-3 con AI Toolkit para VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [Ajuste fino de Phi-3 con Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - [Ajuste fino de Phi-3 con Lora](./md/03.FineTuning/FineTuning_Lora.md) @@ -203,122 +205,125 @@ Sigue estos pasos para empezar a usar estos recursos: - [Ajuste fino con Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Laboratorio práctico de ajuste fino con Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - [Ajuste fino de Phi-3-vision con Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Ajuste fino de Phi-3 con Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - [Ajuste fino de Phi-3-vision (soporte oficial)](./md/03.FineTuning/FineTuning_Vision.md) - - [Ajuste fino de Phi-3 con Kaito AKS, Azure Containers (soporte oficial)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Ajuste fino de Phi-3 con Kaito AKS, contenedores Azure (soporte oficial)](./md/03.FineTuning/FineTuning_Kaito.md) - [Ajuste fino de Phi-3 y 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Laboratorio práctico - [Explorando modelos de vanguardia: LLMs, SLMs, desarrollo local y más](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Desbloqueando el potencial del PLN: ajuste fino con Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Desbloqueando el potencial de NLP: ajuste fino con Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Artículos de investigación académica y publicaciones - - [Textbooks Are All You Need II: informe técnico phi-1.5](https://arxiv.org/abs/2309.05463) - - [Informe técnico Phi-3: un modelo de lenguaje altamente capaz localmente en tu teléfono](https://arxiv.org/abs/2404.14219) - - [Informe técnico Phi-4](https://arxiv.org/abs/2412.08905) - - [Informe técnico Phi-4-Mini: modelos de lenguaje multimodales compactos pero potentes mediante mezcla de LoRAs](https://arxiv.org/abs/2503.01743) - - [Optimizando modelos pequeños de lenguaje para llamadas de función en vehículos](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Ajuste fino de PHI-3 para preguntas de opción múltiple: metodología, resultados y desafíos](https://arxiv.org/abs/2501.01588) - - [Informe técnico Phi-4 razonamiento](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Informe técnico Phi-4-mini razonamiento](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) +- Artículos y publicaciones académicas + - [Los libros de texto son todo lo que necesitas II: informe técnico de phi-1.5](https://arxiv.org/abs/2309.05463) + - [Informe técnico de Phi-3: un modelo de lenguaje altamente capaz localmente en tu teléfono](https://arxiv.org/abs/2404.14219) + - [Informe técnico de Phi-4](https://arxiv.org/abs/2412.08905) + - [Informe técnico de Phi-4-Mini: modelos multimodales compactos pero potentes mediante mezcla de LoRAs](https://arxiv.org/abs/2503.01743) + - [Optimización de modelos de lenguaje pequeños para llamadas de función en vehículo](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Ajuste fino de PHI-3 para respuesta a preguntas de opción múltiple: metodología, resultados y desafíos](https://arxiv.org/abs/2501.01588) + - [Informe técnico de razonamiento Phi-4](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Informe técnico de razonamiento Phi-4-mini](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Usando Modelos Phi +## Uso de modelos Phi ### Phi en Microsoft Foundry -Puedes aprender cómo usar Microsoft Phi y cómo construir soluciones de principio a fin en tus diferentes dispositivos de hardware. Para experimentar Phi por ti mismo, comienza jugando con los modelos y personalizando Phi para tus escenarios utilizando el [Catálogo de modelos Azure AI de Microsoft Foundry](https://aka.ms/phi3-azure-ai). Puedes aprender más en Comenzando con [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Puedes aprender cómo usar Microsoft Phi y cómo construir soluciones E2E en tus diferentes dispositivos de hardware. Para experimentar Phi por ti mismo, comienza jugando con los modelos y personalizando Phi para tus escenarios usando el [Catálogo de modelos de Azure AI de Microsoft Foundry](https://aka.ms/phi3-azure-ai) y puedes aprender más en Comenzando con [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Playground** -Cada modelo tiene un playground dedicado para probar el modelo [Azure AI Playground](https://aka.ms/try-phi3). +**Área de pruebas (Playground)** +Cada modelo tiene un área de pruebas dedicada para probar el modelo [Azure AI Playground](https://aka.ms/try-phi3). ### Phi en modelos de GitHub -Puedes aprender cómo usar Microsoft Phi y cómo construir soluciones de principio a fin en tus diferentes dispositivos de hardware. Para experimentar Phi por ti mismo, comienza jugando con el modelo y personalizando Phi para tus escenarios usando el [Catálogo de modelos de GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Puedes aprender más en Comenzando con [Catálogo de modelos GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) +Puedes aprender cómo usar Microsoft Phi y cómo construir soluciones E2E en tus diferentes dispositivos de hardware. Para experimentar Phi por ti mismo, comienza jugando con el modelo y personalizando Phi para tus escenarios usando el [Catálogo de modelos de GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) y puedes aprender más en Comenzando con [Catálogo de modelos de GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Playground** -Cada modelo tiene un [playground dedicado para probar el modelo](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Área de pruebas (Playground)** +Cada modelo tiene un [área de pruebas dedicada para probar el modelo](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi en Hugging Face También puedes encontrar el modelo en [Hugging Face](https://huggingface.co/microsoft) -**Playground** -[Playground de Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Área de pruebas (Playground)** + [Área de pruebas de Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Otros Cursos + ## 🎒 Otros cursos -¡Nuestro equipo produce otros cursos! Explora: +¡Nuestro equipo produce otros cursos! Mira: -### LangChain -[![LangChain4j para principiantes](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js para principiantes](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain para principiantes](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +### LangChain +[![LangChain4j para principiantes](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js para principiantes](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain para principiantes](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agentes -[![AZD para principiantes](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI para principiantes](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP para principiantes](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents para principiantes](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agentes +[![AZD para principiantes](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI para principiantes](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP para principiantes](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Agentes de IA para principiantes](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Serie de IA generativa -[![IA generativa para principiantes](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![IA generativa (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![IA generativa (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![IA generativa (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### Serie de IA Generativa +[![IA generativa para principiantes](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![IA generativa (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![IA generativa (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![IA generativa (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Aprendizaje básico -[![ML para principiantes](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Ciencia de datos para principiantes](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![IA para principiantes](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Ciberseguridad para principiantes](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Desarrollo web para principiantes](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT para principiantes](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![Desarrollo XR para principiantes](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) + +### Aprendizaje principal +[![ML para principiantes](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Ciencia de datos para principiantes](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![IA para principiantes](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Ciberseguridad para principiantes](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Desarrollo web para principiantes](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT para principiantes](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![Desarrollo XR para principiantes](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Serie Copilot -[![Copilot para programación asistida por IA](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot para C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Aventura Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + +### Serie Copilot +[![Copilot para programación asistida por IA](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot para C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Aventura Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## IA Responsable +## IA Responsable + +Microsoft está comprometido a ayudar a nuestros clientes a usar nuestros productos de IA de manera responsable, compartiendo nuestros aprendizajes y construyendo asociaciones basadas en la confianza mediante herramientas como Notas de Transparencia y Evaluaciones de Impacto. Muchos de estos recursos se pueden encontrar en [https://aka.ms/RAI](https://aka.ms/RAI). +El enfoque de Microsoft para la IA responsable se basa en nuestros principios de IA de equidad, confiabilidad y seguridad, privacidad y seguridad, inclusión, transparencia y responsabilidad. + +Modelos a gran escala de lenguaje natural, imagen y voz - como los usados en este ejemplo - pueden potencialmente comportarse de maneras que sean injustas, poco confiables u ofensivas, causando daños. Por favor, consulte la [nota de transparencia del servicio Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) para informarse sobre riesgos y limitaciones. -Microsoft está comprometido en ayudar a nuestros clientes a usar nuestros productos de IA de manera responsable, compartiendo nuestros aprendizajes y construyendo asociaciones basadas en la confianza a través de herramientas como Notas de Transparencia y Evaluaciones de Impacto. Muchos de estos recursos se pueden encontrar en [https://aka.ms/RAI](https://aka.ms/RAI). -El enfoque de Microsoft para la IA responsable se basa en nuestros principios de IA de equidad, confiabilidad y seguridad, privacidad y seguridad, inclusividad, transparencia y responsabilidad. -Los modelos de lenguaje, imagen y voz a gran escala, como los usados en este ejemplo, pueden comportarse de maneras que no son justas, confiables o que resultan ofensivas, causando daños. Por favor, consulta la [Nota de Transparencia del servicio Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) para estar informado sobre riesgos y limitaciones. -El enfoque recomendado para mitigar estos riesgos es incluir un sistema de seguridad en su arquitectura que pueda detectar y prevenir comportamientos dañinos. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) proporciona una capa independiente de protección, capaz de detectar contenido dañino generado por usuarios y por IA en aplicaciones y servicios. Azure AI Content Safety incluye API de texto e imagen que permiten detectar material que es perjudicial. Dentro de Microsoft Foundry, el servicio Content Safety le permite ver, explorar y probar código de ejemplo para detectar contenido dañino en diferentes modalidades. La siguiente [documentación de inicio rápido](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) le guía para realizar solicitudes al servicio. +El enfoque recomendado para mitigar estos riesgos es incluir un sistema de seguridad en su arquitectura que pueda detectar y prevenir conductas dañinas. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) proporciona una capa independiente de protección, capaz de detectar contenido dañino generado por usuarios y por IA en aplicaciones y servicios. Azure AI Content Safety incluye APIs de texto e imagen que le permiten detectar material perjudicial. Dentro de Microsoft Foundry, el servicio Content Safety le permite ver, explorar y probar código de ejemplo para detectar contenido dañino en diferentes modalidades. La siguiente [documentación de inicio rápido](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) le guía para realizar solicitudes al servicio. -Otro aspecto a tener en cuenta es el rendimiento general de la aplicación. Con aplicaciones multimodales y multimodelos, consideramos que el rendimiento significa que el sistema funciona como usted y sus usuarios esperan, incluido no generar resultados dañinos. Es importante evaluar el rendimiento de su aplicación en general utilizando [evaluadores de Rendimiento y Calidad y de Riesgo y Seguridad](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). También tiene la capacidad de crear y evaluar con [evaluadores personalizados](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Otro aspecto a tener en cuenta es el desempeño general de la aplicación. En aplicaciones multimodales y multi-modelos, consideramos que el desempeño significa que el sistema funciona como usted y sus usuarios esperan, incluyendo no generar resultados dañinos. Es importante evaluar el desempeño de su aplicación general utilizando los [evaluadores de Desempeño y Calidad y de Riesgo y Seguridad](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). También tiene la posibilidad de crear y evaluar con [evaluadores personalizados](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Puede evaluar su aplicación de IA en su entorno de desarrollo utilizando el [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Dado un conjunto de datos de prueba o un objetivo, las generaciones de su aplicación de IA generativa se miden cuantitativamente con evaluadores incorporados o evaluadores personalizados de su elección. Para comenzar con el azure ai evaluation sdk y evaluar su sistema, puede seguir la [guía de inicio rápido](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Una vez que ejecute una evaluación, puede [visualizar los resultados en Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Puede evaluar su aplicación de IA en su entorno de desarrollo utilizando el [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Dado un conjunto de datos de prueba o un objetivo, las generaciones de su aplicación de IA generativa se miden cuantitativamente con evaluadores integrados o evaluadores personalizados de su elección. Para comenzar con el Azure AI Evaluation SDK para evaluar su sistema, puede seguir la [guía de inicio rápido](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Una vez que ejecute una evaluación, puede [visualizar los resultados en Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## Marcas registradas +## Marcas Registradas -Este proyecto puede contener marcas registradas o logotipos de proyectos, productos o servicios. El uso autorizado de las marcas o logotipos de Microsoft está sujeto a y debe seguir las [Directrices de marcas registradas y marca de Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -El uso de marcas o logotipos de Microsoft en versiones modificadas de este proyecto no debe causar confusión ni implicar patrocinio de Microsoft. Cualquier uso de marcas o logotipos de terceros está sujeto a las políticas de esos terceros. +Este proyecto puede contener marcas registradas o logotipos de proyectos, productos o servicios. El uso autorizado de marcas o logotipos de Microsoft está sujeto a y debe seguir las [Directrices de Marcas y Marca Comercial de Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +El uso de marcas o logotipos de Microsoft en versiones modificadas de este proyecto no debe causar confusión ni implicar patrocinio por parte de Microsoft. Cualquier uso de marcas o logotipos de terceros está sujeto a las políticas de esos terceros. -## Obtener ayuda +## Obtener Ayuda -Si se queda atascado o tiene alguna pregunta sobre la creación de aplicaciones de IA, únase a: +Si tiene problemas o alguna pregunta sobre la construcción de aplicaciones de IA, únase a: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Si tiene comentarios sobre el producto o errores durante la creación, visite: +Si tiene comentarios del producto o errores durante la construcción, visite: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Aviso legal**: -Este documento ha sido traducido utilizando el servicio de traducción automática [Co-op Translator](https://github.com/Azure/co-op-translator). Aunque nos esforzamos por la precisión, tenga en cuenta que las traducciones automáticas pueden contener errores o inexactitudes. El documento original en su idioma nativo debe considerarse la fuente autorizada. Para información crítica, se recomienda una traducción profesional realizada por humanos. No nos hacemos responsables de ningún malentendido o interpretación errónea que surja del uso de esta traducción. +**Descargo de responsabilidad**: +Este documento ha sido traducido utilizando el servicio de traducción automática [Co-op Translator](https://github.com/Azure/co-op-translator). Aunque nos esforzamos por la precisión, tenga en cuenta que las traducciones automatizadas pueden contener errores o inexactitudes. El documento original en su idioma nativo debe considerarse la fuente autorizada. Para información crítica, se recomienda una traducción profesional humana. No somos responsables de cualquier malentendido o interpretación errónea que surja del uso de esta traducción. \ No newline at end of file diff --git a/translations/es/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/es/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..5d8633fef --- /dev/null +++ b/translations/es/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 RAG Híbrido con Microsoft phi-4-mini y SQLite FTS5 (SLM sin Nube)\n", + "\n", + "> **Autor:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Enfoque:** Pequeños Modelos de Lenguaje (SLMs), SQLite FTS5 BM25, Embeddings Densos, Fusión de Ranking Recíproco (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivación: El Dilema de Recuperación de Palabras Clave en SLMs Locales\n", + "Las arquitecturas estándar de RAG que dependen únicamente de embeddings vectoriales densos a menudo fallan en recuperar tokens numéricos exactos (ej., `2,340,000 TL`, códigos de contrato, números de cuenta). \n", + "Por otro lado, la búsqueda léxica dispersa (BM25) pasa por alto sinónimos semánticos y preguntas parafraseadas.\n", + "\n", + "Este recetario demuestra cómo implementar un **motor de recuperación híbrido en memoria de alta velocidad** combinando:\n", + "1. **Vectores Densos** (Similitud Coseno)\n", + "2. **Búsqueda Léxica Dispersa** (SQLite FTS5 BM25)\n", + "3. **Fusión de Ranking Recíproco (RRF, $k=60$)**\n", + "4. **Generación de Citas Fundamentadas (`[1]`, `[2]`)** con Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Esquema Dual SQLite (Vectores Densos + Tabla Virtual FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Muestra de Ingesta y Evaluación de Ejecución\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Formulación de instrucciones fundamentadas para Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Descargo de responsabilidad**:\nEste documento ha sido traducido utilizando el servicio de traducción automática [Co-op Translator](https://github.com/Azure/co-op-translator). Aunque nos esforzamos por la precisión, tenga en cuenta que las traducciones automatizadas pueden contener errores o inexactitudes. El documento original en su idioma nativo debe considerarse la fuente autorizada. Para información crítica, se recomienda una traducción profesional humana. No somos responsables de cualquier malentendido o interpretación errónea que surja del uso de esta traducción.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/et/.co-op-translator.json b/translations/et/.co-op-translator.json index fb86a5a79..737367a9f 100644 --- a/translations/et/.co-op-translator.json +++ b/translations/et/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "et" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T07:54:29+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:27:51+00:00", "source_file": "README.md", "language_code": "et" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "et" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:31:28+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "et" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-10-11T11:33:33+00:00", diff --git a/translations/et/README.md b/translations/et/README.md index 1e37d530b..b81cd87e4 100644 --- a/translations/et/README.md +++ b/translations/et/README.md @@ -1,14 +1,14 @@ -# Phi Kokaraamat: Käed-küljes näited Microsofti Phi mudelitega +# Phi kokaraamat: Praktilised näited Microsofti Phi mudelitega -[![Ava ja kasuta näidiseid GitHub Codespaces’is](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Ava Dev Containerites](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Ava ja kasuta näiteid GitHub Codespaces'is](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Ava Dev Containers'is](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub kaastöötajad](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ülesanded](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub probleemid](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub tõmbepäringud](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PR-d on teretulnud](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![PR-d on oodatud](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub vaatlejad](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub jälgijad](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub forkid](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub tähed](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) @@ -16,27 +16,27 @@ Phi on Microsofti arendatud avatud lähtekoodiga tehisintellekti mudelite seeria. -Phi on praegu kõige võimsam ja kuluefektiivsem väike keelemudel (SLM), mis annab väga head tulemused mitmekeelsuses, loogilises mõtlemises, teksti/vestluse genereerimises, programmeerimises, piltides, helis ja muudes olukordades. +Phi on praegu kõige võimsam ja kulutõhusam väike keelemudel (SLM), millel on väga head tulemused mitmekeelses, arutlevas, teksti/vestluse genereerimises, kodeerimises, piltides, helis ja teistes stsenaariumites. -Saad Phi pilve või servaseadmetesse juurutada ning piiratud arvutusvõimsusega hõlpsalt generatiivseid tehisintellekti rakendusi luua. +Saate deploy'da Phi kas pilve või ääreseadmetele ning luua hõlpsasti generatiivseid tehisintellekti rakendusi piiratud arvutusvõimsusega. -Järgmiste sammude abil saad neid ressursse kasutada: -1. **Loo hoidla haru (fork)**: Klõpsa [![GitHub forkid](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Klooni hoidla**: `git clone https://github.com/microsoft/PhiCookBook.git` +Alustamiseks toimige järgmiselt: +1. **Forki hoidla**: Klõpsake [![GitHub forkid](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Kloneeri hoidla**: `git clone https://github.com/microsoft/PhiCookBook.git` 3. [**Liitu Microsoft AI Discord kogukonnaga ja kohtuge ekspertide ning teiste arendajatega**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/et/cover.eb18d1b9605d754b.webp) ### 🌐 Mitmekeelne tugi -#### Toetatud GitHub Actioni kaudu (Automaatne & alati ajakohane) +#### Toetatud läbi GitHub Actioni (automatiseeritud ja alati ajakohane) -[Araabia](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgaaria](../bg/README.md) | [Burma (Myanmar)](../my/README.md) | [Hiina (lihtsustatud)](../zh-CN/README.md) | [Hiina (traditsiooniline, Hongkong)](../zh-HK/README.md) | [Hiina (traditsiooniline, Macau)](../zh-MO/README.md) | [Hiina (traditsiooniline, Taiwan)](../zh-TW/README.md) | [Horvaadi](../hr/README.md) | [Tšehhi](../cs/README.md) | [Taani](../da/README.md) | [Hollandi](../nl/README.md) | [Eesti](./README.md) | [Soome](../fi/README.md) | [Prantsuse](../fr/README.md) | [Saksa](../de/README.md) | [Kreeka](../el/README.md) | [Heebrea](../he/README.md) | [Hindi](../hi/README.md) | [Ungari](../hu/README.md) | [Indoneesia](../id/README.md) | [Itaalia](../it/README.md) | [Jaapani](../ja/README.md) | [Kannada](../kn/README.md) | [Khmeeri](../km/README.md) | [Korea](../ko/README.md) | [Leedu](../lt/README.md) | [Malai](../ms/README.md) | [Malajalami](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigeeria pidžin](../pcm/README.md) | [Norra](../no/README.md) | [Pärsia (Farsi)](../fa/README.md) | [Poola](../pl/README.md) | [Portugali (Brasiilia)](../pt-BR/README.md) | [Portugali (Portugal)](../pt-PT/README.md) | [Pandžabi (Gurmukhi)](../pa/README.md) | [Rumeenia](../ro/README.md) | [Vene](../ru/README.md) | [Serbia (kirillitsa)](../sr/README.md) | [Slovaki](../sk/README.md) | [Sloveeni](../sl/README.md) | [Hispaania](../es/README.md) | [Suahiili](../sw/README.md) | [Rootsi](../sv/README.md) | [Tagaloogi (Filipino)](../tl/README.md) | [Tamili](../ta/README.md) | [Telugu](../te/README.md) | [Tai](../th/README.md) | [Türgi](../tr/README.md) | [Ukraina](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnami](../vi/README.md) +[Araabia](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgaaria](../bg/README.md) | [Birma (Myanmar)](../my/README.md) | [Hiina (lihtsustatud)](../zh-CN/README.md) | [Hiina (traditsiooniline, Hongkong)](../zh-HK/README.md) | [Hiina (traditsiooniline, Macau)](../zh-MO/README.md) | [Hiina (traditsiooniline, Taiwan)](../zh-TW/README.md) | [Horvaadi](../hr/README.md) | [Tšehhi](../cs/README.md) | [Taani](../da/README.md) | [Hollandi](../nl/README.md) | [Eesti](./README.md) | [Soome](../fi/README.md) | [Prantsuse](../fr/README.md) | [Saksa](../de/README.md) | [Kreeka](../el/README.md) | [Heebrea](../he/README.md) | [Hindi](../hi/README.md) | [Ungari](../hu/README.md) | [Indoneesia](../id/README.md) | [Itaalia](../it/README.md) | [Jaapani](../ja/README.md) | [Kannada](../kn/README.md) | [Khmeri](../km/README.md) | [Korea](../ko/README.md) | [Leedu](../lt/README.md) | [Malai](../ms/README.md) | [Malajalami](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigeeria pidžin](../pcm/README.md) | [Norra](../no/README.md) | [Pärsia (Farsi)](../fa/README.md) | [Poola](../pl/README.md) | [Portugali (Brasiilia)](../pt-BR/README.md) | [Portugali (Portugal)](../pt-PT/README.md) | [Pundžabi (Gurmukhi)](../pa/README.md) | [Rumeenia](../ro/README.md) | [Vene](../ru/README.md) | [Serbia (kirillitsa)](../sr/README.md) | [Slovaki](../sk/README.md) | [Sloveeni](../sl/README.md) | [Hispaania](../es/README.md) | [Suaheli](../sw/README.md) | [Rootsi](../sv/README.md) | [Tagalogi (Filipino)](../tl/README.md) | [Tamili](../ta/README.md) | [Telugu](../te/README.md) | [Tai](../th/README.md) | [Türgi](../tr/README.md) | [Ukraina](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnam](../vi/README.md) -> **Eelistad kohaliku kloonimisega?** +> **Eelistate kohalikku kloonimist?** > -> See hoidla sisaldab üle 50 keele tõlkeid, mis suurendavad oluliselt allalaadimise mahtu. Tõlkedeta kloonimiseks kasuta osalist väljavalimist: +> See hoidla sisaldab enam kui 50 keele tõlget, mis suurendab oluliselt allalaadimismahtu. Tõlgeteta kloonimiseks kasutage sparse checkout-i: > > **Bash / macOS / Linux:** > ```bash @@ -52,7 +52,7 @@ Järgmiste sammude abil saad neid ressursse kasutada: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Nii saad kõik vajaliku kursuse läbimiseks palju kiiremalt kätte. +> See annab teile kõik vajaliku kursuse lõpetamiseks palju kiirema allalaadimisega. ## Sisukord @@ -61,194 +61,196 @@ Järgmiste sammude abil saad neid ressursse kasutada: - [Tere tulemast Phi perekonda](./md/01.Introduction/01/01.PhiFamily.md) - [Keskkonna seadistamine](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Oluliste tehnoloogiate mõistmine](./md/01.Introduction/01/01.Understandingtech.md) - - [Tehisintellekti ohutus Phi mudelite jaoks](./md/01.Introduction/01/01.AISafety.md) + - [AI ohutus Phi mudelite jaoks](./md/01.Introduction/01/01.AISafety.md) - [Phi riistvaratugi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi mudelid ja kättesaadavus platvormidel](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Phi mudelid ja kättesaadavus platvormide lõikes](./md/01.Introduction/01/01.Edgeandcloud.md) - [Guidance-ai ja Phi kasutamine](./md/01.Introduction/01/01.Guidance.md) - - [GitHub Marketplace mudelid](https://github.com/marketplace/models) - - [Azure AI mudelikataloog](https://ai.azure.com) + - [GitHub Marketplace'i mudelid](https://github.com/marketplace/models) + - [Azure AI mudelite kataloog](https://ai.azure.com) -- Phi inference erinevates keskkondades - - [Hugging face](./md/01.Introduction/02/01.HF.md) +- Phi järeldamine erinevates keskkondades + - [Hugging Face](./md/01.Introduction/02/01.HF.md) - [GitHub mudelid](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry mudelikataloog](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Microsoft Foundry mudelite kataloog](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [AI tööriistakomplekt VSCode'is (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry lokaalselt](./md/01.Introduction/02/07.FoundryLocal.md) - -- Phi perekonna inference - - [Phi inference iOS-is](./md/01.Introduction/03/iOS_Inference.md) - - [Phi inference Androidis](./md/01.Introduction/03/Android_Inference.md) - - [Phi inference Jetsonis](./md/01.Introduction/03/Jetson_Inference.md) - - [Phi inference AI arvutis](./md/01.Introduction/03/AIPC_Inference.md) - - [Phi inference Apple MLX raamistikuga](./md/01.Introduction/03/MLX_Inference.md) - - [Phi inference lokaalses serveris](./md/01.Introduction/03/Local_Server_Inference.md) - - [Phi inference kaugserveris AI Toolkitiga](./md/01.Introduction/03/Remote_Interence.md) - - [Phi inference Rustiga](./md/01.Introduction/03/Rust_Inference.md) - - [Phi inference–Visioon lokaalselt](./md/01.Introduction/03/Vision_Inference.md) - - [Phi inference Kaito AKS, Azure konteineritega (ametlik tugi)](./md/01.Introduction/03/Kaito_Inference.md) - -- [Phi perekonna kvantimine](./md/01.Introduction/04/QuantifyingPhi.md) + - [Foundry lokaalne](./md/01.Introduction/02/07.FoundryLocal.md) + +- Phi perekonna järeldamine + - [Phi järeldamine iOS-is](./md/01.Introduction/03/iOS_Inference.md) + - [Phi järeldamine Androidis](./md/01.Introduction/03/Android_Inference.md) + - [Phi järeldamine Jetsonis](./md/01.Introduction/03/Jetson_Inference.md) + - [Phi järeldamine AI PC-s](./md/01.Introduction/03/AIPC_Inference.md) + - [Phi järeldamine Apple MLX raamistikuga](./md/01.Introduction/03/MLX_Inference.md) + - [Phi järeldamine lokaalses serveris](./md/01.Introduction/03/Local_Server_Inference.md) + - [Phi järeldamine kaugserveris AI tööriistaga](./md/01.Introduction/03/Remote_Interence.md) + - [Phi järeldamine Rustiga](./md/01.Introduction/03/Rust_Inference.md) + - [Phi visiooni järeldamine lokaalselt](./md/01.Introduction/03/Vision_Inference.md) + - [Phi järeldamine Kaito AKS, Azure konteineritega (ametlik tugi)](./md/01.Introduction/03/Kaito_Inference.md) +- [Phi perekonna kvantifitseerimine](./md/01.Introduction/04/QuantifyingPhi.md) - [Phi-3.5 / 4 kvantimine kasutades llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Phi-3.5 / 4 kvantimine Generative AI laienditega onnxruntime jaoks](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Phi-3.5 / 4 kvantimine kasutades Generative AI laiendusi onnxruntime'ile](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [Phi-3.5 / 4 kvantimine kasutades Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Phi-3.5 / 4 kvantimine Apple MLX raamistiku abil](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Phi-3.5 / 4 kvantimine kasutades Apple MLX raamistikku](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Phi hindamine - [Vastutustundlik AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry hindamiseks](./md/01.Introduction/05/AIFoundry.md) - [Promptflow kasutamine hindamiseks](./md/01.Introduction/05/Promptflow.md) -- RAG Azure AI otsinguga - - [Kuidas kasutada Phi-4-mini ja Phi-4-multimodal(RAG) Azure AI otsinguga](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- RAG kasutamine Azure AI otsingus + - [Kuidas kasutada Phi-4-mini ja Phi-4-multimodal (RAG) Azure AI otsinguga](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Null-pilve kohalik hübriid RAG SQLite FTS5 ja phi-4-mini-ga](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Phi rakenduste arendamise näited - Teksti- ja vestlusrakendused - Phi-4 näited - [📓] [Vestlus Phi-4-mini ONNX mudeliga](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Vestlus Phi-4 lokaalse ONNX mudeliga .NET-is](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Vestlus .NET konsoolirakendusega Phi-4 ONNX ja Semantic Kerneliga](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 näited - - [Lokaalne vestlusbott brauseris kasutades Phi3, ONNX Runtime Web ja WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Mitme Mudeli - Interaktiivne Phi-3-mini ja OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Komplekti ehitamine ja Phi-3 kasutamine koos MLFlow'ga](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Mudeli optimeerimine - Kuidas optimeerida Phi-3-mini mudelit ONNX Runtime Web’ile Olive'iga](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Vestlus Phi-4 lokaalse ONNX mudeliga .NET'is](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Vestluse .NET konsoolirakendus Phi-4 ONNX mudeliga kasutades Semantic Kernelit](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Näidised + - [Kohalik vestlusrobot brauseris, kasutades Phi3, ONNX Runtime Web ja WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [OpenVino Vestlus](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Mitmemudeliline - interaktiivne Phi-3-mini ja OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Katteloori loomine ja Phi-3 kasutamine MLFlow’ga](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Mudelid optimeerimine - Kuidas optimeerida Phi-3-minimudelit ONNX Runtime Web jaoks Olive’iga](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [WinUI3 rakendus Phi-3 mini-4k-instruct-onnx’iga](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Mitme Mudeli tehisintellektil põhinev märkmete rakenduse näidis](https://github.com/microsoft/ai-powered-notes-winui3-sample) + -[WinUI3 mitmemudeliline AI-võimustatud märkmete rakenduse näidis](https://github.com/microsoft/ai-powered-notes-winui3-sample) - [Kohandatud Phi-3 mudelite peenhäälestamine ja integreerimine Prompt flow abil](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - [Kohandatud Phi-3 mudelite peenhäälestamine ja integreerimine Prompt flow abil Microsoft Foundrys](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Peenhäälestatud Phi-3/Phi-3.5 mudeli hindamine Microsoft Foundrys, keskendudes Microsofti vastutustundliku tehisintellekti põhimõtetele](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct keele prognoosimise näidis (hiina/inglise)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Peenhäälestatud Phi-3 / Phi-3.5 mudeli hindamine Microsoft Foundry’s, keskendudes Microsofti vastutustundliku tehisintellekti põhimõtetele](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct keele ennustamise näidis (hiina/inglise keel)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG vestlusrobot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Windowsi GPU kasutamine Prompt flow lahenduse loomiseks Phi-3.5-Instruct ONNX’iga](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Microsoft Phi-3.5 tflite kasutamine Androidi rakenduse loomiseks](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Küsimuste ja vastuste .NET näide kohaliku ONNX Phi-3 mudeli kasutamisega Microsoft.ML.OnnxRuntime abil](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Käsurea vestlusrakendus .NET Semantic Kernel ja Phi-3 abil](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Küsimused ja vastused .NET näidis, kasutades kohalikku ONNX Phi-3 mudelit Microsoft.ML.OnnxRuntime’iga](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Konsooli vestlus .NET rakendus Semantic Kerneliga ja Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI järeldamise SDK põhised näited - - Phi-4 näited - - [📓] [Projekti koodi genereerimine Phi-4-multimodal abil](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 näited + - Azure AI järeldussõlme koodipõhised näidised + - Phi-4 näidised + - [📓] [Projektikoodi genereerimine kasutades Phi-4-multimodaalset](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Phi-3 / 3.5 näidised - [Ehita oma Visual Studio Code GitHub Copilot vestlus Microsoft Phi-3 perekonnaga](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Loo oma Visual Studio Code Chat Copilot agent Phi-3.5 GitHub mudelitega](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - Täiustatud järeldamise näited - - Phi-4 näited - - [📓] [Phi-4-mini-järeldamine või Phi-4-järeldamise näited](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Phi-4-mini-järeldamise peenhäälestamine Microsoft Olive’iga](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-järeldamise peenhäälestamine Apple MLX’iga](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-järeldamine GitHub mudelitega](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-järeldamine Microsoft Foundry mudelitega](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demo’d - - [Phi-4-mini demo’d Hugging Face Spaces’is](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demo’d Hugginge Face Spaces’is](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Nägemisnäited - - Phi-4 näited - - [📓] [Phi-4-multimodal kasutamine piltide lugemiseks ja koodi genereerimiseks](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 näited - - [📓][Phi-3-nägemis-pildi tekst tekstiks](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-nägemis-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-nägemis CLIP sisend](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 ringlussevõtt](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-nägemis - visuaalse keelega assistent - Phi3-Vision ja OpenVINO abil](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Nägemine Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Nägemine OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Nägemine mitme kaadri või mitme pildi näidis](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Nägemine kohalik ONNX mudel Microsoft.ML.OnnxRuntime .NET abil](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menüü-põhine Phi-3 Nägemine kohalik ONNX mudel Microsoft.ML.OnnxRuntime .NET abil](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Järeldamine-nägemisnäited - - Phi-4-Järeldamine-Nägemine-15B - - [📓] [Phi-4-Järeldamine-Nägemine-15B kasutamine jaywalking’u tuvastamiseks](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Phi-4-Järeldamine-Nägemine-15B kasutamine matemaatikas](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-Järeldamine-Nägemine-15B kasutamine kasutajaliidese (UI) tuvastamiseks](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - Matemaatika näited - - Phi-4-Mini-Flash-Järeldamine-Juhised näited [Matemaatika demo Phi-4-Mini-Flash-Järeldamine-Juhistega](./md/02.Application/09.Math/MathDemo.ipynb) - - - Helinäited - - Phi-4 näited - - [📓] [Audio transkriptsioonide ekstraheerimine Phi-4-multimodal abil](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal heli näidis](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal kõne tõlke näidis](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET käsurea rakendus, mis kasutab Phi-4-multimodal heli, heli faili analüüsimiseks ja transkriptsiooni genereerimiseks](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - MOE näited - - Phi-3 / 3.5 näited - - [📓] [Phi-3.5 Ekspertide segu mudelid (MoEs) sotsiaalmeedia näidis](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Retrieval-Augmented Generation (RAG) torujuhtme ehitamine NVIDIA NIM Phi-3 MOE, Azure AI Search ja LlamaIndex abil](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [Loo oma Visual Studio Code vestlus Copilot agent Phi-3.5 abil GitHub mudelitega](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - Täiustatud arutluse näidised + - Phi-4 näidised + - [📓] [Phi-4-mini-arutlus või Phi-4-arutluse näidised](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Phi-4-mini-arutluse peenhäälestamine Microsoft Oliviga](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-arutluse peenhäälestamine Apple MLX’iga](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-arutlus GitHub mudelitega](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-arutlus Microsoft Foundry mudelitega](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Demonstreerimised + - [Phi-4-mini demonstreerimised majutatud Hugging Face Spacesis](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodaalsete demonstreerimised majutatud Hugging Face Spacesis](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Nägemise näidised + - Phi-4 näidised + - [📓] [Kasutage Phi-4-multimodaalset piltide lugemiseks ja koodi genereerimiseks](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 näidised + - [📓][Phi-3-nägemise pildi tekst tekstiks](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-nägemine-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-nägemise CLIP manustus](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [DEMO: Phi-3 taaskasutus](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-nägemine - visuaalne keeleassistent - Phi3-Vision ja OpenVINO abil](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision mitme kaadri või mitme pildi näidis](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision kohalik ONNX mudel Microsoft.ML.OnnxRuntime .NET abil](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Menüüpõhine Phi-3 Vision kohalik ONNX mudel Microsoft.ML.OnnxRuntime .NET abil](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Arutlus-nägemise näidised + - Phi-4-Arutlus-Nägemine-15B + - [📓] [Phi-4-Arutlus-Nägemine-15B kasutamine jaywalkingu tuvastamiseks](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-Arutlus-Nägemine-15B kasutamine matemaatika jaoks](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Phi-4-Arutlus-Nägemine-15B kasutamine kasutajaliidese tuvastamiseks](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - Matemaatika näidised + - Phi-4-Mini-Flash-Arutlus-Juhised näidised [Matemaatika demo Phi-4-Mini-Flash-Arutlus-Juhistega](./md/02.Application/09.Math/MathDemo.ipynb) + + - Audio näidised + - Phi-4 näidised + - [📓] [Audio transkriptide väljavõtmine kasutades Phi-4-multimodaalset](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodaalne audiomudel](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodaalne kõnetõlke näidis](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET konsoolirakendus, kasutades Phi-4-multimodaalset heli analüüsimiseks ja transkripti genereerimiseks](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE näidised + - Phi-3 / 3.5 näidised + - [📓] [Phi-3.5 Ekspertide segu mudelid (MoEs) Sotsiaalmeedia näidis](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Taastamisega rikastatud genereerimise (RAG) torujuhtme loomine NVIDIA NIM Phi-3 MOE, Azure AI Search ja LlamaIndex’iga](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Funktsioonikutsumise näited - - Phi-4 näited 🆕 - - [📓] [Funktsioonikutsumise kasutamine Phi-4-mini puhul](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Funktsioonikutsumise kasutamine mitme agendi loomiseks Phi-4-mini abil](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Funktsioonikutsumise kasutamine Ollama’ga](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Funktsioonikutsumise kasutamine ONNX’iga](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Mitmemodaalne segamine näited - - Phi-4 näited 🆕 - - [📓] [Phi-4-multimodal kasutamine tehnoloogiaajakirjanikuna](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET käsurea rakendus, mis kasutab Phi-4-multimodal piltide analüüsimiseks](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Phi peenhäälestamine näited + - Funktsioonikõne näidised + - Phi-4 näidised 🆕 + - [📓] [Funktsioonikõne kasutamine Phi-4-mini’ga](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Funktsioonikõne kasutamine mitme agendi loomiseks Phi-4-mini’ga](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Funktsioonikõne kasutamine Ollamaga](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Funktsioonikõne kasutamine ONNX’iga](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Multimodaalne segamine näidised + - Phi-4 näidised 🆕 + - [📓] [Phi-4-multimodaalse kasutamine tehnoloogiauudiste ajakirjanikuna](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET konsoolirakendus, kasutades Phi-4-multimodaalset piltide analüüsimiseks](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Phi peenhäälestamine - [Peenhäälestamise stsenaariumid](./md/03.FineTuning/FineTuning_Scenarios.md) - [Peenhäälestamine vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Lase Phi-3-l saada tööstuse eksperdiks peenhäälestamisega](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Phi-3 peenhäälestamine AI Toolkitiga VS Code jaoks](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Lase Phi-3’l saada tööstuse eksperdiks peenhäälestamise abil](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Phi-3 peenhäälestamine AI Toolkit’i abil VS Code’is](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [Phi-3 peenhäälestamine Azure Machine Learning teenusega](./md/03.FineTuning/Introduce_AzureML.md) - - [Phi-3 peenhäälestamine Lora abil](./md/03.FineTuning/FineTuning_Lora.md) - - [Phi-3 peenhäälestamine QLora abil](./md/03.FineTuning/FineTuning_Qlora.md) - - [Phi-3 peenhäälestamine Microsoft Foundry’ga](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Phi-3 peenhäälestamine Loraga](./md/03.FineTuning/FineTuning_Lora.md) + - [Phi-3 peenhäälestamine QLoraga](./md/03.FineTuning/FineTuning_Qlora.md) + - [Phi-3 peenhäälestamine Microsoft Foundry’s](./md/03.FineTuning/FineTuning_AIFoundry.md) - [Phi-3 peenhäälestamine Azure ML CLI/SDK abil](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Peenhäälestamine Microsoft Olive’iga](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive praktiline labori käsiraamat](./md/03.FineTuning/olive-lab/readme.md) - - [Phi-3-vision peenhäälestamine Weights and Bias abil](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Peenhäälestamine Microsoft Oliviga](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Peenhäälestamine Microsoft Olive praktilises laboris](./md/03.FineTuning/olive-lab/readme.md) + - [Phi-3-nägemise peenhäälestamine Weights and Bias’iga](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Phi-3 peenhäälestamine Apple MLX raamistikuga](./md/03.FineTuning/FineTuning_MLX.md) - [Phi-3-vision peenhäälestamine (ametlik tugi)](./md/03.FineTuning/FineTuning_Vision.md) - [Phi-3 peenhäälestamine Kaito AKS-iga, Azure konteinerid (ametlik tugi)](./md/03.FineTuning/FineTuning_Kaito.md) - [Phi-3 ja 3.5 Vision peenhäälestamine](https://github.com/2U1/Phi3-Vision-Finetune) -- Praktikum - - [Uurides uusimaid mudeleid: LLM-id, SLM-id, kohalik arendus ja palju muud](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Looduskeele töötluse potentsiaali avamine: Microsoft Olive’i peenhäälestus](https://github.com/azure/Ignite_FineTuning_workshop) +- Praktiline töötuba + - [Uuenduslike mudelite uurimine: LLMid, SLMid, lokaalarendus ja palju muud](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP potentsiaali avamine: Microsoft Olive'ga peenhäälestamine](https://github.com/azure/Ignite_FineTuning_workshop) -- Akadeemilised uurimused ja publikatsioonid - - [Õpikud on kõik, mida vajad II: phi-1.5 tehniline aruanne](https://arxiv.org/abs/2309.05463) - - [Phi-3 tehniline aruanne: väga võimekas keelemudel sinu telefonis kohapeal](https://arxiv.org/abs/2404.14219) +- Akadeemilised uurimuspaberid ja väljaanded + - [Textbooks Are All You Need II: phi-1.5 tehniline aruanne](https://arxiv.org/abs/2309.05463) + - [Phi-3 tehniline aruanne: võimas keelemudel lokaalselt teie telefonis](https://arxiv.org/abs/2404.14219) - [Phi-4 tehniline aruanne](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini tehniline aruanne: kompaktne, ent võimas multimodaalne keelemudel LoRA seguga](https://arxiv.org/abs/2503.01743) + - [Phi-4-Mini tehniline aruanne: kompaktne kuid võimas multimodaalne keelemudel LoRAde seguga](https://arxiv.org/abs/2503.01743) - [Väikeste keelemudelite optimeerimine sõidukisiseste funktsioonikõnede jaoks](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) PHI-3 peenhäälestamine valikvastustega küsimustele vastamiseks: metoodika, tulemused ja väljakutsed](https://arxiv.org/abs/2501.01588) - - [Phi-4 loogilise mõtlemise tehniline aruanne](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini loogilise mõtlemise tehniline aruanne](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [(WhyPHI) PHI-3 peenhäälestamine mitmevalikuliste küsimuste vastamiseks: meetod, tulemused ja väljakutsed](https://arxiv.org/abs/2501.01588) + - [Phi-4-lähenemine tehniline aruanne](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-lähenemine tehniline aruanne](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Phi mudelite kasutamine -### Phi Microsoft Foundry’s +### Phi Microsoft Foundrys -Õpi, kuidas kasutada Microsoft Phi’d ja kuidas ehitada E2E lahendusi oma erinevates riistvaraseadmetes. Phi proovimiseks alusta mudelitega mängimisest ja Phi kohandamisest oma stsenaariumitele kasutades [Microsoft Foundry Azure AI Model Catalog’ut](https://aka.ms/phi3-azure-ai). Lisateavet leiad juhistest Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Saate õppida, kuidas kasutada Microsoft Phi-d ja kuidas ehitada E2E lahendusi erinevate riistvarade jaoks. Phi kogemiseks alustage mudelitega mängimisest ja Phi kohandamisest oma stsenaariumide jaoks [Microsoft Foundry Azure AI mudelikataloogi](https://aka.ms/phi3-azure-ai) abil, lisateavet leiate juba alustades [Microsoft Foundrys](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) **Mänguväljak** -Iga mudelil on pühendatud mänguväljak mudeli testimiseks [Azure AI Playground](https://aka.ms/try-phi3). +Igal mudelil on pühendatud mänguväljak mudeli testimiseks [Azure AI Playground](https://aka.ms/try-phi3). -### Phi GitHub mudelites +### Phi GitHubi mudelitel -Õpi, kuidas kasutada Microsoft Phi’d ja kuidas ehitada E2E lahendusi oma erinevates riistvaraseadmetes. Phi proovimiseks alusta mudeliga mängimisest ja Phi kohandamisest oma stsenaariumitele kasutades [GitHub Model Catalog’ut](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Lisateavet leiad juhistest Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Saate õppida, kuidas kasutada Microsoft Phi-d ja kuidas ehitada E2E lahendusi erinevate riistvarade jaoks. Phi kogemiseks alustage mudeliga mängimisest ja Phi kohandamisest oma stsenaariumide jaoks [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) abil, lisateavet leiate juba alustades [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) **Mänguväljak** Igal mudelil on pühendatud [mänguväljak mudeli testimiseks](/md/02.QuickStart/GitHubModel_QuickStart.md). -### Phi Hugging Face’il +### Phi Hugging Face'is -Mudelit leiad ka [Hugging Face’ilt](https://huggingface.co/microsoft) +Mudeli leiate ka [Hugging Face'ist](https://huggingface.co/microsoft) **Mänguväljak** -[Hugging Chat mänguväljak](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) + [Hugging Chat mänguväljak](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Muud kursused + ## 🎒 Teised kursused -Meie meeskond toodab ka muid kursuseid! Vaata: +Meie meeskond toodab ka teisi kursuseid! Vaadake: ### LangChain @@ -261,64 +263,67 @@ Meie meeskond toodab ka muid kursuseid! Vaata: [![AZD algajatele](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI algajatele](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP algajatele](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI agendid algajatele](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Tehisintellekti agendid algajatele](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Generatiivse tehisintellekti sari -[![Generatiivne tehisintellekt algajatele](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generatiivne tehisintellekt (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generatiivne tehisintellekt (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generatiivne tehisintellekt (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +[![Generatiivne AI algajatele](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generatiivne AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generatiivne AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generatiivne AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Põhiõpe + +### Põhiteadmised [![Masinõpe algajatele](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Andmeteadus algajatele](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![Tehisintellekt algajatele](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Küberjulgeolek algajatele](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![AI algajatele](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Küberturvalisus algajatele](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Veebiarendus algajatele](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT algajatele](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![Asjade internet algajatele](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![XR arendus algajatele](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Copiloti sari -[![Copilot tehisintellekti paarisprogrammeerimiseks](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot C#/.NET-ile](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) + +### Copilot sari +[![Copilot AI paarisprogrammeerimiseks](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot C#/.NET jaoks](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot seiklus](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Vastutustundlik tehisintellekt +## Vastutustundlik AI + +Microsoft on pühendunud aitama klientidel kasutada meie AI tooteid vastutustundlikult, jagades oma kogemusi ja ehitades usaldusel põhinevaid partnerlusi tööriistadega nagu Läbipaistvuse märkmed ja Mõjuhinnangud. Paljusid neist ressurssidest leiab aadressilt [https://aka.ms/RAI](https://aka.ms/RAI). +Microsofti lähenemine vastutustundlikule AI-le põhineb meie AI põhimõtetel: õiglus, usaldusväärsus ja ohutus, privaatsus ja turvalisus, kaasatus, läbipaistvus ning vastutus. + +Suurte mahtude loomuliku keele-, pildi- ja kõnemudelid - nagu selles näites kasutatud mudelid - võivad käituda ebaõiglaselt, usaldusväärselt või solvavalt, põhjustades kahju. Palun vaadake [Azure OpenAI teenuse läbipaistvuse märkust](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), et olla teadlik riskidest ja piirangutest. -Microsoft on pühendunud aitama oma klientidel kasutada meie tehisintellekti tooteid vastutustundlikult, jagama õppetunde ja ehitama usaldusel põhinevaid partnerlussuhteid tööriistade nagu Läbipaistvusteadete ja Mõju hindamiste kaudu. Paljusid neid ressursse leiab aadressilt [https://aka.ms/RAI](https://aka.ms/RAI). -Microsofti vastutustundliku tehisintellekti lähenemine põhineb meie tehisintellekti põhimõtetel: õiglus, usaldusväärsus ja ohutus, privaatsus ja turvalisus, kaasatus, läbipaistvus ja vastutus. -Suurte looduskeele, pildi- ja kõnemudelite - nagu neid selles näites kasutatavaid - käitumine võib potentsiaalselt olla ebaõiglane, usaldamatu või solvav, põhjustades kahjustusi. Palun tutvu [Azure OpenAI teenuse läbipaistvuse teatisega](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), et saada teavet riskide ja piirangute kohta. -Soovitatav lähenemine nende riskide leevendamiseks on lisada oma arhitektuuri turvasüsteem, mis suudab tuvastada ja takistada kahjulikku käitumist. Azure AI sisuturve pakub sõltumatut kaitsekihki, mis suudab rakendustes ja teenustes tuvastada kahjulikku kasutajate loodud ja AI poolt genereeritud sisu. Azure AI sisuturve sisaldab teksti- ja pildirakendusliideseid, mis võimaldavad tuvastada kahjulikku materjali. Microsoft Foundrys võimaldab Content Safety teenus vaadata, uurida ja proovida näidiskoodi erinevate modaliteetide kahjuliku sisu tuvastamiseks. Järgmine kiirjuhendi dokumentatsioon juhatab sind teenusele päringute tegemisel. +Nende riskide maandamise soovitatav lähenemisviis on lisada oma arhitektuuri ohutussüsteem, mis suudab tuvastada ja vältida kahjulikku käitumist. [Azure AI sisu turvalisus](https://learn.microsoft.com/azure/ai-services/content-safety/overview) pakub sõltumatut kaitsekihi, mis suudab rakendustes ja teenustes tuvastada kahjulikku kasutaja- ja tehisintellekti loodud sisu. Azure AI sisu turvalisusesse kuuluvad tekstide ja piltide API-d, mis võimaldavad tuvastada kahjulikku materjali. Microsoft Foundry raames võimaldab Content Safety teenus vaadata, uurida ja proovida näitekoodi kahjuliku sisu tuvastamiseks erinevates modaliteetides. Järgmine [kiirjuhendi dokumentatsioon](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) juhendab teid teenusele taotluste tegemisel. -Teine oluline aspekt on rakenduse üldine jõudlus. Multi-modaalsete ja multi-mudelite rakenduste puhul peame jõudlust selleks, et süsteem toimiks nii, nagu sina ja su kasutajad ootavad, sealhulgas mitte tekitades kahjulikke väljundeid. On oluline hinnata oma rakenduse üldist jõudlust kasutades jõudluse, kvaliteedi ning riski ja turvalisuse hindajaid. Sul on ka võimalus luua ja hinnata kohandatud hindajatega. +Teine kaalutav aspekt on kogu rakenduse jõudlus. Mitmemodaalsete ja mitmemudelist rakenduste puhul mõistame jõudluse all seda, et süsteem toimib nii, nagu teie ja teie kasutajad ootavad, sealhulgas ei genereeri kahjulikke väljundeid. Oluline on hinnata oma kogu rakenduse jõudlust, kasutades [jõudluse ja kvaliteedi ning riski ja turvalisuse hindajaid](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Teil on ka võimalus luua ja hinnata [kohandatud hindajate](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) abil. -Sa saad hinnata oma AI rakendust oma arenduskeskkonnas, kasutades Azure AI hindamise SDK-d. Kasutades kas testandmestikku või sihtmärki, mõõdetakse sinu generatiivse AI rakenduse genereeringuid kvantitatiivselt sisseehitatud hindajate või sinu valitud kohandatud hindajate abil. Azure AI hindamise SDK-ga alustamiseks ja oma süsteemi hindamiseks võid järgida kiirjuhendit. Kui oled hindamise käigu käivitanud, saad tulemusi visualiseerida Microsoft Foundrys. +Võite oma AI-rakendust hinnata arenduskeskkonnas, kasutades [Azure AI hindamis-SDK-d](https://microsoft.github.io/promptflow/index.html). Kasutades testandmekogu või eesmärki, mõõdetakse teie generatiivse AI-rakenduse tulemusi kvantitatiivselt kas sisseehitatud või teie valitud kohandatud hindajatega. Alustamiseks Azure AI hindamis-SDK-ga süsteemi hindamiseks võite järgida [kiirjuhendit](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Kui olete hindamiskorra käivitanud, saate [tulemusi visualiseerida Microsoft Foundry's](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Kaubamärgid -See projekt võib sisaldada kaubamärke või logosid projektide, toodete või teenuste jaoks. Microsofti kaubamärkide või logode ametlik kasutamine peab toimuma vastavalt ja järgima Microsofti kaubamärgi ja brändi juhiseid. Microsofti kaubamärkide või logode kasutamine selle projekti muudetud versioonides ei tohi tekitada segadust ega jätta muljet Microsofti sponsorlusest. Kolmandate osapoolte kaubamärkide või logode kasutamine allub nende kolmandate osapoolte poliitikatele. +See projekt võib sisaldada projekte, tooteid või teenuseid tähistavaid kaubamärke või logosid. Microsofti kaubamärkide või logode autoriseeritud kasutamine allub ja peab järgima [Microsofti kaubamärgi- ja brändijuhiseid](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Microsofti kaubamärkide või logode kasutamine selle projekti muudetud versioonides ei tohi tekitada segadust ega viidata Microsofti sponsorlusele. Kolmandate osapoolte kaubamärkide või logode kasutamine allub nende kolmandate osapoolte poliitikatele. ## Abi saamine -Kui sul tekib takistusi või küsimusi AI-rakenduste arendamisel, liitu: +Kui teil tekib küsimusi või teil on raskusi AI-rakenduste loomisel, liituge: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Kui sul on tootepalju või ehitamisel esineb vigu, külasta: +Kui teil on toodete tagasisidet või vigu arendamisel, külastage: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Lahtiütlus**: -See dokument on tõlgitud tehisintellekti tõlketeenuse [Co-op Translator](https://github.com/Azure/co-op-translator) abil. Kuigi püüdleme täpsuse poole, palun arvestage, et automatiseeritud tõlked võivad sisaldada vigu või ebatäpsusi. Originaaldokument oma emakeeles tuleks pidada autoriteetseks allikaks. Kriitilise teabe puhul on soovitatav kasutada professionaalset inimtõlget. Me ei vastuta selle tõlke kasutamisest tulenevate arusaamatuste ega väärarusaamade eest. +**Lahtiütlus**: +See dokument on tõlgitud kasutades AI tõlketeenust [Co-op Translator](https://github.com/Azure/co-op-translator). Kuigi me püüdleme täpsuse poole, palun pange tähele, et automatiseeritud tõlgetes võib esineda vigu või ebatäpsusi. Originaaldokument selle emakeeles tuleks pidada autoriteetseks allikaks. Olulise teabe puhul soovitatakse kasutada professionaalset inimtõlget. Me ei vastuta selle tõlkega seotud eksimustest või valesti mõistmistest. \ No newline at end of file diff --git a/translations/et/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/et/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..96ddaa14c --- /dev/null +++ b/translations/et/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hübriid RAG Microsoft phi-4-mini ja SQLite FTS5-ga (Null-Pilve SLM)\n", + "\n", + "> **Autor:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Fookus:** Väikesed keelemudelid (SLMid), SQLite FTS5 BM25, Tihedad manused, Vastastikuse astme ühtlustamine (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivatsioon: Märksõnade meenutamise dilemma kohalikes SLMides\n", + "Tavalised RAG arhitektuurid, mis tuginevad ainult tihedatele vektorpõhistele manustele, ei suuda sageli leida täpseid numbrilisi märke (nt `2,340,000 TL`, lepingukoodid, kontonumbrid). \n", + "Vastupidiselt sellele jätab harvaesinev leksikaalne otsing (BM25) kahe silma vahele semantilised sünonüümid ja ümberpööratud küsimused.\n", + "\n", + "See teejuht näitab, kuidas rakendada **kõrge kiirusega mälus töötavat hübriidseid otsingumootorit**, ühendades:\n", + "1. **Tihedad vektorid** (Kosinus-sarnasuse alusel)\n", + "2. **Haarav leksikaalne otsing** (SQLite FTS5 BM25)\n", + "3. **Vastastikuse astme ühtlustamine (RRF, $k=60$)**\n", + "4. **Tõendatud viidete genereerimine (`[1]`, `[2]`)** Microsoft `phi-4-mini` abil.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Kahepoolne SQLite skeem (tihedad vektorid + virtuaalne FTS5 BM25 tabel)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Näidise põimumise ja täitmise võrdlustest\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Microsoft phi-4-mini jaoks rahuldavate küsimuste koostamine\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Lahtiütlus**:\nSee dokument on tõlgitud kasutades AI tõlketeenust [Co-op Translator](https://github.com/Azure/co-op-translator). Kuigi me püüdleme täpsuse poole, palun pange tähele, et automatiseeritud tõlgetes võib esineda vigu või ebatäpsusi. Originaaldokument selle emakeeles tuleks pidada autoriteetseks allikaks. Olulise teabe puhul soovitatakse kasutada professionaalset inimtõlget. Me ei vastuta selle tõlkega seotud eksimustest või valesti mõistmistest.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/fa/.co-op-translator.json b/translations/fa/.co-op-translator.json index 7c182880c..867925767 100644 --- a/translations/fa/.co-op-translator.json +++ b/translations/fa/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "fa" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T07:41:37+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:48:26+00:00", "source_file": "README.md", "language_code": "fa" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "fa" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:53:21+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "fa" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:29:40+00:00", diff --git a/translations/fa/README.md b/translations/fa/README.md index 86d819223..d02e115de 100644 --- a/translations/fa/README.md +++ b/translations/fa/README.md @@ -1,44 +1,44 @@ -# کتابچه آشپزی Phi: مثال‌های عملی با مدل‌های Phi مایکروسافت +# کتاب آشپزی Phi: مثال‌های عملی با مدل‌های Phi مایکروسافت -[![باز کردن و استفاده از نمونه‌ها در GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![باز کردن در Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![نمونه‌ها را در GitHub Codespaces باز و استفاده کنید](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![باز کردن در Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=بازکردن&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![همکاران GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![مشارکت‌کنندگان GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![مسائل GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![درخواست‌های pull در GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![خوش‌آمدگویی به PRها](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![درخواست‌های pull GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![درخواست‌های pull خوش‌آمدید](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![تماشاگران GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![ناظران GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![انشعابات GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![ستاره‌های GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) -[![Discord بنیاد مایکروسافت](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) +[![دیسکورد Microsoft Foundry](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi مجموعه‌ای از مدل‌های هوش مصنوعی متن‌باز است که توسط مایکروسافت توسعه یافته‌اند. +Phi مجموعه‌ای از مدل‌های هوش مصنوعی متن‌باز است که توسط مایکروسافت توسعه یافته است. -Phi در حال حاضر قدرتمندترین و اقتصادی‌ترین مدل زبان کوچک (SLM) است، با معیارهای بسیار خوب در چندزبانگی، استدلال، تولید متن/چت، کدنویسی، تصاویر، صدا و سایر سناریوها. +Phi در حال حاضر قدرتمندترین و مقرون به صرفه‌ترین مدل زبان کوچک (SLM) است که در معیارها در چند زبان، استدلال، تولید متن/چت، کدنویسی، تصاویر، صوت و سایر سناریوها عملکرد بسیار خوبی دارد. -شما می‌توانید Phi را در فضای ابری یا دستگاه‌های لبه پیاده‌سازی کنید و به‌سادگی برنامه‌های هوش مصنوعی مولد را با قدرت محاسباتی محدود بسازید. +شما می‌توانید Phi را در ابر یا دستگاه‌های لبه مستقر کنید و به‌راحتی برنامه‌های هوش مصنوعی مولد را با قدرت محاسباتی محدود بسازید. -برای شروع استفاده از این منابع، این مراحل را دنبال کنید: -1. **شعبه (Fork) مخزن را انجام دهید**: کلیک کنید [![انشعابات GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +برای شروع استفاده از این منابع مراحل زیر را دنبال کنید: +1. **انشعاب (Fork) مخزن را ایجاد کنید**: کلیک کنید [![انشعابات GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **مخزن را کلون کنید**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**به جامعه Discord هوش مصنوعی مایکروسافت بپیوندید و با کارشناسان و توسعه‌دهندگان دیگر آشنا شوید**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**به جامعه دیسکورد هوش مصنوعی مایکروسافت بپیوندید و با کارشناسان و همکاران توسعه‌دهنده ملاقات کنید**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/fa/cover.eb18d1b9605d754b.webp) +![کاور](../../translated_images/fa/cover.eb18d1b9605d754b.webp) ### 🌐 پشتیبانی چندزبانه -#### پشتیبانی شده از طریق GitHub Action (خودکار و همیشه به‌روز) +#### پشتیبانی از طریق فعل و انفعال GitHub (خودکار و همواره به‌روز) -[عربی](../ar/README.md) | [بنگالی](../bn/README.md) | [بلغاری](../bg/README.md) | [برمه‌ای (میانمار)](../my/README.md) | [چینی (ساده‌شده)](../zh-CN/README.md) | [چینی (سنتی، هنگ‌کنگ)](../zh-HK/README.md) | [چینی (سنتی، ماکائو)](../zh-MO/README.md) | [چینی (سنتی، تایوان)](../zh-TW/README.md) | [کرواتی](../hr/README.md) | [چکی](../cs/README.md) | [دانمارکی](../da/README.md) | [هلندی](../nl/README.md) | [استونیایی](../et/README.md) | [فنلاندی](../fi/README.md) | [فرانسوی](../fr/README.md) | [آلمانی](../de/README.md) | [یونانی](../el/README.md) | [عبری](../he/README.md) | [هندی](../hi/README.md) | [مجارستانی](../hu/README.md) | [اندونزیایی](../id/README.md) | [ایتالیایی](../it/README.md) | [ژاپنی](../ja/README.md) | [کانادایی](../kn/README.md) | [خمری](../km/README.md) | [کره‌ای](../ko/README.md) | [لیتوانیایی](../lt/README.md) | [مالایی](../ms/README.md) | [مالایالام](../ml/README.md) | [مراتی](../mr/README.md) | [نپالی](../ne/README.md) | [نیجریه‌ای پجدین](../pcm/README.md) | [نروژی](../no/README.md) | [فارسی (Farsi)](./README.md) | [لهستانی](../pl/README.md) | [پرتغالی (برزیل)](../pt-BR/README.md) | [پرتغالی (پرتغال)](../pt-PT/README.md) | [پنجابی (Gurmukhi)](../pa/README.md) | [رومانیایی](../ro/README.md) | [روسی](../ru/README.md) | [صربی (سیریلیک)](../sr/README.md) | [اسلواکی](../sk/README.md) | [اسلوونیایی](../sl/README.md) | [اسپانیایی](../es/README.md) | [سواحیلی](../sw/README.md) | [سوئدی](../sv/README.md) | [تاگالوگ (فیلیپینی)](../tl/README.md) | [تامیل](../ta/README.md) | [تلوگو](../te/README.md) | [تایلندی](../th/README.md) | [ترکی](../tr/README.md) | [اوکراینی](../uk/README.md) | [اردو](../ur/README.md) | [ویتنامی](../vi/README.md) +[عربی](../ar/README.md) | [بنگالی](../bn/README.md) | [بلغاری](../bg/README.md) | [برمه‌ای (میانمار)](../my/README.md) | [چینی (ساده‌شده)](../zh-CN/README.md) | [چینی (سنتی، هنگ‌کنگ)](../zh-HK/README.md) | [چینی (سنتی، ماکائو)](../zh-MO/README.md) | [چینی (سنتی، تایوان)](../zh-TW/README.md) | [کرواسی](../hr/README.md) | [چکی](../cs/README.md) | [دانمارکی](../da/README.md) | [هلندی](../nl/README.md) | [استونیایی](../et/README.md) | [فنلاندی](../fi/README.md) | [فرانسوی](../fr/README.md) | [آلمانی](../de/README.md) | [یونانی](../el/README.md) | [عبری](../he/README.md) | [هندی](../hi/README.md) | [مجارستانی](../hu/README.md) | [اندونزیایی](../id/README.md) | [ایتالیایی](../it/README.md) | [ژاپنی](../ja/README.md) | [کنادا](../kn/README.md) | [خمری](../km/README.md) | [کره‌ای](../ko/README.md) | [لیتوانیایی](../lt/README.md) | [مالایی](../ms/README.md) | [مالایالامی](../ml/README.md) | [مراتی](../mr/README.md) | [نپالی](../ne/README.md) | [پیدگین نیجریه‌ای](../pcm/README.md) | [نروژی](../no/README.md) | [فارسی (Farsi)](./README.md) | [لهستانی](../pl/README.md) | [پرتغالی (برزیل)](../pt-BR/README.md) | [پرتغالی (پرتغال)](../pt-PT/README.md) | [پنجابی (Gurmukhi)](../pa/README.md) | [رومانیایی](../ro/README.md) | [روسی](../ru/README.md) | [صرب (سیریلیک)](../sr/README.md) | [اسلواکی](../sk/README.md) | [اسلوونیایی](../sl/README.md) | [اسپانیایی](../es/README.md) | [سواحیلی](../sw/README.md) | [سوئدی](../sv/README.md) | [تاگالوگ (فیلیپینی)](../tl/README.md) | [تامیل](../ta/README.md) | [تلگو](../te/README.md) | [تایلندی](../th/README.md) | [ترکی](../tr/README.md) | [اوکراینی](../uk/README.md) | [اردو](../ur/README.md) | [ویتنامی](../vi/README.md) -> **ترجیح می‌دهید محلی کلون کنید؟** +> **ترجیح می‌دهید به‌صورت محلی کلون کنید؟** > -> این مخزن شامل بیش از ۵۰ ترجمه زبانی است که حجم دانلود را به طور قابل توجهی افزایش می‌دهد. برای کلون بدون ترجمه‌ها، از sparse checkout استفاده کنید: +> این مخزن شامل بیش از ۵۰ ترجمه زبانی است که اندازه دانلود را به طور قابل توجهی افزایش می‌دهد. برای کلون کردن بدون ترجمه‌ها، از بررسی پراکنده (sparse checkout) استفاده کنید: > -> **Bash / macOS / Linux:** +> **Bash / macOS / لینوکس:** > ```bash > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook @@ -52,30 +52,30 @@ Phi در حال حاضر قدرتمندترین و اقتصادی‌ترین م > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> این به شما همه چیز لازم برای تکمیل دوره را با دانلود بسیار سریع‌تر می‌دهد. +> این به شما همه چیز مورد نیاز برای تکمیل دوره را با دانلود بسیار سریع‌تر می‌دهد. ## فهرست مطالب - معرفی - - [خوش‌آمدگویی به خانواده Phi](./md/01.Introduction/01/01.PhiFamily.md) + - [خوش‌آمدید به خانواده Phi](./md/01.Introduction/01/01.PhiFamily.md) - [راه‌اندازی محیط شما](./md/01.Introduction/01/01.EnvironmentSetup.md) - [درک فناوری‌های کلیدی](./md/01.Introduction/01/01.Understandingtech.md) - [ایمنی هوش مصنوعی برای مدل‌های Phi](./md/01.Introduction/01/01.AISafety.md) - [پشتیبانی سخت‌افزاری Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [مدل‌های Phi و دسترسی آن‌ها در پلتفرم‌ها](./md/01.Introduction/01/01.Edgeandcloud.md) + - [مدل‌های Phi و در دسترس بودن در پلتفرم‌ها](./md/01.Introduction/01/01.Edgeandcloud.md) - [استفاده از Guidance-ai و Phi](./md/01.Introduction/01/01.Guidance.md) - [مدل‌های بازار GitHub](https://github.com/marketplace/models) - - [کاتالوگ مدل هوش مصنوعی Azure](https://ai.azure.com) + - [کاتالوگ مدل‌های AI Azure](https://ai.azure.com) - استنتاج Phi در محیط‌های مختلف - [Hugging face](./md/01.Introduction/02/01.HF.md) - [مدل‌های GitHub](./md/01.Introduction/02/02.GitHubModel.md) - - [کاتالوگ مدل Foundry مایکروسافت](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [کاتالوگ مدل مایکروسافت فاندری](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [جعبه‌ابزار هوش مصنوعی VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [ابزار هوش مصنوعی VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry محلی](./md/01.Introduction/02/07.FoundryLocal.md) + - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - استنتاج خانواده Phi - [استنتاج Phi در iOS](./md/01.Introduction/03/iOS_Inference.md) @@ -84,141 +84,144 @@ Phi در حال حاضر قدرتمندترین و اقتصادی‌ترین م - [استنتاج Phi در رایانه هوش مصنوعی](./md/01.Introduction/03/AIPC_Inference.md) - [استنتاج Phi با چارچوب Apple MLX](./md/01.Introduction/03/MLX_Inference.md) - [استنتاج Phi در سرور محلی](./md/01.Introduction/03/Local_Server_Inference.md) - - [استنتاج Phi در سرور راه دور با استفاده از جعبه‌ابزار هوش مصنوعی](./md/01.Introduction/03/Remote_Interence.md) + - [استنتاج Phi در سرور راه دور با استفاده از ابزار هوش مصنوعی](./md/01.Introduction/03/Remote_Interence.md) - [استنتاج Phi با Rust](./md/01.Introduction/03/Rust_Inference.md) - - [استنتاج Phi-دید در محلی](./md/01.Introduction/03/Vision_Inference.md) + - [استنتاج Phi--Vision در محل](./md/01.Introduction/03/Vision_Inference.md) - [استنتاج Phi با Kaito AKS، Azure Containers (پشتیبانی رسمی)](./md/01.Introduction/03/Kaito_Inference.md) -- [کوانتیزه کردن خانواده Phi](./md/01.Introduction/04/QuantifyingPhi.md) +- [کوانتیزه‌کردن خانواده Phi](./md/01.Introduction/04/QuantifyingPhi.md) - [کوانتیزه کردن Phi-3.5 / 4 با استفاده از llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [کوانتیزه کردن Phi-3.5 / 4 با استفاده از افزونه‌های هوش مصنوعی مولد برای onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [کوانتیزه کردن Phi-3.5 / 4 با استفاده از افزونه‌های هوش مولد برای onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [کوانتیزه کردن Phi-3.5 / 4 با استفاده از Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - [کوانتیزه کردن Phi-3.5 / 4 با استفاده از چارچوب Apple MLX](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - ارزیابی Phi - - [هوش مصنوعی مسئولانه](./md/01.Introduction/05/ResponsibleAI.md) - - [Foundry مایکروسافت برای ارزیابی](./md/01.Introduction/05/AIFoundry.md) + - [هوش مصنوعی مسئول](./md/01.Introduction/05/ResponsibleAI.md) + - [مایکروسافت فاندری برای ارزیابی](./md/01.Introduction/05/AIFoundry.md) - [استفاده از Promptflow برای ارزیابی](./md/01.Introduction/05/Promptflow.md) - RAG با جستجوی هوش مصنوعی Azure - - [چگونه از Phi-4-mini و Phi-4-multimodal (RAG) با جستجوی هوش مصنوعی Azure استفاده کنیم](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [نحوه استفاده از Phi-4-mini و Phi-4-multimodal (RAG) با جستجوی هوش مصنوعی Azure](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG هیبرید محلی صفرابری با SQLite FTS5 و phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - نمونه‌های توسعه برنامه Phi - برنامه‌های متن و چت - - نمونه‌های Phi-4 + - نمونه‌های Phi-4 - [📓] [چت با مدل ONNX Phi-4-mini](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [چت با مدل محلی ONNX Phi-4 با دات‌نت](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [برنامه کنسول چت دات‌نت با ONNX Phi-4 با استفاده از Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - نمونه‌های Phi-3 / 3.5 + - [چت با مدل محلی ONNX Phi-4 .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [برنامه کنسول چت .NET با استفاده از Phi-4 ONNX و Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - نمونه‌های فی-3 / 3.5 - [چت‌بات محلی در مرورگر با استفاده از Phi3، ONNX Runtime Web و WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [چت اوپن‌وینو](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [مدل چندگانه - فی-3‌-مینی تعاملی و اوپن‌ای وی ویسپر](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - ساخت یک پوشش دهنده و استفاده از فی-3 با MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [بهینه‌سازی مدل - چگونه مدل Phi-3-min را برای اجرای وب ONNX با Olive بهینه کنیم](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [برنامه WinUI3 با Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[نمونه برنامه یادداشت‌های چندمدل WinUI3 با هوش مصنوعی](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [تنظیم دقیق و یکپارچه‌سازی مدل‌های سفارشی Phi-3 با Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [تنظیم دقیق و یکپارچه‌سازی مدل‌های سفارشی Phi-3 با Prompt flow در مایکروسافت فاندری](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [ارزیابی مدل‌های تنظیم دقیق شده Phi-3 / Phi-3.5 در مایکروسافت فاندری با تمرکز بر اصول مسئولیت‌پذیری هوش مصنوعی مایکروسافت](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [چت OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [چند مدل – Phi-3-mini تعاملی و OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow – ساخت یک Wrappper و استفاده از Phi-3 با MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [بهینه‌سازی مدل – چگونه مدل Phi-3-min را برای ONNX Runtime Web با Olive بهینه کنیم](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [اپلیکیشن WinUI3 با Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + - [نمونه اپلیکیشن یادداشت چند مدلی با WinUI3 قدرت‌گرفته از هوش مصنوعی](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [فاین‌تیون و ادغام مدل‌های سفارشی Phi-3 با Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [فاین‌تیون و ادغام مدل‌های سفارشی Phi-3 با Prompt flow در Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [ارزیابی مدل فاین‌تیون شده Phi-3 / Phi-3.5 در Microsoft Foundry با تمرکز بر اصول هوش مصنوعی مسئولانه مایکروسافت](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [نمونه پیش‌بینی زبان Phi-3.5-mini-instruct (چینی/انگلیسی)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [چت‌بات RAG وب‌جی‌پی‌یو Phi-3.5-Instruct](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [استفاده از GPU ویندوز برای ایجاد راه‌حل Prompt flow با Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [استفاده از Microsoft Phi-3.5 tflite برای ساخت برنامه اندروید](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [مثال پرسش و پاسخ در دات‌نت با استفاده از مدل ONNX محلی Phi-3 با Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [برنامه کنسول چت دات‌نت با Semantic Kernel و Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - - نمونه‌های کد مبتنی بر SDK استنتاج Azure AI - - نمونه‌های Phi-4 + - [چت‌بات RAG تحت وب Phi-3.5-Instruct(WebGPU)](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [استفاده از GPU ویندوز برای ساخت راه‌حل Prompt flow با Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [ساخت اپلیکیشن اندروید با Microsoft Phi-3.5 tflite](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [نمونه Q&A دات‌نت با استفاده از مدل محلی ONNX Phi-3 با Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [اپلیکیشن کنسول چت دات‌نت با Semantic Kernel و Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + + - نمونه‌های مبتنی بر کد Azure AI Inference SDK + - نمونه‌های فی-4 - [📓] [تولید کد پروژه با استفاده از Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - نمونه‌های Phi-3 / 3.5 - - [ساخت چت GitHub Copilot در Visual Studio Code با خانواده Phi-3 مایکروسافت](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [ساخت عامل چت GitHub Copilot خودتان در Visual Studio Code با مدل‌های Phi-3.5](./md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - نمونه‌های فی-3 / 3.5 + - [ساخت چت GitHub Copilot برای ویژوال استودیو کد با خانواده Phi-3 مایکروسافت](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [ساخت نماینده چت GitHub Copilot برای ویژوال استودیو کد با Phi-3.5 توسط مدل‌های GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - نمونه‌های استدلال پیشرفته - - نمونه‌های Phi-4 - - [📓] [نمونه‌های استدلال Phi-4-mini یا Phi-4](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [تنظیم دقیق Phi-4-mini-reasoning با Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [تنظیم دقیق Phi-4-mini-reasoning با Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - نمونه‌های فی-4 + - [📓] [نمونه‌های Phi-4-mini-reasoning یا Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [فاین‌تیون Phi-4-mini-reasoning با Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [فاین‌تیون Phi-4-mini-reasoning با Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-reasoning با مدل‌های GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-reasoning با مدل‌های Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - دموها - [دموهای Phi-4-mini میزبانی شده در Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - [دموهای Phi-4-multimodal میزبانی شده در Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - نمونه‌های بینایی - - نمونه‌های Phi-4 + - نمونه‌های فی-4 - [📓] [استفاده از Phi-4-multimodal برای خواندن تصاویر و تولید کد](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - نمونه‌های Phi-3 / 3.5 - - [📓][متن به متن تصویر بینایی Phi-3](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3 بینایی ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][تعبیه CLIP برای بینایی Phi-3](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - نمونه‌های فی-3 / 3.5 + - [📓][Phi-3-vision-تبدیل متن تصویر به متن](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vision جاسازی CLIP](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [دمو: بازیافت Phi-3](https://github.com/jennifermarsman/PhiRecycling/) - [Phi-3-vision - دستیار زبان بصری - با Phi3-Vision و OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][نمونه چندفریم یا چندتصویر بینایی Phi-3.5](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [مدل ONNX محلی Phi-3 Vision با استفاده از Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [مدل ONNX محلی Phi-3 Vision با منوی مبتنی بر Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][نمونه چند فریمی یا چند تصویری Phi-3.5 Vision](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [مدل محلی ONNX Phi-3 Vision با استفاده از Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [مدل محلی ONNX Phi-3 Vision با منوی مبتنی بر Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - نمونه‌های استدلال-بینایی - Phi-4-Reasoning-Vision-15B - [📓] [استفاده از Phi-4-Reasoning-Vision-15B برای تشخیص عبور غیرمجاز از خیابان](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [استفاده از Phi-4-Reasoning-Vision-15B برای ریاضی](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [استفاده از Phi-4-Reasoning-Vision-15B برای ریاضیات](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - [📓] [استفاده از Phi-4-Reasoning-Vision-15B برای تشخیص رابط کاربری](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - نمونه‌های ریاضی - - نمونه‌های Phi-4-Mini-Flash-Reasoning-Instruct [دموی ریاضی با Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - نمونه‌های Phi-4-Mini-Flash-Reasoning-Instruct [دموی ریاضی با Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - نمونه‌های صوتی - - نمونه‌های Phi-4 - - [📓] [استخراج رونوشتی صوتی با استفاده از Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - نمونه‌های فی-4 + - [📓] [استخراج رونوشت‌های صوتی با استفاده از Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [نمونه صوتی Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [نمونه ترجمه گفتار Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [برنامه کنسول .NET با استفاده از Phi-4-multimodal برای تجزیه و تحلیل فایل صوتی و تولید رونوشتی](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [اپلیکیشن کنسول دات‌نت با استفاده از Phi-4-multimodal برای تحلیل فایل صوتی و تولید رونوشت](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - نمونه‌های MOE - - نمونه‌های Phi-3 / 3.5 - - [📓] [نمونه مدل‌های ترکیب کارشناسان Phi-3.5 Mixture of Experts (MoEs) برای رسانه‌های اجتماعی](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [ساخت خط لوله تولید بازیابی‌شده (RAG) با NVIDIA NIM Phi-3 MOE، جستجوی Azure AI و LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - نمونه‌های فی-3 / 3.5 + - [📓] [نمونه مدل‌های ترکیبی کارشناسان Phi-3.5 (MoEs) برای رسانه‌های اجتماعی](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [ساخت خط لوله تولید تکمیل بازیابی‌شده (RAG) با NVIDIA NIM Phi-3 MOE، Azure AI Search و LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - نمونه‌های فراخوانی توابع - - نمونه‌های Phi-4 🆕 - - [📓] [استفاده از فراخوانی تابع با Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [استفاده از فراخوانی تابع برای ایجاد چند عامل با Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [استفاده از فراخوانی تابع با Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [استفاده از فراخوانی تابع با ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - نمونه‌های مخلوط چندرسانه‌ای - - نمونه‌های Phi-4 🆕 - - [📓] [استفاده از Phi-4-multimodal به عنوان روزنامه‌نگار فناوری](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [برنامه کنسول .NET با استفاده از Phi-4-multimodal برای تجزیه و تحلیل تصاویر](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- تنظیم دقیق نمونه‌های Phi - - [سناریوهای تنظیم دقیق](./md/03.FineTuning/FineTuning_Scenarios.md) - - [تنظیم دقیق در مقابل RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [بگذارید Phi-3 به یک متخصص صنعتی تبدیل شود](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [تنظیم دقیق Phi-3 با ابزار AI برای VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [تنظیم دقیق Phi-3 با سرویس یادگیری ماشین Azure](./md/03.FineTuning/Introduce_AzureML.md) - - [تنظیم دقیق Phi-3 با لورا](./md/03.FineTuning/FineTuning_Lora.md) - - [تنظیم دقیق Phi-3 با QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [تنظیم دقیق Phi-3 با Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [تنظیم دقیق Phi-3 با Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [تنظیم دقیق با Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [آزمایشگاه عملی تنظیم دقیق با Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [تنظیم دقیق Phi-3-vision با Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - نمونه‌های فراخوانی تابع + - نمونه‌های فی-4 🆕 + - [📓] [استفاده از فراخوانی تابع با Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [استفاده از فراخوانی تابع برای ساخت چند نماینده با Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [استفاده از فراخوانی تابع با Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [استفاده از فراخوانی تابع با ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - نمونه‌های ترکیب چندرسانه‌ای + - نمونه‌های فی-4 🆕 + - [📓] [استفاده از Phi-4-multimodal به عنوان خبرنگار فناوری](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [اپلیکیشن کنسول دات‌نت برای تحلیل تصاویر با استفاده از Phi-4-multimodal](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- فاین‌تیون نمونه‌های Phi + - [سناریوهای فاین‌تیون](./md/03.FineTuning/FineTuning_Scenarios.md) + - [فاین‌تیون در مقابل RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [اجازه دهید Phi-3 تبدیل به یک کارشناس صنعتی شود](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [فاین‌تیون Phi-3 با AI Toolkit برای VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [فاین‌تیون Phi-3 با Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [فاین‌تیون Phi-3 با Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [فاین‌تیون Phi-3 با QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [فاین‌تیون Phi-3 با Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [فاین‌تیون Phi-3 با Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [فاین‌تیون با Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [کارگاه عملی فاین‌تیون با Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) + - [فاین‌تیون Phi-3-vision با Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [تنظیم دقیق Phi-3 با چارچوب Apple MLX](./md/03.FineTuning/FineTuning_MLX.md) - [تنظیم دقیق Phi-3-vision (پشتیبانی رسمی)](./md/03.FineTuning/FineTuning_Vision.md) - [تنظیم دقیق Phi-3 با Kaito AKS ، کانتینرهای Azure (پشتیبانی رسمی)](./md/03.FineTuning/FineTuning_Kaito.md) - - [تنظیم دقیق Phi-3 و Vision 3.5](https://github.com/2U1/Phi3-Vision-Finetune) + - [تنظیم دقیق Phi-3 و 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - آزمایشگاه عملی - - [کشف مدل‌های نوین: LLM ، SLM ، توسعه محلی و بیشتر](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [آزادسازی پتانسیل NLP: تنظیم دقیق با Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [بررسی مدل‌های پیشرفته: LLMها، SLMها، توسعه محلی و بیشتر](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [آزادسازی ظرفیت NLP: تنظیم دقیق با Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - مقالات و انتشارات پژوهشی دانشگاهی - - [کتاب‌های درسی همه چیز هستند II: گزارش فنی phi-1.5](https://arxiv.org/abs/2309.05463) - - [گزارش فنی Phi-3: یک مدل زبان بسیار توانمند به صورت محلی در تلفن شما](https://arxiv.org/abs/2404.14219) + - [کتاب‌های درسی همه چیز مورد نیاز است II: گزارش فنی phi-1.5](https://arxiv.org/abs/2309.05463) + - [گزارش فنی Phi-3: مدل زبان بسیار توانمند به صورت محلی روی گوشی شما](https://arxiv.org/abs/2404.14219) - [گزارش فنی Phi-4](https://arxiv.org/abs/2412.08905) - - [گزارش فنی Phi-4-Mini: مدل‌های زبان چندرسانه‌ای جمع‌وجور اما قدرتمند از طریق مخلوط LoRAها](https://arxiv.org/abs/2503.01743) - - [بهینه‌سازی مدل‌های زبان کوچک برای فراخوانی عملکرد درون‌خودرویی](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) تنظیم دقیق PHI-3 برای پاسخ به سوالات چندگزینه‌ای: روش‌شناسی، نتایج و چالش‌ها](https://arxiv.org/abs/2501.01588) + - [گزارش فنی Phi-4-Mini: مدل‌های چندرسانه‌ای قدرتمند و جمع‌وجور از طریق مخلوط LoRAها](https://arxiv.org/abs/2503.01743) + - [بهینه‌سازی مدل‌های زبان کوچک برای فراخوانی عملکرد در خودرو](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) تنظیم دقیق PHI-3 برای پاسخگویی به پرسش‌های چندجوابی: متدولوژی، نتایج و چالش‌ها](https://arxiv.org/abs/2501.01588) - [گزارش فنی استدلال Phi-4](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [گزارش فنی استدلال Phi-4-mini](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) @@ -226,28 +229,28 @@ Phi در حال حاضر قدرتمندترین و اقتصادی‌ترین م ### Phi در Microsoft Foundry -می‌توانید بیاموزید چگونه از Microsoft Phi استفاده کنید و چگونه راهکارهای انتها به انتها را در دستگاه‌های سخت‌افزاری مختلف خود بسازید. برای تجربه Phi برای خودتان، با بازی کردن با مدل‌ها و سفارشی‌سازی Phi برای سناریوهای خود شروع کنید با استفاده از [کاتالوگ مدل AI در Microsoft Foundry Azure](https://aka.ms/phi3-azure-ai) اطلاعات بیشتر در شروع کار با [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) قابل دسترسی است. +شما می‌توانید بیاموزید چگونه از Microsoft Phi استفاده کنید و چگونه راهکارهای پایان به پایان را در دستگاه‌های سخت‌افزاری مختلف خود بسازید. برای تجربه Phi خودتان، با آزمایش مدل‌ها و سفارشی‌سازی Phi برای سناریوهای خودتان شروع کنید با استفاده از [کاتالوگ مدل‌های Azure AI در Microsoft Foundry](https://aka.ms/phi3-azure-ai). می‌توانید اطلاعات بیشتری در شروع کار با [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) کسب کنید. **محیط آزمایش** -هر مدل یک محیط آزمایش اختصاصی برای تست مدل دارد [Azure AI Playground](https://aka.ms/try-phi3). +هر مدل دارای محیط آزمایشی اختصاصی برای تست مدل است [Azure AI Playground](https://aka.ms/try-phi3). ### Phi در مدل‌های GitHub -می‌توانید بیاموزید چگونه از Microsoft Phi استفاده کنید و چگونه راهکارهای انتها به انتها را در دستگاه‌های سخت‌افزاری مختلف خود بسازید. برای تجربه Phi برای خودتان، با بازی کردن با مدل و سفارشی‌سازی Phi برای سناریوهای خود شروع کنید با استفاده از [کاتالوگ مدل GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) اطلاعات بیشتر در شروع کار با [کاتالوگ مدل GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) موجود است. +شما می‌توانید بیاموزید چگونه از Microsoft Phi استفاده کنید و چگونه راهکارهای پایان به پایان را در دستگاه‌های سخت‌افزاری مختلف خود بسازید. برای تجربه Phi خودتان، با آزمایش مدل و سفارشی‌سازی Phi برای سناریوهای خودتان شروع کنید با استفاده از [کاتالوگ مدل GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). می‌توانید اطلاعات بیشتری در شروع کار با [کاتالوگ مدل GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) کسب کنید. **محیط آزمایش** -هر مدل دارای یک [محیط آزمایش اختصاصی برای تست مدل](/md/02.QuickStart/GitHubModel_QuickStart.md) می‌باشد. +هر مدل دارای [محیط آزمایشی اختصاصی برای تست مدل](/md/02.QuickStart/GitHubModel_QuickStart.md) است. ### Phi در Hugging Face -می‌توانید مدل را در [Hugging Face](https://huggingface.co/microsoft) نیز پیدا کنید. +همچنین می‌توانید مدل را در [Hugging Face](https://huggingface.co/microsoft) پیدا کنید. **محیط آزمایش** -[محیط بازی Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) + [محیط آزمایش چت Hugging](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 دوره‌های دیگر + ## 🎒 دوره‌های دیگر -تیم ما دوره‌های دیگری نیز تولید می‌کند! نگاهی بیندازید: +تیم ما دوره‌های دیگری نیز تولید می‌کند! مشاهده کنید: ### LangChain @@ -256,11 +259,11 @@ Phi در حال حاضر قدرتمندترین و اقتصادی‌ترین م [![LangChain برای مبتدیان](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agents +### Azure / Edge / MCP / نمایندگان [![AZD برای مبتدیان](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI برای مبتدیان](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP برای مبتدیان](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents برای مبتدیان](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![نمایندگان هوش مصنوعی برای مبتدیان](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- @@ -272,53 +275,55 @@ Phi در حال حاضر قدرتمندترین و اقتصادی‌ترین م --- -### آموزش‌های پایه +### یادگیری بنیادی [![یادگیری ماشین برای مبتدیان](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![علم داده برای مبتدیان](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![هوش مصنوعی برای مبتدیان](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![امنیت سایبری برای مبتدیان](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![توسعه وب برای مبتدیان](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![اینترنت اشیاء برای مبتدیان](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![اینترنت اشیا برای مبتدیان](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![توسعه XR برای مبتدیان](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### سری Copilot -[![Copilot برای برنامه‌نویسی جفتی هوش مصنوعی](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot برای C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![ماجراجویی Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +### سری همیار برنامه‌نویسی +[![همیار برنامه‌نویسی برای برنامه‌نویسی زوجی AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![همیار برای C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![ماجراجویی همیار](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## هوش مصنوعی مسئولانه -مایکروسافت متعهد است به کمک به مشتریان خود برای استفاده مسئولانه از محصولات هوش مصنوعی ما، اشتراک‌گذاری یادگیری‌هایمان و ایجاد شراکت‌های مبتنی بر اعتماد از طریق ابزارهایی مانند یادداشت‌های شفافیت و ارزیابی‌های تأثیر. بسیاری از این منابع را می‌توانید در [https://aka.ms/RAI](https://aka.ms/RAI) بیابید. -رویکرد مایکروسافت به هوش مصنوعی مسئولانه بر اصول هوش مصنوعی ما مبنی بر عدالت، قابلیت اطمینان و ایمنی، حریم خصوصی و امنیت، شمول‌پذیری، شفافیت و پاسخگویی استوار است. +مایکروسافت متعهد است به کمک به مشتریان خود برای استفاده مسئولانه از محصولات هوش مصنوعی، به اشتراک‌گذاری یادگیری‌های ما، و ساخت مشارکت‌های مبتنی بر اعتماد از طریق ابزارهایی مانند یادداشت‌های شفافیت و ارزیابی‌های تاثیر. بسیاری از این منابع را می‌توان در [https://aka.ms/RAI](https://aka.ms/RAI) یافت. +رویکرد مایکروسافت به هوش مصنوعی مسئولانه مبتنی بر اصول هوش مصنوعی ما یعنی عدالت، قابلیت اطمینان و ایمنی، حریم خصوصی و امنیت، فراگیری، شفافیت، و پاسخگویی است. + +مدل‌های بزرگ زبان طبیعی، تصویر و گفتار - مانند آن‌هایی که در این نمونه استفاده شده‌اند - ممکن است به‌ شیوه‌هایی رفتار کنند که ناعادلانه، غیرقابل اطمینان یا توهین‌آمیز باشند و در نتیجه آسیب‌هایی ایجاد کنند. لطفاً برای اطلاع از خطرات و محدودیت‌ها به [یادداشت شفافیت سرویس Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) مراجعه کنید. + -مدل‌های بزرگ زبان طبیعی، تصویر و گفتار - مانند مدل‌هایی که در این نمونه استفاده شده‌اند - ممکن است به گونه‌ای رفتار کنند که ناعادلانه، غیرقابل اعتماد یا توهین‌آمیز باشد و در نتیجه آسیب‌هایی به دنبال داشته باشد. لطفاً یادداشت شفافیت سرویس Azure OpenAI را برای آگاهی از ریسک‌ها و محدودیت‌ها مطالعه نمایید: [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text). -رویکرد پیشنهادی برای کاهش این ریسک‌ها، شامل کردن یک سیستم ایمنی در معماری شما است که بتواند رفتارهای مضر را تشخیص داده و جلوگیری کند. Azure AI Content Safety یک لایه مستقل از محافظت ارائه می‌دهد که قادر است محتوای مضر تولیدشده توسط کاربر و هوش مصنوعی را در برنامه‌ها و خدمات تشخیص دهد. Azure AI Content Safety شامل APIهای متن و تصویر است که به شما امکان تشخیص محتوای مضر را می‌دهد. در Microsoft Foundry، سرویس Content Safety به شما اجازه می‌دهد تا نمونه کدهای تشخیص محتوای مضر در حالت‌های مختلف را مشاهده، بررسی و امتحان کنید. مستندات شروع سریع زیر، شما را در ارسال درخواست‌ها به این سرویس راهنمایی می‌کند. +رویکرد پیشنهادی برای کاهش این ریسک‌ها، وارد کردن یک سیستم ایمنی در معماری شما است که بتواند رفتارهای مضر را تشخیص داده و جلوگیری کند. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) یک لایه محافظت مستقل فراهم می‌کند که قادر به شناسایی محتوای مضر تولید شده توسط کاربر و هوش مصنوعی در برنامه‌ها و خدمات است. Azure AI Content Safety شامل API‌های متن و تصویر است که به شما امکان می‌دهد محتوای مضر را شناسایی کنید. در Microsoft Foundry، سرویس Content Safety به شما امکان مشاهده، بررسی و آزمایش نمونه کد برای تشخیص محتوای مضر در مدالیتی‌های مختلف را می‌دهد. [مستندات شروع سریع](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) زیر شما را در ارسال درخواست به این سرویس راهنمایی می‌کند. -یکی دیگر از جنبه‌هایی که باید در نظر گرفته شود، عملکرد کلی برنامه است. در برنامه‌های چندمودال و چندمدلی، عملکرد به معنی این است که سیستم همانطور که شما و کاربران‌تان انتظار دارید عمل کند، از جمله عدم تولید خروجی‌های مضر. ارزیابی عملکرد کلی برنامه خود با استفاده از ارزیاب‌های عملکرد، کیفیت، ریسک و ایمنی اهمیت دارد. همچنین شما قادر به ایجاد و ارزیابی با ارزیاب‌های سفارشی هستید. +جنبه دیگری که باید در نظر گرفته شود عملکرد کلی برنامه است. با برنامه‌های چند مدالیتی و چند مدلی، عملکرد به این معناست که سیستم همانطور که شما و کاربران‌تان انتظار دارید عمل کند، از جمله عدم تولید خروجی‌های مضر. ارزیابی عملکرد کلی برنامه خود با استفاده از [ارزیاب‌های عملکرد، کیفیت، ریسک و ایمنی](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) مهم است. همچنین شما این توانایی را دارید که با [ارزیاب‌های سفارشی](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) بسازید و ارزیابی کنید. -شما می‌توانید برنامه هوش مصنوعی خود را در محیط توسعه خود با استفاده از Azure AI Evaluation SDK ارزیابی کنید. با توجه به یک مجموعه داده تست یا هدف مشخص، خروجی‌های تولیدشده توسط برنامه هوش مصنوعی مولد شما به طور کمی با ارزیاب‌های داخلی یا ارزیاب‌های سفارشی انتخابی شما اندازه‌گیری می‌شوند. برای شروع کار با Azure AI Evaluation SDK جهت ارزیابی سیستم خود، می‌توانید از راهنمای شروع سریع پیروی کنید. پس از اجرای یک ارزیابی، می‌توانید نتایج را در Microsoft Foundry مشاهده کنید. +می‌توانید برنامه هوش مصنوعی خود را در محیط توسعه خود با استفاده از [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ارزیابی کنید. با داشتن یک مجموعه داده آزمایشی یا هدف، تولیدهای برنامه هوش مصنوعی تولیدی شما به طور کمی با استفاده از ارزیاب‌های داخلی یا ارزیاب‌های سفارشی انتخابی شما اندازه‌گیری می‌شود. برای شروع کار با azure ai evaluation sdk جهت ارزیابی سیستم خود، می‌توانید از [راهنمای شروع سریع](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) پیروی کنید. پس از اجرای یک ارزیابی، می‌توانید [نتایج را در Microsoft Foundry مشاهده کنید](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## علائم تجاری -این پروژه ممکن است حاوی علائم تجاری یا لوگوهایی برای پروژه‌ها، محصولات یا خدمات باشد. استفاده مجاز از علائم تجاری یا لوگوهای مایکروسافت مشروط به رعایت راهنمای علائم تجاری و برند مایکروسافت است. -استفاده از علائم تجاری یا لوگوهای مایکروسافت در نسخه‌های تغییر یافته این پروژه نباید موجب سردرگمی شود یا حمایت مایکروسافت را القا کند. هرگونه استفاده از علائم تجاری یا لوگوهای طرف سوم تابع سیاست‌های آن طرف‌های سوم است. +این پروژه ممکن است شامل علائم تجاری یا لوگوهایی برای پروژه‌ها، محصولات یا خدمات باشد. استفاده مجاز از علائم تجاری یا لوگوهای مایکروسافت منوط به و باید مطابق با [راهنمای علائم تجاری و برند مایکروسافت](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) باشد. +استفاده از علائم تجاری یا لوگوهای مایکروسافت در نسخه‌های تغییر یافته این پروژه نباید باعث سردرگمی شود یا به حمایت مایکروسافت دلالت کند. هرگونه استفاده از علائم تجاری یا لوگوهای شخص ثالث تابع سیاست‌های آن شخص ثالث است. ## دریافت کمک -اگر به مشکل برخوردید یا سوالی درباره ساخت برنامه‌های هوش مصنوعی دارید، بپیوندید به: +اگر در ساخت برنامه‌های هوش مصنوعی به مشکل برخوردید یا سوالی داشتید، به موارد زیر بپیوندید: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -اگر بازخورد محصول یا خطا هنگام ساخت داشتید، به مراجعه کنید: +اگر بازخورد محصول یا خطایی هنگام ساخت دارید، بازدید کنید: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**توضیح مهم**: -این سند با استفاده از سرویس ترجمه هوش مصنوعی [Co-op Translator](https://github.com/Azure/co-op-translator) ترجمه شده است. در حالی که ما در تلاش برای دقت هستیم، لطفاً توجه داشته باشید که ترجمه‌های خودکار ممکن است شامل خطاها یا نادرستی‌هایی باشند. سند اصلی به زبان اصلی آن باید به عنوان منبع معتبر در نظر گرفته شود. برای اطلاعات حیاتی، ترجمه حرفه‌ای انسانی توصیه می‌شود. ما مسئول هیچ گونه سوءتفاهم یا تفسیر نادرستی که از استفاده این ترجمه ناشی شود، نیستیم. +**سلب مسئولیت**: +این سند با استفاده از سرویس ترجمه هوش مصنوعی [Co-op Translator](https://github.com/Azure/co-op-translator) ترجمه شده است. در حالی که ما در تلاش برای دقت هستیم، لطفاً توجه داشته باشید که ترجمه‌های خودکار ممکن است شامل خطاها یا نادرستی‌هایی باشند. سند اصلی به زبان مادری خود باید به عنوان منبع معتبر در نظر گرفته شود. برای اطلاعات حیاتی، ترجمه حرفه‌ای انسانی توصیه می‌شود. ما در قبال هرگونه سوء تفاهم یا برداشت نادرست ناشی از استفاده از این ترجمه مسئولیتی نداریم. \ No newline at end of file diff --git a/translations/fa/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/fa/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..40451a468 --- /dev/null +++ b/translations/fa/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 بازیابی ترکیبی RAG با مدل Microsoft phi-4-mini و SQLite FTS5 (مدل‌های زبان کوچک بدون کلود)\n", + "\n", + "> **نویسنده:** Çağrı Giray Keşان ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **تمرکز:** مدل‌های زبان کوچک (SLM)، SQLite FTS5 BM25، تعبیه‌های چگال، ادغام رتبه متقابل (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 ۱. انگیزه: معضل بازیابی کلیدواژه‌ها در مدل‌های محلی SLM\n", + "معماری‌های استاندارد RAG که صرفاً به تعبیه‌های بردار چگال بستگی دارند اغلب در بازیابی توکن‌های عددی دقیق (مثلاً `۲٬۳۴۰٬۰۰۰ TL`، کدهای قرارداد، شماره حساب) ناکام می‌مانند. \n", + "در مقابل، جستجوی واژگانی پراکنده (BM25) هم‌معناها و سوالات بازنویسی‌شده را از دست می‌دهد.\n", + "\n", + "این راهنما نشان می‌دهد چگونه یک **موتور بازیابی ترکیبی پرسرعت در حافظه** پیاده‌سازی کنیم که شامل:\n", + "۱. **بردارهای چگال** (شباهت کسینوسی)\n", + "۲. **جستجوی واژگانی پراکنده** (SQLite FTS5 BM25)\n", + "۳. **ادغام رتبه متقابل (RRF, $k=60$)**\n", + "۴. **تولید استنادهای مستدل (`[1]`, `[2]`)** با مدل Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ ۲. طرح‌واره دوگانه SQLite (بردارهای چگال + جدول مجازی FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 ۳. معیار دریافت نمونه و اجرای آن\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 ۴. تدوین پرامپت مبتنی بر داده برای Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**سلب مسئولیت**:\nاین سند با استفاده از سرویس ترجمه هوش مصنوعی [Co-op Translator](https://github.com/Azure/co-op-translator) ترجمه شده است. در حالی که ما در تلاش برای دقت هستیم، لطفاً توجه داشته باشید که ترجمه‌های خودکار ممکن است شامل خطاها یا نادرستی‌هایی باشند. سند اصلی به زبان مادری خود باید به عنوان منبع معتبر در نظر گرفته شود. برای اطلاعات حیاتی، ترجمه حرفه‌ای انسانی توصیه می‌شود. ما در قبال هرگونه سوء تفاهم یا برداشت نادرست ناشی از استفاده از این ترجمه مسئولیتی نداریم.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/fi/.co-op-translator.json b/translations/fi/.co-op-translator.json index 932918ccb..b41e9db5e 100644 --- a/translations/fi/.co-op-translator.json +++ b/translations/fi/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "fi" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:09:18+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:18:44+00:00", "source_file": "README.md", "language_code": "fi" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "fi" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:19:37+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "fi" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:19+00:00", diff --git a/translations/fi/README.md b/translations/fi/README.md index 89d8924dc..62db3e5db 100644 --- a/translations/fi/README.md +++ b/translations/fi/README.md @@ -1,42 +1,42 @@ # Phi Cookbook: Käytännön esimerkkejä Microsoftin Phi-malleilla -[![Avaa ja käytä esimerkkejä GitHub Codespacesissa](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Avaa ja käytä näytteitä GitHub Codespacesissa](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Avaa Dev Containersissa](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub-yhteistyöntekijät](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub-ongelmat](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub-pyyntöjen vetopyynnöt](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHubin kontribuuttorit](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHubin ongelmat](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHubin pull-pyynnöt](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![PR:t tervetulleita](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub-tarkkailijat](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub-haarat](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub-tähdet](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHubin tarkkailijat](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHubin haarukat](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHubin tähdet](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi on Microsoftin kehittämä sarja avoimen lähdekoodin tekoälymalleja. +Phi on Microsoftin kehittämä avoimen lähdekoodin tekoälymallien sarja. -Phi on tällä hetkellä tehokkain ja kustannustehokkain pieni kielimalli (SLM), jolla on erittäin hyvät suorituskykyvertailut monikielisyydessä, päättelyssä, tekstin/keskustelun generoinnissa, koodauksessa, kuvissa, äänissä ja muissa käyttötapauksissa. +Phi on tällä hetkellä tehokkain ja kustannustehokkain pieni kielimalli (SLM), jolla on erittäin hyvät vertailuarvot monikielisyydessä, päättelyssä, tekstin/keskustelun generoinnissa, koodauksessa, kuvissa, äänissä ja muissa käyttötapauksissa. -Voit ottaa Phin käyttöön pilvessä tai reunalaitteissa, ja voit helposti rakentaa generatiivisia tekoälysovelluksia rajallisella laskentateholla. +Voit ottaa Filip käyttöön pilvessä tai reunalaitteilla ja rakentaa helposti generatiivisia tekoälysovelluksia rajallisella laskentateholla. -Aloita käyttämällä näitä resursseja seuraavasti: -1. **Forkkaa repositorio**: Klikkaa [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +Seuraa näitä vaiheita aloittaaksesi näiden resurssien käytön: +1. **Haarauta repositorio**: Klikkaa [![GitHubin haarukat](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Kloonaa repositorio**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Liity Microsoft AI Discord -yhteisöön ja tapaa asiantuntijoita ja muita kehittäjiä**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Liity Microsoft AI Discord -yhteisöön tapaa asiantuntijoita ja muita kehittäjiä**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/fi/cover.eb18d1b9605d754b.webp) +![kansi](../../translated_images/fi/cover.eb18d1b9605d754b.webp) -### 🌐 Monikielinen tuki +### 🌐 Monikielituki -#### Tuettu GitHub Actionin kautta (automaattinen & aina ajan tasalla) +#### Tuettu GitHub Actionin kautta (automaattinen ja aina ajan tasalla) -[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](./README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) +[Arabia](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgaria](../bg/README.md) | [Burma (Myanmar)](../my/README.md) | [Kiina (yksinkertaistettu)](../zh-CN/README.md) | [Kiina (perinteinen, Hong Kong)](../zh-HK/README.md) | [Kiina (perinteinen, Macao)](../zh-MO/README.md) | [Kiina (perinteinen, Taiwan)](../zh-TW/README.md) | [Kroatia](../hr/README.md) | [Tšekki](../cs/README.md) | [Tanska](../da/README.md) | [Hollanti](../nl/README.md) | [Viro](../et/README.md) | [Suomi](./README.md) | [Ranska](../fr/README.md) | [Saksa](../de/README.md) | [Kreikka](../el/README.md) | [Heprean kieli](../he/README.md) | [Hindi](../hi/README.md) | [Unkari](../hu/README.md) | [Indonesia](../id/README.md) | [Italia](../it/README.md) | [Japani](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korea](../ko/README.md) | [Liettua](../lt/README.md) | [Malaiji](../ms/README.md) | [Malajalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norja](../no/README.md) | [Persia (Farsi)](../fa/README.md) | [Puola](../pl/README.md) | [Portugali (Brasilia)](../pt-BR/README.md) | [Portugali (Portugali)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romania](../ro/README.md) | [Venäjä](../ru/README.md) | [Serbia (kyrillinen)](../sr/README.md) | [Slovakki](../sk/README.md) | [Sloveeni](../sl/README.md) | [Espanja](../es/README.md) | [Swahili](../sw/README.md) | [Ruotsi](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamili](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkki](../tr/README.md) | [Ukraina](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnam](../vi/README.md) > **Haluatko mieluummin kloonata paikallisesti?** > -> Tämä repositorio sisältää yli 50 kielen käännökset, mikä lisää merkittävästi latauskoon määrää. Jos haluat kloonata ilman käännöksiä, käytä sparse checkoutia: +> Tämä repositorio sisältää yli 50 kielikäännöstä, mikä lisää huomattavasti latauskokoa. Jos haluat kloonata ilman käännöksiä, käytä sparse checkout -toimintoa: > > **Bash / macOS / Linux:** > ```bash @@ -52,202 +52,205 @@ Aloita käyttämällä näitä resursseja seuraavasti: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Saat näin kaiken tarvitsemasi kurssin suorittamiseen paljon nopeammalla latauksella. +> Tämä antaa sinulle kaiken tarvittavan kurssin läpiviemiseksi paljon nopeammalla latauksella. ## Sisällysluettelo - Johdanto - [Tervetuloa Phi-perheeseen](./md/01.Introduction/01/01.PhiFamily.md) - - [Ympäristön asennus](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Keskeisten teknologioiden ymmärtäminen](./md/01.Introduction/01/01.Understandingtech.md) + - [Ympäristön asetukset](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Ymmärtäen keskeisiä teknologioita](./md/01.Introduction/01/01.Understandingtech.md) - [Tekoälyn turvallisuus Phi-malleissa](./md/01.Introduction/01/01.AISafety.md) - - [Phi-laitetuki](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi-mallit ja saatavuus alustoilla](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Phi-laitteistotuki](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Phi-mallit ja saatavuus eri alustoilla](./md/01.Introduction/01/01.Edgeandcloud.md) - [Guidance-ai:n ja Phin käyttö](./md/01.Introduction/01/01.Guidance.md) - [GitHub Marketplace -mallit](https://github.com/marketplace/models) - - [Azure AI -malliluettelo](https://ai.azure.com) + - [Azure AI Malliluettelo](https://ai.azure.com) -- Phi-päättely eri ympäristöissä - - [Hugging Face](./md/01.Introduction/02/01.HF.md) +- Phi-päätelmät eri ympäristöissä + - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub-mallit](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry Model Catalog](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Microsoft Foundryn malliluettelo](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Phi-perheen päättely - - [Phi-päättely iOS:ssä](./md/01.Introduction/03/iOS_Inference.md) - - [Phi-päättely Androidissa](./md/01.Introduction/03/Android_Inference.md) - - [Phi-päättely Jetsonissa](./md/01.Introduction/03/Jetson_Inference.md) - - [Phi-päättely AI-PC:ssä](./md/01.Introduction/03/AIPC_Inference.md) - - [Phi-päättely Apple MLX -kehyksellä](./md/01.Introduction/03/MLX_Inference.md) - - [Phi-päättely paikallisella palvelimella](./md/01.Introduction/03/Local_Server_Inference.md) - - [Phi-päättely etäpalvelimella AI Toolkitin avulla](./md/01.Introduction/03/Remote_Interence.md) - - [Phi-päättely Rustilla](./md/01.Introduction/03/Rust_Inference.md) - - [Phi-päättely – Vision paikallisesti](./md/01.Introduction/03/Vision_Inference.md) - - [Phi-päättely Kaiton AKS:llä, Azure Containersilla (virallinen tuki)](./md/01.Introduction/03/Kaito_Inference.md) -- [Phin kvantisointi](./md/01.Introduction/04/QuantifyingPhi.md) - - [Phi-3.5 / 4:n kvantisointi llama.cpp:llä](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Phi-3.5 / 4:n kvantisointi Generative AI -laajennuksilla onnxruntimeen](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Phi-3.5 / 4:n kvantisointi Intel OpenVINO:lla](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Phi-3.5 / 4:n kvantisointi Apple MLX -kehyksellä](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) +- Phi-perheen päätelmät + - [Phi-päätelmät iOS:ssä](./md/01.Introduction/03/iOS_Inference.md) + - [Phi-päätelmät Androidissa](./md/01.Introduction/03/Android_Inference.md) + - [Phi-päätelmät Jetsossa](./md/01.Introduction/03/Jetson_Inference.md) + - [Phi-päätelmät AI-PC:ssä](./md/01.Introduction/03/AIPC_Inference.md) + - [Phi-päätelmät Apple MLX -kehyksen avulla](./md/01.Introduction/03/MLX_Inference.md) + - [Phi-päätelmät paikallisella palvelimella](./md/01.Introduction/03/Local_Server_Inference.md) + - [Phi-päätelmät etäpalvelimella AI Toolkitin avulla](./md/01.Introduction/03/Remote_Interence.md) + - [Phi-päätelmät Rustilla](./md/01.Introduction/03/Rust_Inference.md) + - [Phi-päätelmät Visio paikallisena](./md/01.Introduction/03/Vision_Inference.md) + - [Phi-päätelmät Kaito AKS:n ja Azure Containersin (virallinen tuki) kanssa](./md/01.Introduction/03/Kaito_Inference.md) +- [Phi-perheen kvantisointi](./md/01.Introduction/04/QuantifyingPhi.md) + - [Phi-3.5 / 4 kvantisointi llama.cpp:n avulla](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Phi-3.5 / 4 kvantisointi Generatiivisen tekoälyn laajennuksilla onnxruntimeen](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Phi-3.5 / 4 kvantisointi Intel OpenVINO:lla](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Phi-3.5 / 4 kvantisointi Apple MLX -kehyksellä](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Phi:n arviointi - [Vastuullinen tekoäly](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry arviointiin](./md/01.Introduction/05/AIFoundry.md) - - [Promptflow’n käyttö arvioinnissa](./md/01.Introduction/05/Promptflow.md) + - [Promptflow:n käyttö arvioinnissa](./md/01.Introduction/05/Promptflow.md) -- RAG Azure AI Searchin kanssa - - [Miten käyttää Phi-4-mini ja Phi-4-multimodal (RAG) Azure AI Searchin kanssa](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- RAG Azure AI Haulla + - [Phi-4-mini ja Phi-4-multimodal (RAG) käyttö Azure AI Haun kanssa](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud Paikallinen Hybrid RAG SQLite FTS5:n ja phi-4-minin kanssa](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi-sovelluskehityksen esimerkit +- Phi-sovelluskehityksen näytteet - Teksti- ja keskustelusovellukset - - Phi-4-esimerkit + - Phi-4-näytteet - [📓] [Keskustele Phi-4-mini ONNX -mallin kanssa](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Keskustele Phi-4 paikallisen ONNX-mallin kanssa .NET:llä](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Keskustelu .NET-konsolisovelluksella Phi-4 ONNX:n kanssa käyttäen Semantic Kernelia](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 -esimerkit - - [Paikallinen chatbot selaimessa käyttäen Phi3:ta, ONNX Runtime Webiä ja WebGPU:ta](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Keskustelusovellus .NET-konsolissa Phi-4 ONNX:n kanssa käyttäen Semantic Kernelia](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Näytteet + - [Paikallinen chatbot selaimessa käyttäen Phi3, ONNX Runtime Web ja WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - Interaktiivinen Phi-3-mini ja OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Kääreen rakentaminen ja Phi-3:n käyttäminen MLFlow'n kanssa](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Mallin optimointi - Kuinka optimoida Phi-3-minimalli ONNX Runtime Webille käyttäen Olivea](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3-sovellus Phi-3 mini-4k-instruct-onnx:llä](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Monimallin tekoälyllä varustettu muistiinpanosovellusnäyte](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Phi-3 -mallien hienosäätö ja integrointi Prompt Flow'n avulla](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Phi-3 -mallien hienosäätö ja integrointi Prompt Flow'n avulla Microsoft Foundryssa](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Hienosäädetyn Phi-3 / Phi-3.5 mallin arviointi Microsoft Foundryssa keskittyen Microsoftin vastuullisen tekoälyn periaatteisiin](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct kieliennusteenäyte (kiina/englanti)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Monimalli - Interaktiivinen Phi-3-mini ja OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Rungon rakentaminen ja Phi-3:n käyttäminen MLFlow'n kanssa](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Mallin optimointi - Kuinka optimoida Phi-3-minimalli ONNX Runtime Webille Oliven avulla](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3-sovellus Phi-3 mini-4k-instruct-onnx:n kanssa](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 Monimalli AI Powered Notes -sovellusnäyte](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Hienosäädä ja integroi mukautetut Phi-3-mallit Prompt flown kanssa](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Hienosäädä ja integroi mukautetut Phi-3-mallit Prompt flown kanssa Microsoft Foundryssä](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Arvioi hienosäädetty Phi-3 / Phi-3.5 malli Microsoft Foundryssä keskittyen Microsoftin vastuullisen tekoälyn periaatteisiin](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct kieliennustusnäyte (kiina/englanti)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU:n käyttäminen Prompt Flow -ratkaisun luomiseen Phi-3.5-Instruct ONNX:llä](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Windowsin GPU:n käyttäminen Prompt flow -ratkaisun luomiseen Phi-3.5-Instruct ONNX:n kanssa](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Microsoft Phi-3.5 tflite:n käyttäminen Android-sovelluksen luomiseen](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Kysymys & vastaus .NET-esimerkki paikallisella ONNX Phi-3 -mallilla käyttäen Microsoft.ML.OnnxRuntimea](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Konsolipohjainen chat .NET-sovellus Semantic Kernelillä ja Phi-3:lla](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Kyselyt ja vastaukset .NET-esimerkki paikallisen ONNX Phi-3 -mallin kanssa käyttäen Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Konsolipohjainen chat .NET-sovellus Semantic Kernelin ja Phi-3:n kanssa](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK:n koodipohjaiset näytteet - - Phi-4-näytteet - - [📓] [Projektikoodin generointi käyttäen Phi-4-multimodaalia](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 näytteet - - [Rakenna oma Visual Studio Code GitHub Copilot Chat Microsoft Phi-3 -perheen kanssa](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Luo oma Visual Studio Code Chat Copilot -agentti Phi-3.5:llä GitHub-mallien avulla](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - Azure AI Inference SDK -koodinäytteet + - Phi-4 Näytteet + - [📓] [Luo projektikoodi käyttäen Phi-4-multimodelia](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Phi-3 / 3.5 Näytteet + - [Rakenna oma Visual Studio Code GitHub Copilot Chat Microsoft Phi-3 -perheellä](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Luo oma Visual Studio Code Chat Copilot -agentti Phi-3.5:n avulla GitHub-malleilla](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Edistyneet päättelynäytteet - - Phi-4-näytteet - - [📓] [Phi-4-mini-päättely tai Phi-4-päättelynäytteet](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Phi-4-mini-päättelyn hienosäätö Microsoft Oliven avulla](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-päättelyn hienosäätö Apple MLX:n avulla](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-päättely GitHub-malleilla](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-päättely Microsoft Foundry -malleilla](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demos - - [Phi-4-mini esittelyt hostattuna Hugging Face Spacesissa](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal esittelyt hostattuna Hugginge Face Spacesissa](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Näkökykynäytteet - - Phi-4-näytteet - - [📓] [Käytä Phi-4-multimodaalia lukemaan kuvia ja generoimaan koodia](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 näytteet - - [📓][Phi-3-näkökuvan teksti tekstiksi](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-näkö ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-näkö CLIP upotus](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - Phi-4 Näytteet + - [📓] [Phi-4-mini-päättely- tai Phi-4-päättelynäytteet](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Hienosäätö Phi-4-mini-päättelylle Microsoft Oliven avulla](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Hienosäätö Phi-4-mini-päättelylle Apple MLX:n avulla](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-päättely GitHub-mallien kanssa](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-päättely Microsoft Foundry -mallien kanssa](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Demoja + - [Phi-4-mini demot Hugging Face Spaces -palvelussa](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal demot Hugging Face Spaces -palvelussa](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Vision Näytteet + - Phi-4 Näytteet + - [📓] [Käytä Phi-4-multimodalia kuvien lukemiseen ja koodin generointiin](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 Näytteet + - [📓][Phi-3-vision-Kuva tekstiksi-tekstiksi](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vision CLIP -upotus](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [DEMO: Phi-3 kierrätys](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-näkö - Visuaalinen kieliavustaja - Phi3-Näkö ja OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Näkö Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Näkö OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Näkö monikehys- tai monikuvaesimerkki](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Näkö paikallinen ONNX-malli käyttäen Microsoft.ML.OnnxRuntime .NETissä](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Valikkopohjainen Phi-3 Näkö paikallinen ONNX-malli käyttäen Microsoft.ML.OnnxRuntime .NETissä](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Päättely-Näkö Näytteet - - Phi-4-Päättely-Näkö-15B - - [📓] [Käyttämällä Phi-4-Päättely-Näkö-15B:tä vilkkaan liikenteen havaitsemiseen](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Käyttämällä Phi-4-Päättely-Näkö-15B:tä matematiikkaan](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Käyttämällä Phi-4-Päättely-Näkö-15B:tä käyttöliittymän tunnistamiseen](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - Matematiikan näytteet - - Phi-4-Mini-Flash-Päättely-Ohje Näytteet [Matematiikan demo Phi-4-Mini-Flash-Päättely-Ohjeella](./md/02.Application/09.Math/MathDemo.ipynb) - - - Ääninäytteet - - Phi-4-näytteet - - [📓] [Äänitallenteiden puhtaaksikirjoitus Phi-4-multimodaalilla](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodaalinen ääninäyte](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodaalinen puheen käännösnäyte](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET-konsolisovellus Phi-4-multimodaalista äänen analysointiin ja puhtaaksikirjoituksen luomiseen](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - MOE-näytteet - - Phi-3 / 3.5 näytteet - - [📓] [Phi-3.5 Eksperttisekoitusmallit (MoEs) sosiaalisen median näyte](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Hakua parantavan generoinnin (RAG) putken rakentaminen NVIDIA NIM Phi-3 MOE:n, Azure AI Searchin ja LlamaIndexin avulla](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [Phi-3-vision - Visuaalinen kieliavustaja - Phi3-Visionin ja OpenVINO:n kanssa](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision monikehys- tai monikuva-näyte](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision paikallinen ONNX-malli käyttäen Microsoft.ML.OnnxRuntime .NET:ää](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Valikkopohjainen Phi-3 Vision paikallinen ONNX-malli käyttäen Microsoft.ML.OnnxRuntime .NET:ää](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Päättely-Vision Näytteet + - Phi-4-Päättely-Vision-15B + - [📓] [Phi-4-Päättely-Vision-15B:n käyttäminen vilkun tunnistukseen](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-Päättely-Vision-15B:n käyttäminen matematiikkaan](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Phi-4-Päättely-Vision-15B:n käyttäminen käyttöliittymän tunnistukseen](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - Matematiikka Näytteet + - Phi-4-Mini-Flash-Päättely-Ohjeistettu Näytteet [Matematiikka Demo Phi-4-Mini-Flash-Päättely-Ohjeistetulla](./md/02.Application/09.Math/MathDemo.ipynb) + + - Audio Näytteet + - Phi-4 Näytteet + - [📓] [Äänitekstien purku Phi-4-multimodalilla](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal Audio Näyte](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal Puheentekstitys näyte](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET-konsolisovellus käyttäen Phi-4-multimodalia analysoimaan äänitiedosto ja generoimaan tekstityksen](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE Näytteet + - Phi-3 / 3.5 Näytteet + - [📓] [Phi-3.5 Asiantuntijoiden sekoitus (MoEs) Sosiaalisen median näyte](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Rakentaminen RAG-putkisto (Retrieval-Augmented Generation) NVIDIA NIM Phi-3 MOE:n, Azure AI Searchin ja LlamaIndexin kanssa](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Funktiokutsunäytteet - - Phi-4-näytteet 🆕 + - Funktiokutsut Näytteet + - Phi-4 Näytteet 🆕 - [📓] [Funktiokutsun käyttäminen Phi-4-minin kanssa](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Funktiokutsun käyttäminen moniagenttien luomiseen Phi-4-minin kanssa](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Funktiokutsun käyttäminen moniagenttien luomiseksi Phi-4-minillä](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Funktiokutsun käyttäminen Ollaman kanssa](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [Funktiokutsun käyttäminen ONNX:n kanssa](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Monimodaalisen yhdistämisen näytteet - - Phi-4-näytteet 🆕 - - [📓] [Phi-4-multimodaalin käyttäminen teknologiatoimittajana](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET-konsolisovellus, joka käyttää Phi-4-multimodaalia kuvien analysointiin](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - Monimodaalinen sekoittaminen Näytteet + - Phi-4 Näytteet 🆕 + - [📓] [Phi-4-multimodalin käyttäminen teknologiatoimittajana](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET-konsolisovellus käyttäen Phi-4-multimodalia kuvien analysointiin](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - Phi-mallien hienosäätö - - [Hienosäätötapahtumat](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Hienosäätötilanteet](./md/03.FineTuning/FineTuning_Scenarios.md) - [Hienosäätö vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - [Hienosäätö: Anna Phi-3:n tulla alan asiantuntijaksi](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Phi-3:n hienosäätö AI Toolkitilla VS Codeen](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Phi-3:n hienosäätö Azure Machine Learning -palvelulla](./md/03.FineTuning/Introduce_AzureML.md) - - [Phi-3:n hienosäätö Loralla](./md/03.FineTuning/FineTuning_Lora.md) - - [Phi-3:n hienosäätö QLoralla](./md/03.FineTuning/FineTuning_Qlora.md) - - [Phi-3:n hienosäätö Microsoft Foundryllä](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Phi-3:n hienosäätö Azure ML CLI/SDK:lla](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Hienosäätö Phi-3 AI Toolkitin avulla VS Codeen](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Hienosäätö Phi-3 Azure Machine Learning Service:n avulla](./md/03.FineTuning/Introduce_AzureML.md) + - [Hienosäätö Phi-3 Loran avulla](./md/03.FineTuning/FineTuning_Lora.md) + - [Hienosäätö Phi-3 QLoran avulla](./md/03.FineTuning/FineTuning_Qlora.md) + - [Hienosäätö Phi-3 Microsoft Foundryn avulla](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Hienosäätö Phi-3 Azure ML CLI/SDK:n avulla](./md/03.FineTuning/FineTuning_MLSDK.md) - [Hienosäätö Microsoft Olivella](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Oliven kanssa tehtävä hienosäätö -käytännön laboratorio](./md/03.FineTuning/olive-lab/readme.md) - - [Phi-3-näkömallin hienosäätö Weights and Biasen kanssa](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Phi-3:n hienosäätö Apple MLX Frameworkilla](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-näön hienosäätö (virallinen tuki)](./md/03.FineTuning/FineTuning_Vision.md) - - [Phi-3 hienosäätö Kaiton AKS:llä, Azure Containers (virallinen tuki)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Hienosäätö Microsoft Olive Hands-On Labin kanssa](./md/03.FineTuning/olive-lab/readme.md) + - [Hienosäätö Phi-3-vision Weights and Biasin kanssa](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Phi-3 hienosäätö Apple MLX -kehyksellä](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision hienosäätö (virallinen tuki)](./md/03.FineTuning/FineTuning_Vision.md) + - [Phi-3 hienosäätö Kaito AKS:llä, Azure Containers (virallinen tuki)](./md/03.FineTuning/FineTuning_Kaito.md) - [Phi-3 ja 3.5 Vision hienosäätö](https://github.com/2U1/Phi3-Vision-Finetune) -- Käytännön laboratorio - - [Edistyneiden mallien tutkiminen: LLM:t, SLM:t, paikallinen kehitys ja muuta](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP:n mahdollisuuksien avaaminen: Hienosäätö Microsoft Olivella](https://github.com/azure/Ignite_FineTuning_workshop) +- Käytännön harjoitukset + - [Huippuluokan mallien tutkiminen: LLM:t, SLM:t, paikallinen kehitys ja muuta](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP:n potentiaalin avaaminen: hienosäätö Microsoft Oliven avulla](https://github.com/azure/Ignite_FineTuning_workshop) - Akateemiset tutkimuspaperit ja julkaisut - [Textbooks Are All You Need II: phi-1.5 tekninen raportti](https://arxiv.org/abs/2309.05463) - - [Phi-3 tekninen raportti: erittäin kykenevä kielimalli paikallisesti puhelimellasi](https://arxiv.org/abs/2404.14219) + - [Phi-3 tekninen raportti: erittäin kyvykäs kielimalli paikallisesti puhelimellasi](https://arxiv.org/abs/2404.14219) - [Phi-4 tekninen raportti](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini tekninen raportti: tiiviit mutta tehokkaat monimodaaliset kielimallit Mixture-of-LoRAs-menetelmällä](https://arxiv.org/abs/2503.01743) - - [Pienten kielimallien optimointi ajoneuvojen toimintokutsuja varten](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) PHI-3 hienosäätö monivalintakysymyksiin vastaamiseen: metodologia, tulokset ja haasteet](https://arxiv.org/abs/2501.01588) - - [Phi-4-päätelmä tekninen raportti](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini-päätelmä tekninen raportti](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Phi-4-Mini tekninen raportti: Kompakti mutta tehokas monimodaalinen kielimalli Mixture-of-LoRA -tekniikalla](https://arxiv.org/abs/2503.01743) + - [Pienten kielimallien optimointi ajoneuvon sisäisiin toiminnonkutsuihin](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) PHI-3 hienosäätö monivalintakysymyksiin: menetelmät, tulokset ja haasteet](https://arxiv.org/abs/2501.01588) + - [Phi-4-päättelytekninen raportti](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-päättelytekninen raportti](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Phi-mallien käyttö +## Phi-mallien käyttäminen ### Phi Microsoft Foundryssa -Voit oppia käyttämään Microsoft Phitä ja rakentamaan E2E-ratkaisuja eri laitteillasi. Kokeaksesi Phiä itse, aloita leikkimällä malleilla ja räätälöimällä Phi tilanteisiisi käyttämällä [Microsoft Foundry Azure AI Model Catalogia](https://aka.ms/phi3-azure-ai). Lisätietoja löydät aloittamisesta [Microsoft Foundryn kanssa](/md/02.QuickStart/AzureAIFoundry_QuickStart.md). +Voit oppia käyttämään Microsoft Phi:tä ja rakentamaan E2E-ratkaisuja eri laitteillasi. Kokeillaksesi Phi:tä itse, aloita leikkimällä malleilla ja muokkaamalla Phi:tä omiin käyttötapauksiisi [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) kautta, lisätietoja on aloitusohjeissa [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md). **Leikkikenttä** -Jokaisella mallilla on oma leikkikenttä mallin testaamiseen [Azure AI Playground](https://aka.ms/try-phi3). +Jokaisella mallilla on oma testileikkipaikkansa [Azure AI Playground](https://aka.ms/try-phi3). ### Phi GitHub-malleissa -Voit oppia käyttämään Microsoft Phitä ja rakentamaan E2E-ratkaisuja eri laitteillasi. Kokeaksesi Phiä itse, aloita leikkimällä mallilla ja räätälöimällä Phi tilanteisiisi käyttämällä [GitHub Model Catalogia](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Lisätietoja löydät aloittamisesta [GitHub Model Catalogin kanssa](/md/02.QuickStart/GitHubModel_QuickStart.md). +Voit oppia käyttämään Microsoft Phi:tä ja rakentamaan E2E-ratkaisuja eri laitteillasi. Kokeillaksesi Phi:tä itse, aloita leikkimällä mallilla ja muokkaamalla Phi:tä omiin tilanteisiisi [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) kautta, lisätietoja on aloitusohjeissa [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md). **Leikkikenttä** -Jokaisella mallilla on oma [leikkikenttä mallin testaamiseen](/md/02.QuickStart/GitHubModel_QuickStart.md). +Jokaisella mallilla on oma [leikkikenttä testaukseen](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi Hugging Facessa -Mallin löydät myös [Hugging Facesta](https://huggingface.co/microsoft) +Mallin voi myös löytää [Hugging Facesta](https://huggingface.co/microsoft). **Leikkikenttä** [Hugging Chat -leikkikenttä](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) ## 🎒 Muut kurssit -Tiimimme tuottaa myös muita kursseja! Tutustu: +Tiimimme tarjoaa myös muita kursseja! Tutustu: ### LangChain @@ -264,7 +267,7 @@ Tiimimme tuottaa myös muita kursseja! Tutustu: --- -### Generatiivinen tekoäly -sarja +### Generatiivisen tekoälyn sarja [![Generatiivinen tekoäly aloittelijoille](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Generatiivinen tekoäly (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) [![Generatiivinen tekoäly (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) @@ -272,8 +275,8 @@ Tiimimme tuottaa myös muita kursseja! Tutustu: --- -### Ydinkoulutus -[![ML aloittelijoille](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +### Perustaidot +[![Koneoppiminen aloittelijoille](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Data Science aloittelijoille](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![Tekoäly aloittelijoille](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![Kyberturvallisuus aloittelijoille](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) @@ -285,33 +288,36 @@ Tiimimme tuottaa myös muita kursseja! Tutustu: ### Copilot-sarja [![Copilot tekoälyn pariohjelmointiin](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot C#/.NET:lle](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot-seikkailu](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![Copilot C#/.NET-kehittäjille](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot seikkailu](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## Vastuullinen tekoäly -Microsoft on sitoutunut auttamaan asiakkaitamme käyttämään tekoälytuotteitamme vastuullisesti, jakamaan oppimaamme ja rakentamaan luottamukseen perustuvia kumppanuuksia työkaluilla, kuten Läpinäkyvyysmuistiot ja Vaikutusten arvioinnit. Monet näistä resursseista löytyvät osoitteesta [https://aka.ms/RAI](https://aka.ms/RAI). -Microsoftin lähestymistapa vastuulliseen tekoälyyn pohjautuu tekoälyn periaatteisiimme: oikeudenmukaisuuteen, luotettavuuteen ja turvallisuuteen, yksityisyyteen ja tietoturvaan, osallisuuteen, läpinäkyvyyteen ja vastuullisuuteen. +Microsoft on sitoutunut auttamaan asiakkaitamme käyttämään tekoälytuotteitamme vastuullisesti, jakamaan oppejamme ja rakentamaan luottamukseen perustuvia kumppanuuksia läpinäkyvyysmuistiinpanojen ja vaikutusten arviointien kaltaisten työkalujen kautta. Monet näistä resursseista löytyvät osoitteesta [https://aka.ms/RAI](https://aka.ms/RAI). +Microsoftin lähestymistapa vastuulliseen tekoälyyn perustuu tekoälyn periaatteisiimme: oikeudenmukaisuus, luotettavuus ja turvallisuus, yksityisyys ja tietoturva, osallisuus, läpinäkyvyys ja vastuullisuus. + +Laajamittaiset luonnollisen kielen, kuvan ja puheen mallit – kuten tässä esimerkissä käytetyt – voivat mahdollisesti käyttäytyä epäoikeudenmukaisesti, epäluotettavasti tai loukkaavasti, mikä voi aiheuttaa haittaa. Tutustu [Azure OpenAI -palvelun läpinäkyvyysmuistiinpanoon](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ollaksesi tietoinen riskeistä ja rajoituksista. + -Suuret luonnollisen kielen, kuvan ja puheen mallit - kuten tässä esimerkissä käytetyt - voivat käyttäytyä tavoilla, jotka ovat epäreiluja, epäluotettavia tai loukkaavia, mikä voi aiheuttaa haittaa. Tutustu [Azure OpenAI -palvelun läpinäkyvyysmuistioon](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), jotta saat tietoa riskeistä ja rajoituksista. -Suositeltu lähestymistapa näiden riskien lieventämiseksi on sisällyttää arkkitehtuuriisi turvajärjestelmä, joka voi havaita ja estää haitallista käyttäytymistä. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) tarjoaa itsenäisen suojakerroksen, joka pystyy havaitsemaan haitallisen käyttäjän luoman ja tekoälyn luoman sisällön sovelluksissa ja palveluissa. Azure AI Content Safety sisältää teksti- ja kuva-API:t, jotka mahdollistavat haitallisen materiaalin havaitsemisen. Microsoft Foundryn sisällä Content Safety -palvelu mahdollistaa haitallisen sisällön tunnistamiseen liittyvän esimerkkikoodin tarkastelun, tutkimisen ja kokeilemisen eri modalityjen välillä. Seuraava [nopean käynnistyksen dokumentaatio](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) opastaa sinua tekemään pyyntöjä palveluun. +Suositeltu tapa riskien vähentämiseksi on sisällyttää arkkitehtuuriisi turvajärjestelmä, joka voi havaita ja estää haitallisen käyttäytymisen. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) tarjoaa itsenäisen suojakerroksen, joka pystyy havaitsemaan haitallisen käyttäjän tai tekoälyn luoman sisällön sovelluksissa ja palveluissa. Azure AI Content Safety sisältää teksti- ja kuva-API:t, joiden avulla voit havaita haitallista materiaalia. Microsoft Foundryn sisällä Content Safety -palvelu antaa sinun tarkastella, tutkia ja kokeilla esimerkkikoodeja haitallisen sisällön havaitsemiseksi eri muodoissa. Seuraava [aloitusdokumentaatio](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ohjaa sinut tekemään pyyntöjä palveluun. -Toinen huomioon otettava näkökulma on sovelluksen kokonaisvaltainen suorituskyky. Monimodaalisissa ja monimallipohjaisissa sovelluksissa suorituskyvyllä tarkoitetaan sitä, että järjestelmä toimii niin kuin sinä ja käyttäjäsi odotatte, mukaan lukien haitallisten tulosteiden välttäminen. On tärkeää arvioida sovelluksesi kokonaisvaltainen suorituskyky käyttämällä [Suorituskyky- ja Laatu- sekä Riski- ja Turvallisuusarvioijia](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Sinulla on myös mahdollisuus luoda ja arvioida [mukautetuilla arvioijilla](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Toinen huomioon otettava seikka on sovelluksen yleinen suorituskyky. Monimuoto- ja monimallisten sovellusten kohdalla suorituskyvyllä tarkoitetaan, että järjestelmä toimii odotetusti sinulta ja käyttäjiltäsi, mukaan lukien haitallisten tulosten välttäminen. On tärkeää arvioida koko sovelluksesi suorituskykyä käyttämällä [suorituskykyä, laatua, riskiä ja turvallisuutta arvioivia työkaluja](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Lisäksi voit luoda ja arvioida [omilla arvioijilla](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Voit arvioida tekoälysovellustasi kehitysympäristössäsi käyttämällä [Azure AI Evaluation SDK:ta](https://microsoft.github.io/promptflow/index.html). Kun käytössäsi on joko testiaineisto tai kohde, generatiivisten tekoälysovellustesi luomukset mitataan määrällisesti sisäänrakennettujen arvioijien tai valitsemiesi mukautettujen arvioijien avulla. Aloittaaksesi järjestelmäsi arvioinnin azure ai evaluation sdk:lla, voit seurata [nopean käynnistyksen opasta](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Kun suoritat arviointikierroksen, voit [visualisoida tulokset Microsoft Foundryssa](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Voit arvioida tekoälysovellustasi kehitysympäristössäsi käyttämällä [Azure AI Evaluation SDK:ta](https://microsoft.github.io/promptflow/index.html). Antamasi testidatan tai tavoitteen perusteella generatiivisen tekoälysovelluksesi tuotokset mitataan määrällisesti sisäänrakennetuilla tai valitsemillasi omilla arvioijilla. Pääset alkuun Azure AI Evaluation SDK:n kanssa arvioidaksesi järjestelmääsi seuraamalla [aloitusopasta](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Kun suoritat arviointikierroksen, voit [visualisoida tulokset Microsoft Foundryssa](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## Tavaranmerkit +## Tavaramerkit -Tämä projekti saattaa sisältää tavaramerkkejä tai logoja projekteista, tuotteista tai palveluista. Microsoftin tavaramerkkien tai logojen käyttö on sallittua ja sen on noudatettava [Microsoftin tavaramerkkien ja brändiohjeiden](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) ehtoja. Microsoftin tavaramerkkien tai logojen käyttö muokatuissa versioissa tästä projektista ei saa aiheuttaa sekaannusta tai antaa vaikutelmaa Microsoftin sponsoroinnista. Kolmansien osapuolten tavaramerkkien tai logojen käyttö on näiden osapuolten sääntöjen alaista. +Tämä projekti saattaa sisältää tavaramerkkejä tai logoja projekteille, tuotteille tai palveluille. Microsoftin tavaramerkkien tai logojen käyttö on sallittua ainoastaan noudattaen [Microsoftin tavaramerkki- ja brändiohjeita](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Microsoftin tavaramerkkien tai logojen käyttö muokatuissa versioissa tästä projektista ei saa aiheuttaa sekaannusta tai antaa vaikutelmaa Microsoftin sponsoroinnista. Kolmansien osapuolten tavaramerkkien tai logojen käyttöä säätelevät kyseisten osapuolten ohjeistukset. ## Apua saatavilla -Jos jäät jumiin tai sinulla on kysyttävää tekoälysovellusten rakentamisesta, liity: +Jos jumitut tai sinulla on kysyttävää tekoälysovellusten rakentamisesta, liity seuraaviin yhteisöihin: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Jos sinulla on tuotepalautetta tai kohtaat virheitä rakentamisen aikana, käy: +Jos sinulla on palautetta tuotteesta tai huomaat virheitä rakentamisen aikana, käy seuraavassa: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) @@ -319,5 +325,5 @@ Jos sinulla on tuotepalautetta tai kohtaat virheitä rakentamisen aikana, käy: **Vastuuvapauslauseke**: -Tämä asiakirja on käännetty käyttämällä tekoälypohjaista käännöspalvelua [Co-op Translator](https://github.com/Azure/co-op-translator). Vaikka pyrimme tarkkuuteen, huomioithan, että automaattiset käännökset saattavat sisältää virheitä tai epätarkkuuksia. Alkuperäinen asiakirja omalla kielellään on virallinen lähde. Tärkeiden tietojen osalta suositellaan ammattimaista ihmiskääntäjää. Emme ole vastuussa mahdollisista väärinymmärryksistä tai tulkinnoista, jotka johtuvat tämän käännöksen käytöstä. +Tämä asiakirja on käännetty käyttämällä tekoälypohjaista käännöspalvelua [Co-op Translator](https://github.com/Azure/co-op-translator). Vaikka pyrimme tarkkuuteen, otathan huomioon, että automaattiset käännökset saattavat sisältää virheitä tai epätarkkuuksia. Alkuperäinen asiakirja sen alkuperäiskielellä on virallinen lähde. Tärkeissä asioissa suositellaan ammattimaista ihmiskäännöstä. Emme ole vastuussa tämän käännöksen käytöstä aiheutuvista väärinymmärryksistä tai tulkinnoista. \ No newline at end of file diff --git a/translations/fi/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/fi/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..0055a655b --- /dev/null +++ b/translations/fi/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybridinen RAG Microsoft phi-4-mini & SQLite FTS5:n kanssa (Zero-Cloud SLM)\n", + "\n", + "> **Kirjoittaja:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Keskittyminen:** Pienet kielimallit (SLM), SQLite FTS5 BM25, Tiheät upotukset, Vastavuoroinen sijoituksen yhdistäminen (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivaatio: Avainsanojen hakemisen ongelma paikallisissa SLM:issä\n", + "Tavalliset RAG-arkkitehtuurit, jotka perustuvat pelkästään tiheisiin vektoriupotuksiin, eivät usein löydä täsmällisiä numeerisia tunnisteita (esim. `2,340,000 TL`, sopimuskoodit, tilinumerot). \n", + "Toisaalta harva leksikaalinen haku (BM25) ohittaa semanttiset synonyymit ja uudelleenmuotoillut kysymykset.\n", + "\n", + "Tämä ohjekirja näyttää, kuinka toteuttaa **nopea, muistissa toimiva hybridihakumoottori**, joka yhdistää:\n", + "1. **Tiheät vektorit** (kosinisimilariteetti)\n", + "2. **Harva leksikaalinen haku** (SQLite FTS5 BM25)\n", + "3. **Vastavuoroinen sijoituksen yhdistäminen (RRF, $k=60$)**\n", + "4. **Perustellun lähdeviittauksen luominen (`[1]`, `[2]`)** Microsoftin `phi-4-mini`:llä.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Kaksinkertainen SQLite-skeema (Tiiviit vektorit + Virtuaalinen FTS5 BM25 -taulu)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Näytteenotto ja suorituksen vertailuarvo\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Maadoitetun kehotteen laatiminen Microsoft phi-4-mini -mallille\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Vastuuvapauslauseke**:\nTämä asiakirja on käännetty käyttämällä tekoälypohjaista käännöspalvelua [Co-op Translator](https://github.com/Azure/co-op-translator). Vaikka pyrimme tarkkuuteen, otathan huomioon, että automaattiset käännökset saattavat sisältää virheitä tai epätarkkuuksia. Alkuperäinen asiakirja sen alkuperäiskielellä on virallinen lähde. Tärkeissä asioissa suositellaan ammattimaista ihmiskäännöstä. Emme ole vastuussa tämän käännöksen käytöstä aiheutuvista väärinymmärryksistä tai tulkinnoista.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/fr/.co-op-translator.json b/translations/fr/.co-op-translator.json index cc32b63e2..8f89106a7 100644 --- a/translations/fr/.co-op-translator.json +++ b/translations/fr/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "fr" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T07:34:39+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T16:07:30+00:00", "source_file": "README.md", "language_code": "fr" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "fr" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T16:12:07+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "fr" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:29:19+00:00", diff --git a/translations/fr/README.md b/translations/fr/README.md index a33bb760d..0a3ccffe1 100644 --- a/translations/fr/README.md +++ b/translations/fr/README.md @@ -1,11 +1,11 @@ -# Phi Cookbook : Exemples pratiques avec les modèles Phi de Microsoft +# Guide Phi : Exemples pratiques avec les modèles Phi de Microsoft -[![Ouvrir et utiliser les exemples dans GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Ouvrez et utilisez les exemples dans GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Ouvrir dans Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![Contributeurs GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![Issues GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Pull-requests GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![Problèmes GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![Demandes de tirage GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![PRs Bienvenus](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![Observateurs GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) @@ -14,29 +14,29 @@ [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi est une série de modèles IA open source développés par Microsoft. +Phi est une série de modèles d’IA open source développés par Microsoft. -Phi est actuellement le modèle de langage petit (SLM) le plus puissant et le plus rentable, avec de très bons benchmarks en multi-langues, raisonnement, génération de texte/chat, codage, images, audio et autres scénarios. +Phi est actuellement le modèle de langage petit (SLM) le plus puissant et rentable, avec d’excellents résultats dans les domaines multilingue, raisonnement, génération de texte/chat, codage, images, audio et autres scénarios. -Vous pouvez déployer Phi dans le cloud ou sur des appareils edge, et vous pouvez facilement créer des applications d’IA générative avec une puissance de calcul limitée. +Vous pouvez déployer Phi en cloud ou sur des appareils périphériques, et vous pouvez facilement créer des applications d’IA générative avec une puissance de calcul limitée. Suivez ces étapes pour commencer à utiliser ces ressources : 1. **Forkez le dépôt** : Cliquez sur [![Forks GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Clonez le dépôt** : `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Rejoignez la communauté Discord Microsoft AI et rencontrez des experts et développeurs**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Rejoignez la communauté Microsoft AI Discord et rencontrez des experts et développeurs**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![couverture](../../translated_images/fr/cover.eb18d1b9605d754b.webp) +![cover](../../translated_images/fr/cover.eb18d1b9605d754b.webp) ### 🌐 Support multilingue -#### Pris en charge via GitHub Action (Automatisé & Toujours à jour) +#### Pris en charge via GitHub Action (automatisé & toujours à jour) -[Arabe](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgare](../bg/README.md) | [Birman (Myanmar)](../my/README.md) | [Chinois (Simplifié)](../zh-CN/README.md) | [Chinois (Traditionnel, Hong Kong)](../zh-HK/README.md) | [Chinois (Traditionnel, Macau)](../zh-MO/README.md) | [Chinois (Traditionnel, Taïwan)](../zh-TW/README.md) | [Croate](../hr/README.md) | [Tchèque](../cs/README.md) | [Danois](../da/README.md) | [Néerlandais](../nl/README.md) | [Estonien](../et/README.md) | [Finnois](../fi/README.md) | [Français](./README.md) | [Allemand](../de/README.md) | [Grec](../el/README.md) | [Hébreu](../he/README.md) | [Hindi](../hi/README.md) | [Hongrois](../hu/README.md) | [Indonésien](../id/README.md) | [Italien](../it/README.md) | [Japonais](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Coréen](../ko/README.md) | [Lituanien](../lt/README.md) | [Malais](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Népalais](../ne/README.md) | [Pidgin Nigérian](../pcm/README.md) | [Norvégien](../no/README.md) | [Persan (Farsi)](../fa/README.md) | [Polonais](../pl/README.md) | [Portugais (Brésil)](../pt-BR/README.md) | [Portugais (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Roumain](../ro/README.md) | [Russe](../ru/README.md) | [Serbe (Cyrillique)](../sr/README.md) | [Slovaque](../sk/README.md) | [Slovène](../sl/README.md) | [Espagnol](../es/README.md) | [Swahili](../sw/README.md) | [Suédois](../sv/README.md) | [Tagalog (Philippin)](../tl/README.md) | [Tamoul](../ta/README.md) | [Télougou](../te/README.md) | [Thaï](../th/README.md) | [Turc](../tr/README.md) | [Ukrainien](../uk/README.md) | [Ourdou](../ur/README.md) | [Vietnamien](../vi/README.md) +[Arabe](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgare](../bg/README.md) | [Birman (Myanmar)](../my/README.md) | [Chinois (simplifié)](../zh-CN/README.md) | [Chinois (traditionnel, Hong Kong)](../zh-HK/README.md) | [Chinois (traditionnel, Macao)](../zh-MO/README.md) | [Chinois (traditionnel, Taiwan)](../zh-TW/README.md) | [Croate](../hr/README.md) | [Tchèque](../cs/README.md) | [Danois](../da/README.md) | [Néerlandais](../nl/README.md) | [Estonien](../et/README.md) | [Finnois](../fi/README.md) | [Français](./README.md) | [Allemand](../de/README.md) | [Grec](../el/README.md) | [Hébreu](../he/README.md) | [Hindi](../hi/README.md) | [Hongrois](../hu/README.md) | [Indonésien](../id/README.md) | [Italien](../it/README.md) | [Japonais](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Coréen](../ko/README.md) | [Lituanien](../lt/README.md) | [Malais](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Népalais](../ne/README.md) | [Pidgin nigérian](../pcm/README.md) | [Norvégien](../no/README.md) | [Persan (Farsi)](../fa/README.md) | [Polonais](../pl/README.md) | [Portugais (Brésil)](../pt-BR/README.md) | [Portugais (Portugal)](../pt-PT/README.md) | [Pendjabi (Gurmukhi)](../pa/README.md) | [Roumain](../ro/README.md) | [Russe](../ru/README.md) | [Serbe (Cyrillique)](../sr/README.md) | [Slovaque](../sk/README.md) | [Slovène](../sl/README.md) | [Espagnol](../es/README.md) | [Swahili](../sw/README.md) | [Suédois](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamoul](../ta/README.md) | [Télougou](../te/README.md) | [Thaï](../th/README.md) | [Turc](../tr/README.md) | [Ukrainien](../uk/README.md) | [Ourdou](../ur/README.md) | [Vietnamien](../vi/README.md) > **Préférez-vous cloner localement ?** > -> Ce dépôt inclut plus de 50 traductions ce qui augmente significativement la taille du téléchargement. Pour cloner sans les traductions, utilisez le sparse checkout : +> Ce dépôt comprend plus de 50 traductions linguistiques qui augmentent considérablement la taille du téléchargement. Pour cloner sans traductions, utilisez le checkout éparse : > > **Bash / macOS / Linux :** > ```bash @@ -52,24 +52,24 @@ Suivez ces étapes pour commencer à utiliser ces ressources : > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Cela vous donnera tout ce dont vous avez besoin pour terminer le cours avec un téléchargement beaucoup plus rapide. +> Cela vous donne tout ce dont vous avez besoin pour compléter le cours avec un téléchargement beaucoup plus rapide. ## Table des matières - Introduction - [Bienvenue dans la famille Phi](./md/01.Introduction/01/01.PhiFamily.md) - - [Configuration de votre environnement](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Configurer votre environnement](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Comprendre les technologies clés](./md/01.Introduction/01/01.Understandingtech.md) - [Sécurité IA pour les modèles Phi](./md/01.Introduction/01/01.AISafety.md) - [Support matériel Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Modèles Phi & Disponibilité sur les plateformes](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Modèles Phi & disponibilité sur les plateformes](./md/01.Introduction/01/01.Edgeandcloud.md) - [Utilisation de Guidance-ai et Phi](./md/01.Introduction/01/01.Guidance.md) - [Modèles GitHub Marketplace](https://github.com/marketplace/models) - - [Catalogue de modèles Azure AI](https://ai.azure.com) + - [Catalogue des modèles Azure AI](https://ai.azure.com) - Inférence Phi dans différents environnements - - [Hugging Face](./md/01.Introduction/02/01.HF.md) + - [Hugging face](./md/01.Introduction/02/01.HF.md) - [Modèles GitHub](./md/01.Introduction/02/02.GitHubModel.md) - [Catalogue Microsoft Foundry Model](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) @@ -77,123 +77,125 @@ Suivez ces étapes pour commencer à utiliser ces ressources : - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Inférence Famille Phi +- Inférence Phi Family - [Inférence Phi sur iOS](./md/01.Introduction/03/iOS_Inference.md) - [Inférence Phi sur Android](./md/01.Introduction/03/Android_Inference.md) - [Inférence Phi sur Jetson](./md/01.Introduction/03/Jetson_Inference.md) - [Inférence Phi sur AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Inférence Phi avec le framework Apple MLX](./md/01.Introduction/03/MLX_Inference.md) + - [Inférence Phi avec Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - [Inférence Phi sur serveur local](./md/01.Introduction/03/Local_Server_Inference.md) - [Inférence Phi sur serveur distant avec AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - [Inférence Phi avec Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inférence Phi–Vision en local](./md/01.Introduction/03/Vision_Inference.md) + - [Inférence Phi--Vision en local](./md/01.Introduction/03/Vision_Inference.md) - [Inférence Phi avec Kaito AKS, Azure Containers (support officiel)](./md/01.Introduction/03/Kaito_Inference.md) -- [Quantification Famille Phi](./md/01.Introduction/04/QuantifyingPhi.md) +- [Quantification Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) - [Quantification Phi-3.5 / 4 avec llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Quantification Phi-3.5 / 4 avec extensions IA générative pour onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Quantification Phi-3.5 / 4 avec Extensions IA générative pour onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [Quantification Phi-3.5 / 4 avec Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Quantification Phi-3.5 / 4 avec le framework Apple MLX](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Quantification Phi-3.5 / 4 avec Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Évaluation Phi - [IA Responsable](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry pour l’évaluation](./md/01.Introduction/05/AIFoundry.md) - - [Utiliser Promptflow pour l’évaluation](./md/01.Introduction/05/Promptflow.md) + - [Utilisation de Promptflow pour l’évaluation](./md/01.Introduction/05/Promptflow.md) - RAG avec Azure AI Search - [Comment utiliser Phi-4-mini et Phi-4-multimodal (RAG) avec Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Hybride local zéro-cloud RAG avec SQLite FTS5 et phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Échantillons de développement d’applications Phi +- Exemples de développement d’applications Phi - Applications texte & chat - - Échantillons Phi-4 - - [📓] [Discussion avec modèle Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Discussion avec modèle Phi-4 ONNX local .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - Exemples Phi-4 + - [📓] [Discussion avec le modèle Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Discussion avec le modèle Phi-4 ONNX local .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - [Application console chat .NET avec Phi-4 ONNX utilisant Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Échantillons Phi-3 / 3.5 - [Chatbot local dans le navigateur utilisant Phi3, ONNX Runtime Web et WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Chat OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [Multi Modèle - Phi-3-mini interactif et OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Création d'un wrapper et utilisation de Phi-3 avec MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Optimisation du modèle - Comment optimiser le modèle Phi-3-min pour ONNX Runtime Web avec Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [MLFlow - Création d’un wrapper et utilisation de Phi-3 avec MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Optimisation de modèle - Comment optimiser le modèle Phi-3-min pour ONNX Runtime Web avec Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [App WinUI3 avec Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Exemple d'application de prise de notes multi-modèle WinUI3 alimentée par IA](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Affinage et intégration de modèles Phi-3 personnalisés avec Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Affinage et intégration de modèles Phi-3 personnalisés avec Prompt flow dans Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Évaluation du modèle Phi-3 / Phi-3.5 affiné dans Microsoft Foundry en se concentrant sur les principes d'IA responsable de Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + -[Exemple d’application de notes alimentée par IA multi-modèle WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Affiner et intégrer des modèles Phi-3 personnalisés avec Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Affiner et intégrer des modèles Phi-3 personnalisés avec Prompt flow dans Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Évaluer le modèle Phi-3 / Phi-3.5 affiné dans Microsoft Foundry en se concentrant sur les principes d’IA responsable de Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Exemple de prédiction linguistique Phi-3.5-mini-instruct (Chinois/Anglais)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Chatbot RAG Phi-3.5-Instruct WebGPU](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Utilisation du GPU Windows pour créer une solution Prompt flow avec Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Utilisation de Microsoft Phi-3.5 tflite pour créer une application Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Exemple Q&R .NET utilisant le modèle ONNX Phi-3 local via Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Utiliser le GPU Windows pour créer une solution Prompt flow avec Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Utiliser Microsoft Phi-3.5 tflite pour créer une application Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Exemple Q&R .NET utilisant modèle local ONNX Phi-3 avec Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - [Application console chat .NET avec Semantic Kernel et Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Exemples de code basés sur le SDK Azure AI Inference - - Exemples Phi-4 - - [📓] [Générer le code de projet en utilisant Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Exemples Phi-3 / 3.5 - - [Construisez votre propre chat GitHub Copilot Visual Studio Code avec Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Créez votre propre agent Copilot Chat Visual Studio Code avec Phi-3.5 par GitHub Models](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - Exemples basés sur le code Azure AI Inference SDK + - Échantillons Phi-4 + - [📓] [Générer le code projet avec Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Échantillons Phi-3 / 3.5 + - [Construisez votre propre Visual Studio Code GitHub Copilot Chat avec la famille Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Créez votre propre agent Chat Copilot Visual Studio Code avec Phi-3.5 par les modèles GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Exemples de raisonnement avancé - - Exemples Phi-4 + - Échantillons Phi-4 - [📓] [Exemples Phi-4-mini-reasoning ou Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Affinage de Phi-4-mini-reasoning avec Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Affinage de Phi-4-mini-reasoning avec Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning avec GitHub Models](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning avec Microsoft Foundry Models](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [📓] [Affinage Phi-4-mini-reasoning avec Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Affinage Phi-4-mini-reasoning avec Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-reasoning avec modèles GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-reasoning avec modèles Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Démos - [Démos Phi-4-mini hébergées sur Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - [Démos Phi-4-multimodal hébergées sur Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Exemples Vision - - Exemples Phi-4 - - [📓] [Utiliser Phi-4-multimodal pour lire des images et générer du code](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Exemples Phi-3 / 3.5 - - [📓][Phi-3-vision-Image texte à texte](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - Échantillons Vision + - Échantillons Phi-4 + - [📓] [Utiliser Phi-4-multimodal pour lire des images et générer du code](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Échantillons Phi-3 / 3.5 + - [📓][Phi-3-vision-Texte image vers texte](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DÉMO : Recycling Phi-3](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Assistant visuel en langage - avec Phi3-Vision et OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [DÉMO : Phi-3 Recycling](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - Assistant de langage visuel - avec Phi3-Vision et OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Exemple Phi-3.5 Vision multi-frames ou multi-images](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision Modèle ONNX local utilisant Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Modèle ONNX local Phi-3 Vision à base de menu utilisant Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Échantillon Phi-3.5 Vision multi-frame ou multi-image](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Modèle local Phi-3 Vision ONNX utilisant Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Modèle local Phi-3 Vision ONNX avec menu utilisant Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - Exemples Raisonnement-Vision - - Phi-4-Raisonnement-Vision-15B - - [📓] [Utiliser Phi-4-Raisonnement-Vision-15B pour détecter le jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - Échantillons Raisonnement-Vision + - Phi-4-Raisonnement-Vision-15B + - [📓] [Utiliser Phi-4-Raisonnement-Vision-15B pour détecter le passage piéton intempestif](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [Utiliser Phi-4-Raisonnement-Vision-15B pour les mathématiques](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Utiliser Phi-4-Raisonnement-Vision-15B pour détecter l'interface utilisateur](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [Utiliser Phi-4-Raisonnement-Vision-15B pour détecter l’interface utilisateur](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - Exemples Mathématiques - - Exemples Phi-4-Mini-Flash-Reasoning-Instruct [Démonstration Math avec Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Échantillons Mathématiques + - Échantillons Phi-4-Mini-Flash-Raisonnement-Instruct [Démo Math avec Phi-4-Mini-Flash-Raisonnement-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - Exemples Audio - - Exemples Phi-4 + - Échantillons Audio + - Échantillons Phi-4 - [📓] [Extraction de transcriptions audio avec Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Exemple audio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Exemple de traduction vocale Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - [Application console .NET utilisant Phi-4-multimodal Audio pour analyser un fichier audio et générer une transcription](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - Exemples MOE - - Exemples Phi-3 / 3.5 - - [📓] [Exemples Phi-3.5 Mixture of Experts Models (MoEs) sur les réseaux sociaux](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Création d'un pipeline Retrieval-Augmented Generation (RAG) avec NVIDIA NIM Phi-3 MOE, Azure AI Search et LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - Échantillons MOE + - Échantillons Phi-3 / 3.5 + - [📓] [Modèles Mixture of Experts Phi-3.5 (MoEs) Exemple réseaux sociaux](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Création d’un pipeline de génération augmentée par récupération (RAG) avec NVIDIA NIM Phi-3 MOE, Azure AI Search, et LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Exemples d'appel de fonctions - - Exemples Phi-4 🆕 - - [📓] [Utilisation de l'appel de fonction avec Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Utilisation de l'appel de fonction pour créer des agents multiples avec Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Utilisation de l'appel de fonction avec Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Utilisation de l'appel de fonction avec ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Exemples de mixage multimodal - - Exemples Phi-4 🆕 - - [📓] [Utilisation de Phi-4-multimodal en tant que journaliste technologique](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - Échantillons d’appel de fonction + - Échantillons Phi-4 🆕 + - [📓] [Utilisation de l’appel de fonction avec Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Utilisation de l’appel de fonction pour créer des agents multiples avec Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Utilisation de l’appel de fonction avec Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Utilisation de l’appel de fonction avec ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Échantillons mélange multimodal + - Échantillons Phi-4 🆕 + - [📓] [Utiliser Phi-4-multimodal en tant que journaliste technologique](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [Application console .NET utilisant Phi-4-multimodal pour analyser des images](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Affinage des exemples Phi - - [Scénarios d'affinage](./md/03.FineTuning/FineTuning_Scenarios.md) +- Affinage des échantillons Phi + - [Scénarios d’affinage](./md/03.FineTuning/FineTuning_Scenarios.md) - [Affinage vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Affinage : Faire de Phi-3 un expert industriel](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Affinage : faire de Phi-3 un expert industriel](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - [Affinage de Phi-3 avec AI Toolkit pour VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [Affinage de Phi-3 avec Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - [Affinage de Phi-3 avec Lora](./md/03.FineTuning/FineTuning_Lora.md) @@ -201,53 +203,54 @@ Suivez ces étapes pour commencer à utiliser ces ressources : - [Affinage de Phi-3 avec Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - [Affinage de Phi-3 avec Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - [Affinage avec Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Atelier pratique d'affinage avec Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [Affinage Phi-3-vision avec Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Affinage de Phi-3 avec Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Affinage Phi-3-vision (support officiel)](./md/03.FineTuning/FineTuning_Vision.md) - - [Affinage de Phi-3 avec Kaito AKS, Conteneurs Azure (support officiel)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Affinage de Phi-3 et 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + - [Affinage avec Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) + - [Affinage de Phi-3-vision avec Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Affiner Phi-3 avec Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Affiner Phi-3-vision (support officiel)](./md/03.FineTuning/FineTuning_Vision.md) + - [Affiner Phi-3 avec Kaito AKS, Azure Containers (support officiel)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Affiner Phi-3 et 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Atelier pratique - - [Exploration des modèles de pointe : LLMs, SLMs, développement local et plus](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Explorer les modèles de pointe : LLMs, SLMs, développement local et plus](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - [Libérer le potentiel du NLP : Affinage avec Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Articles de recherche académiques et publications +- Articles de recherche académique et publications - [Les manuels sont tout ce dont vous avez besoin II : rapport technique phi-1.5](https://arxiv.org/abs/2309.05463) - - [Rapport technique Phi-3 : Un modèle de langage très performant localement sur votre téléphone](https://arxiv.org/abs/2404.14219) + - [Rapport technique Phi-3 : un modèle de langage très performant localement sur votre téléphone](https://arxiv.org/abs/2404.14219) - [Rapport technique Phi-4](https://arxiv.org/abs/2412.08905) - [Rapport technique Phi-4-Mini : Modèles de langage multimodaux compacts mais puissants via Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Optimisation des petits modèles de langage pour l’appel de fonctions embarquées](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Affinage de PHI-3 pour la réponse aux questions à choix multiple : méthodologie, résultats et défis](https://arxiv.org/abs/2501.01588) - - [Rapport technique Phi-4-raisonnement](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Optimisation des petits modèles de langage pour l’appel de fonctions dans les véhicules](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Affinage de PHI-3 pour la réponse aux questions à choix multiples : méthodologie, résultats et défis](https://arxiv.org/abs/2501.01588) + - [Rapport technique Phi-4 raisonnement](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Rapport technique Phi-4-mini-raisonnement](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Utilisation des modèles Phi ### Phi sur Microsoft Foundry -Vous pouvez apprendre à utiliser Microsoft Phi et comment construire des solutions E2E sur vos différents appareils matériels. Pour expérimenter Phi vous-même, commencez par vous amuser avec les modèles et personnaliser Phi pour vos scénarios en utilisant le [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) ; vous pouvez en savoir plus dans la section Démarrage avec [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md). +Vous pouvez apprendre comment utiliser Microsoft Phi et comment construire des solutions de bout en bout sur différents dispositifs matériels. Pour découvrir Phi par vous-même, commencez par expérimenter avec les modèles et personnaliser Phi pour vos scénarios en utilisant le [Catalogue Microsoft Foundry Azure AI Model](https://aka.ms/phi3-azure-ai) ; vous pouvez en savoir plus dans la rubrique Démarrage rapide avec [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Playground** -Chaque modèle dispose d’un playground dédié pour tester le modèle [Azure AI Playground](https://aka.ms/try-phi3). +**Espace de test** +Chaque modèle dispose d’un espace de test dédié pour évaluer le modèle [Azure AI Playground](https://aka.ms/try-phi3). ### Phi sur les modèles GitHub -Vous pouvez apprendre à utiliser Microsoft Phi et comment construire des solutions E2E sur vos différents appareils matériels. Pour expérimenter Phi vous-même, commencez par vous amuser avec le modèle et personnaliser Phi pour vos scénarios en utilisant le [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ; vous pouvez en savoir plus dans la section Démarrage avec [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md). +Vous pouvez apprendre comment utiliser Microsoft Phi et comment construire des solutions de bout en bout sur différents dispositifs matériels. Pour découvrir Phi par vous-même, commencez par expérimenter avec le modèle et personnaliser Phi pour vos scénarios en utilisant le [Catalogue des modèles GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ; vous pouvez en savoir plus dans la rubrique Démarrage rapide avec [Catalogue des modèles GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Playground** -Chaque modèle dispose d’un [playground dédié pour tester le modèle](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Espace de test** +Chaque modèle dispose d’un [espace de test dédié pour évaluer le modèle](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi sur Hugging Face -Vous pouvez aussi trouver le modèle sur [Hugging Face](https://huggingface.co/microsoft). +Vous pouvez également trouver le modèle sur [Hugging Face](https://huggingface.co/microsoft) -**Playground** -[Playground Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Espace de test** + [Espace de test Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Autres cours + ## 🎒 Autres cours -Notre équipe produit d’autres cours ! Consultez : +Notre équipe produit d’autres cours ! Découvrez : ### LangChain @@ -264,17 +267,17 @@ Notre équipe produit d’autres cours ! Consultez : --- -### Série Intelligence Artificielle Générative -[![IA générative pour débutants](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![IA générative (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![IA générative (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![IA générative (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### Série IA Générative +[![IA Générative pour débutants](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![IA Générative (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![IA Générative (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![IA Générative (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- -### Apprentissage de base +### Apprentissage fondamental [![ML pour débutants](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Sciences des données pour débutants](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![Science des données pour débutants](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![IA pour débutants](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![Cybersécurité pour débutants](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Développement web pour débutants](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) @@ -289,36 +292,38 @@ Notre équipe produit d’autres cours ! Consultez : [![Aventure Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## IA responsable +## IA responsable + +Microsoft s’engage à aider ses clients à utiliser ses produits d’IA de manière responsable, à partager nos enseignements et à construire des partenariats basés sur la confiance grâce à des outils tels que les Notes de transparence et les évaluations d’impact. Beaucoup de ces ressources sont disponibles sur [https://aka.ms/RAI](https://aka.ms/RAI). +L’approche de Microsoft en matière d’IA responsable repose sur nos principes d’IA : équité, fiabilité et sécurité, confidentialité et sécurité, inclusion, transparence et responsabilité. + +Les modèles à grande échelle pour le langage naturel, les images et la parole - comme ceux utilisés dans cet exemple - peuvent potentiellement se comporter de manière injuste, peu fiable ou offensante, causant ainsi des préjudices. Veuillez consulter la [note de transparence du service Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) pour vous informer des risques et des limites. -Microsoft s’engage à aider ses clients à utiliser nos produits d’IA de manière responsable, à partager nos apprentissages et à construire des partenariats basés sur la confiance grâce à des outils comme les Notes de transparence et les évaluations d’impact. Beaucoup de ces ressources sont disponibles sur [https://aka.ms/RAI](https://aka.ms/RAI). -L’approche de Microsoft pour une IA responsable repose sur nos principes d’IA : équité, fiabilité et sécurité, confidentialité et sécurité, inclusion, transparence et responsabilité. -Les modèles à grande échelle de langage naturel, d’image et de parole - comme ceux utilisés dans cet exemple - peuvent potentiellement se comporter de manière injuste, peu fiable ou offensante, causant ainsi des préjudices. Veuillez consulter la [note de transparence du service Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) pour être informé des risques et des limites. -L'approche recommandée pour atténuer ces risques est d'inclure un système de sécurité dans votre architecture qui peut détecter et prévenir les comportements nuisibles. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) fournit une couche de protection indépendante, capable de détecter le contenu nuisible généré par les utilisateurs et par l'IA dans les applications et services. Azure AI Content Safety inclut des API de texte et d'image qui vous permettent de détecter le contenu nuisible. Au sein de Microsoft Foundry, le service Content Safety vous permet de visualiser, d'explorer et d'essayer du code d'exemple pour détecter le contenu nuisible à travers différentes modalités. La [documentation de démarrage rapide](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) suivante vous guide dans la réalisation de requêtes vers le service. +L'approche recommandée pour atténuer ces risques est d'inclure un système de sécurité dans votre architecture capable de détecter et de prévenir les comportements nuisibles. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) fournit une couche de protection indépendante, capable de détecter les contenus nuisibles générés par les utilisateurs et par l'IA dans les applications et services. Azure AI Content Safety comprend des API de texte et d'images qui vous permettent de détecter les contenus nuisibles. Au sein de Microsoft Foundry, le service Content Safety vous permet de visualiser, d'explorer et d'essayer des exemples de code pour détecter les contenus nuisibles à travers différentes modalités. La [documentation de démarrage rapide](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) suivante vous guide pour effectuer des requêtes vers le service. -Un autre aspect à prendre en compte est la performance globale de l'application. Avec les applications multi-modales et multi-modèles, la performance signifie que le système fonctionne comme vous et vos utilisateurs l’attendez, y compris en ne générant pas de sorties nuisibles. Il est important d'évaluer la performance de votre application globale en utilisant les [évaluateurs de Performance et Qualité ainsi que Risque et Sécurité](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Vous avez également la possibilité de créer et d'évaluer avec des [évaluateurs personnalisés](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Un autre aspect à prendre en compte est la performance globale de l'application. Avec des applications multi-modales et multi-modèles, nous considérons que la performance signifie que le système fonctionne comme vous et vos utilisateurs l'attendez, y compris ne pas générer de résultats nuisibles. Il est important d'évaluer la performance de votre application globale en utilisant les [évaluateurs de Performance, Qualité, Risque et Sécurité](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Vous avez également la possibilité de créer et d'évaluer avec des [évaluateurs personnalisés](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Vous pouvez évaluer votre application d'IA dans votre environnement de développement en utilisant le [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Que vous disposiez d'un jeu de données de test ou d’une cible, les générations de votre application d’IA générative sont mesurées quantitativement avec des évaluateurs intégrés ou des évaluateurs personnalisés de votre choix. Pour commencer avec l'azure ai evaluation sdk afin d'évaluer votre système, vous pouvez suivre le [guide de démarrage rapide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Une fois que vous avez exécuté une évaluation, vous pouvez [visualiser les résultats dans Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Vous pouvez évaluer votre application IA dans votre environnement de développement en utilisant le [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Donné un jeu de données de test ou un objectif, vos générations d'application IA générative sont mesurées quantitativement avec des évaluateurs intégrés ou des évaluateurs personnalisés de votre choix. Pour commencer avec le sdk d'évaluation azure ai pour évaluer votre système, vous pouvez suivre le [guide de démarrage rapide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Une fois que vous exécutez une session d'évaluation, vous pouvez [visualiser les résultats dans Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Marques déposées -Ce projet peut contenir des marques déposées ou des logos de projets, produits ou services. L'utilisation autorisée des marques déposées ou logos Microsoft est soumise à et doit respecter les [Directives sur les marques et la marque de Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -L'utilisation des marques déposées ou logos Microsoft dans des versions modifiées de ce projet ne doit pas entraîner de confusion ni impliquer un parrainage de Microsoft. Toute utilisation de marques déposées ou logos de tiers est soumise aux politiques de ces tiers. +Ce projet peut contenir des marques déposées ou logos de projets, produits ou services. L'utilisation autorisée des marques déposées ou logos Microsoft est soumise aux [Directives sur les marques et la marque](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) de Microsoft. +L'utilisation des marques ou logos Microsoft dans des versions modifiées de ce projet ne doit pas causer de confusion ni impliquer un parrainage de Microsoft. Toute utilisation des marques ou logos de tiers est soumise aux politiques de ces tiers. -## Obtenir de l’aide +## Obtenir de l'aide -Si vous êtes bloqué ou si vous avez des questions sur la création d'applications IA, rejoignez : +Si vous êtes bloqué ou avez des questions sur la création d'applications IA, rejoignez : [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Si vous avez des retours sur le produit ou des erreurs lors de la création, consultez : +Si vous avez des retours produits ou des erreurs lors de la construction, rendez-vous sur : [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Avertissement** : -Ce document a été traduit à l'aide du service de traduction automatique [Co-op Translator](https://github.com/Azure/co-op-translator). Bien que nous nous efforcions d'assurer l'exactitude, veuillez noter que les traductions automatiques peuvent contenir des erreurs ou des inexactitudes. Le document original dans sa langue native doit être considéré comme la source faisant foi. Pour les informations critiques, une traduction professionnelle effectuée par un humain est recommandée. Nous ne saurions être tenus responsables des malentendus ou des erreurs d'interprétation résultant de l'utilisation de cette traduction. +**Avertissement** : +Ce document a été traduit à l'aide du service de traduction automatique [Co-op Translator](https://github.com/Azure/co-op-translator). Bien que nous nous efforçions d'assurer l'exactitude, veuillez noter que les traductions automatisées peuvent contenir des erreurs ou des inexactitudes. Le document original dans sa langue native doit être considéré comme la source faisant autorité. Pour les informations critiques, il est recommandé de recourir à une traduction professionnelle réalisée par un humain. Nous ne saurions être tenus responsables des malentendus ou erreurs d'interprétation découlant de l'utilisation de cette traduction. \ No newline at end of file diff --git a/translations/fr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/fr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..717e076a0 --- /dev/null +++ b/translations/fr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 RAG Hybride avec Microsoft phi-4-mini & SQLite FTS5 (SLM sans Cloud)\n", + "\n", + "> **Auteur :** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Sujet :** Petits Modèles de Langage (SLM), SQLite FTS5 BM25, Embeddings Denses, Fusion par Rang Réciproque (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivation : Le Dilemme du Rappel par Mot-Clé dans les SLM Locaux\n", + "Les architectures RAG standard s'appuyant uniquement sur des embeddings vectoriels denses échouent souvent à récupérer des jetons numériques exacts (par ex., `2 340 000 TL`, codes de contrat, numéros de compte). \n", + "En revanche, la recherche lexicale parcimonieuse (BM25) manque les synonymes sémantiques et les questions paraphrasées.\n", + "\n", + "Ce guide pratique montre comment implémenter un **moteur de récupération hybride ultra-rapide en mémoire** combinant :\n", + "1. **Vecteurs Denses** (Similarité Cosinus)\n", + "2. **Recherche Lexicale Parcimonieuse** (SQLite FTS5 BM25)\n", + "3. **Fusion par Rang Réciproque (RRF, $k=60$)**\n", + "4. **Génération de Citations Fondées (`[1]`, `[2]`)** avec Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Schéma SQLite double (Vecteurs denses + Table virtuelle FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Benchmark d'ingestion et d'exécution d'échantillons\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Formulation d’invite ancrée pour Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Avertissement** :\nCe document a été traduit à l'aide du service de traduction automatique [Co-op Translator](https://github.com/Azure/co-op-translator). Bien que nous nous efforçions d'assurer l'exactitude, veuillez noter que les traductions automatisées peuvent contenir des erreurs ou des inexactitudes. Le document original dans sa langue native doit être considéré comme la source faisant autorité. Pour les informations critiques, il est recommandé de recourir à une traduction professionnelle réalisée par un humain. Nous ne saurions être tenus responsables des malentendus ou erreurs d'interprétation découlant de l'utilisation de cette traduction.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/he/.co-op-translator.json b/translations/he/.co-op-translator.json index e7bd46e75..69b9c3c15 100644 --- a/translations/he/.co-op-translator.json +++ b/translations/he/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "he" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:06:28+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:05:41+00:00", "source_file": "README.md", "language_code": "he" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "he" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:08:30+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "he" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:27+00:00", diff --git a/translations/he/README.md b/translations/he/README.md index 95d707280..33471a94c 100644 --- a/translations/he/README.md +++ b/translations/he/README.md @@ -1,42 +1,42 @@ -# ספר מתכונים לפאי: דוגמאות מעשיות עם דגמי הפאי של מיקרוסופט +# ספר המתכונים Phi: דוגמאות מעשיות עם דגמי Phi של מיקרוסופט -[![פתח והשתמש בדוגמאות ב-GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![פתח ב-Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![פתח והשתמש בדוגמאות בקוד ספייס של GitHub](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![פתח במיכלי פיתוח](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![תורמים ב-GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![נושאים ב-GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![תורמים ל-GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![בעיות ב-GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) [![בקשות משיכה ב-GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![ברוכים הבאים לבקשות משיכה](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![עוקבים ב-GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![פורקים ב-GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![צופים ב-GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![מפיצלים ב-GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![כוכבים ב-GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) -[![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) +[![Discord של Microsoft Foundry](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -פאי היא סדרה של דגמי בינה מלאכותית בקוד פתוח שפותחו על ידי מיקרוסופט. +Phi היא סדרת דגמי בינה מלאכותית בקוד פתוח שפותחה על ידי מיקרוסופט. -פאי היא כיום מודל שפה קטן (SLM) החזק והיעיל ביותר מבחינת עלות, עם מדדים טובים מאוד במגוון שפות, חשיבה, יצירת טקסט/שיחה, קידוד, תמונות, אודיו וסצנריואים נוספים. +Phi היא כיום דגם שפה קטן (SLM) החזק והיעיל ביותר מבחינת עלות, עם ביצועים מאוד טובים במבחנים בשפות מרובות, היסקות, יצירת טקסט/צ'אט, קידוד, תמונות, שמע ותסריטים נוספים. -ניתן לפרוס את פאי בענן או על מכשירי קצה, וניתן לבנות בקלות יישומי בינה מלאכותית מפורטים עם כוח חישובי מוגבל. +ניתן לפרוס את Phi בענן או במכשירי קצה, וניתן בקלות לבנות יישומי בינה מלאכותית יצירתית עם יכולת חישוב מוגבלת. -עקבו אחרי השלבים האלה כדי להתחיל להשתמש במשאבים הללו: -1. **העמיקו את המאגר**: לחצו על [![פורקים ב-GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +עקבו אחר הצעדים הללו כדי להתחיל להשתמש במשאבים אלו: +1. **פצלו את המאגר**: לחצו על [![מפיצלים ב-GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **שכפלו את המאגר**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**הצטרפו לקהילת ה-Discord של AI מיקרוסופט ופגשו מומחים ומפתחים נוספים**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**הצטרפו לקהילת Discord של מיקרוסופט AI ופגשו מומחים ומפתחים נוספים**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/he/cover.eb18d1b9605d754b.webp) ### 🌐 תמיכה בריבוי שפות -#### נתמך באמצעות GitHub Action (אוטומטי ותמיד מעודכן) +#### נתמך באמצעות פעולה ב-GitHub (מאוד עדכני ואוטומטי) -[ערבית](../ar/README.md) | [בנגלית](../bn/README.md) | [בולגרית](../bg/README.md) | [בורמזית (מיאנמר)](../my/README.md) | [סינית (מפושטת)](../zh-CN/README.md) | [סינית (מסורתית, הונג קונג)](../zh-HK/README.md) | [סינית (מסורתית, מקאו)](../zh-MO/README.md) | [סינית (מסורתית, טאיוואן)](../zh-TW/README.md) | [קרואטית](../hr/README.md) | [צ'כית](../cs/README.md) | [דנית](../da/README.md) | [הולנדית](../nl/README.md) | [אסטונית](../et/README.md) | [פינית](../fi/README.md) | [צרפתית](../fr/README.md) | [גרמנית](../de/README.md) | [יוונית](../el/README.md) | [עברית](./README.md) | [הינדי](../hi/README.md) | [הונגרית](../hu/README.md) | [אינדונזית](../id/README.md) | [איטלקית](../it/README.md) | [יפנית](../ja/README.md) | [קנדה](../kn/README.md) | [חמרית](../km/README.md) | [קוריאנית](../ko/README.md) | [ליטאית](../lt/README.md) | [מלאית](../ms/README.md) | [מאליאלאם](../ml/README.md) | [מרתית](../mr/README.md) | [נפאלית](../ne/README.md) | [פידג'ן ניגרי](../pcm/README.md) | [נורווגית](../no/README.md) | [פרסית (פרסית)](../fa/README.md) | [פולנית](../pl/README.md) | [פורטוגזית (ברזיל)](../pt-BR/README.md) | [פורטוגזית (פורטוגל)](../pt-PT/README.md) | [פונז'בית (ג'רמוכי)](../pa/README.md) | [רומנית](../ro/README.md) | [רוסית](../ru/README.md) | [סרבית (קירילית)](../sr/README.md) | [סלובקית](../sk/README.md) | [סלובנית](../sl/README.md) | [ספרדית](../es/README.md) | [סווהילי](../sw/README.md) | [שוודית](../sv/README.md) | [טגלוג (פיליפינית)](../tl/README.md) | [טמילית](../ta/README.md) | [טלוגו](../te/README.md) | [תאית](../th/README.md) | [טורקית](../tr/README.md) | [אוקראינית](../uk/README.md) | [אורדו](../ur/README.md) | [וייטנאמית](../vi/README.md) +[ערבית](../ar/README.md) | [בנגלית](../bn/README.md) | [בולגרית](../bg/README.md) | [בורמזית (מיאנמר)](../my/README.md) | [סינית (מפושטת)](../zh-CN/README.md) | [סינית (מסורתית, הונג קונג)](../zh-HK/README.md) | [סינית (מסורתית, מאקאו)](../zh-MO/README.md) | [סינית (מסורתית, טייוואן)](../zh-TW/README.md) | [קרואטית](../hr/README.md) | [צ'כית](../cs/README.md) | [דנית](../da/README.md) | [הולנדית](../nl/README.md) | [אסטונית](../et/README.md) | [פינית](../fi/README.md) | [צרפתית](../fr/README.md) | [גרמנית](../de/README.md) | [יוונית](../el/README.md) | [עברית](./README.md) | [הודית (הינדית)](../hi/README.md) | [הונגרית](../hu/README.md) | [אינדונזית](../id/README.md) | [איטלקית](../it/README.md) | [יפנית](../ja/README.md) | [קנדה](../kn/README.md) | [חמרית](../km/README.md) | [קוריאנית](../ko/README.md) | [ליטאית](../lt/README.md) | [מלאית](../ms/README.md) | [מלאיאלאם](../ml/README.md) | [מרטהית](../mr/README.md) | [נפאלית](../ne/README.md) | [פידג'ין ניגרי](../pcm/README.md) | [נורווגית](../no/README.md) | [פרסית (פארסי)](../fa/README.md) | [פולנית](../pl/README.md) | [פורטוגזית (ברזיל)](../pt-BR/README.md) | [פורטוגזית (פורטוגל)](../pt-PT/README.md) | [פונג'בית (ג'רמוקי)](../pa/README.md) | [רומנית](../ro/README.md) | [רוסית](../ru/README.md) | [סרבית (קירילית)](../sr/README.md) | [סלובקית](../sk/README.md) | [סלובנית](../sl/README.md) | [ספרדית](../es/README.md) | [סוואהילי](../sw/README.md) | [שוודית](../sv/README.md) | [טגלוג (פיליפינית)](../tl/README.md) | [טמילית](../ta/README.md) | [טלאוגו](../te/README.md) | [תאית](../th/README.md) | [טורקית](../tr/README.md) | [אוקראינית](../uk/README.md) | [אורדו](../ur/README.md) | [וייטנאמית](../vi/README.md) > **מעדיפים לשכפל מקומית?** > -> מאגר זה כולל למעלה מ-50 תרגומים לשפות שונות שמגדילים משמעותית את גודל ההורדה. כדי לשכפל ללא תרגומים, השתמשו ב-checkout חסר: +> מאגר זה כולל למעלה מ-50 תרגומים לשפות שמגדילים משמעותית את גודל ההורדה. כדי לשכפל ללא תרגומים, השתמשו בבדיקת sparse: > > **Bash / macOS / Linux:** > ```bash @@ -52,202 +52,205 @@ > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> זה נותן לכם את כל מה שאתם צריכים כדי להשלים את הקורס בהורדה מהירה יותר. +> זה נותן לכם את כל מה שצריך כדי להשלים את הקורס עם הורדה הרבה יותר מהירה. ## תוכן העניינים -- מבוא - - [ברוכים הבאים למשפחת פאי](./md/01.Introduction/01/01.PhiFamily.md) - - [הגדרת הסביבה שלכם](./md/01.Introduction/01/01.EnvironmentSetup.md) +- הקדמה + - [ברוכים הבאים למשפחת Phi](./md/01.Introduction/01/01.PhiFamily.md) + - [הגדרת הסביבה שלך](./md/01.Introduction/01/01.EnvironmentSetup.md) - [הבנת טכנולוגיות מפתח](./md/01.Introduction/01/01.Understandingtech.md) - - [בטיחות בינה מלאכותית לדגמי פאי](./md/01.Introduction/01/01.AISafety.md) - - [תמיכה בחומרה של פאי](./md/01.Introduction/01/01.Hardwaresupport.md) - - [דגמי פאי וזמינות בפלטפורמות שונות](./md/01.Introduction/01/01.Edgeandcloud.md) - - [שימוש ב-Guidance-ai ובפאי](./md/01.Introduction/01/01.Guidance.md) - - [דגמי Marketplace ב-GitHub](https://github.com/marketplace/models) - - [קטלוג דגמים של Azure AI](https://ai.azure.com) - -- אינפרנס פאי בסביבות שונות + - [בטיחות AI עבור דגמי Phi](./md/01.Introduction/01/01.AISafety.md) + - [תמיכה בחומרה של Phi](./md/01.Introduction/01/01.Hardwaresupport.md) + - [דגמי Phi וזמינות בפלטפורמות שונות](./md/01.Introduction/01/01.Edgeandcloud.md) + - [שימוש ב-Guidance-ai ו-Phi](./md/01.Introduction/01/01.Guidance.md) + - [דגמים ב-GitHub Marketplace](https://github.com/marketplace/models) + - [קטלוג דגמי Azure AI](https://ai.azure.com) + +- הסקת מידע עם Phi בסביבות שונות - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [דגמי GitHub](./md/01.Introduction/02/02.GitHubModel.md) + - [דגמים ב-GitHub](./md/01.Introduction/02/02.GitHubModel.md) - [קטלוג דגמי Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [ערכת כלים AI ל-VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [כלי AI ב-VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry מקומי](./md/01.Introduction/02/07.FoundryLocal.md) -- אינפרנס משפחת פאי - - [אינפרנס פאי ב-iOS](./md/01.Introduction/03/iOS_Inference.md) - - [אינפרנס פאי באנדרואיד](./md/01.Introduction/03/Android_Inference.md) - - [אינפרנס פאי ב-Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [אינפרנס פאי במחשב AI](./md/01.Introduction/03/AIPC_Inference.md) - - [אינפרנס פאי עם Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [אינפרנס פאי בשרת מקומי](./md/01.Introduction/03/Local_Server_Inference.md) - - [אינפרנס פאי בשרת מרוחק באמצעות AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [אינפרנס פאי עם Rust](./md/01.Introduction/03/Rust_Inference.md) - - [אינפרנס פאי--חזון במחשב מקומי](./md/01.Introduction/03/Vision_Inference.md) - - [אינפרנס פאי עם Kaito AKS, מכולות Azure (תמיכה רשמית)](./md/01.Introduction/03/Kaito_Inference.md) -- [כימות משפחת פאי](./md/01.Introduction/04/QuantifyingPhi.md) +- משפחת ההסקה Phi + - [הסקת Phi באייפון](./md/01.Introduction/03/iOS_Inference.md) + - [הסקת Phi במערכת אנדרואיד](./md/01.Introduction/03/Android_Inference.md) + - [הסקת Phi ב-Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [הסקת Phi במחשב AI](./md/01.Introduction/03/AIPC_Inference.md) + - [הסקת Phi עם מסגרת Apple MLX](./md/01.Introduction/03/MLX_Inference.md) + - [הסקת Phi בשרת מקומי](./md/01.Introduction/03/Local_Server_Inference.md) + - [הסקת Phi בשרת מרוחק באמצעות כלי AI](./md/01.Introduction/03/Remote_Interence.md) + - [הסקת Phi עם Rust](./md/01.Introduction/03/Rust_Inference.md) + - [הסקת Phi--ראייה מקומית](./md/01.Introduction/03/Vision_Inference.md) + - [הסקת Phi עם Kaito AKS, מיכלי Azure (תמיכה רשמית)](./md/01.Introduction/03/Kaito_Inference.md) +- [כימות משפחת Phi](./md/01.Introduction/04/QuantifyingPhi.md) - [כימות Phi-3.5 / 4 באמצעות llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [כימות Phi-3.5 / 4 באמצעות תוספי בינה מלאכותית גנרטיביים ל-onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [כימות Phi-3.5 / 4 באמצעות הרחבות Generative AI ל-onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [כימות Phi-3.5 / 4 באמצעות Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [כימות Phi-3.5 / 4 באמצעות Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [כימות Phi-3.5 / 4 באמצעות מסגרת Apple MLX](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- הערכת פאי - - [בינה מלאכותית אחראית](./md/01.Introduction/05/ResponsibleAI.md) +- הערכת Phi + - [AI שאפתני](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry להערכה](./md/01.Introduction/05/AIFoundry.md) - [שימוש ב-Promptflow להערכה](./md/01.Introduction/05/Promptflow.md) - RAG עם Azure AI Search - - [כיצד להשתמש ב-Phi-4-mini ו-Phi-4-multimodal (RAG) עם Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [כיצד להשתמש ב-Phi-4-mini וב-Phi-4-multimodal(RAG) עם Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG היברידי מקומי ללא ענן עם SQLite FTS5 ו-Phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- דוגמאות לפיתוח יישומי פאי - - יישומי טקסט ושיחה +- דוגמאות לפיתוח יישומי Phi + - יישומי טקסט וצ'אט - דוגמאות Phi-4 - - [📓] [שיחה עם מודל Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [שיחה עם מודל ONNX מקומי Phi-4 .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [אפליקציית שיחה במצב Console .NET עם Phi-4 ONNX באמצעות Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - דוגמאות Phi-3 / 3.5 + - [📓] [צ'אט עם דגם Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [צ'אט עם דגם Phi-4 ONNX מקומי ב-.NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [יישום קונסולה ב-.NET לצ'אט עם Phi-4 ONNX באמצעות Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 דוגמאות - [צ'אטבוט מקומי בדפדפן באמצעות Phi3, ONNX Runtime Web ו-WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [ריבוי מודלים - אינטראקטיבי Phi-3-mini ו-OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - בניית עטיפה ושימוש ב-Phi-3 עם MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [אופטימיזציית מודל - איך לאופטימיזציית מודל Phi-3-min עבור ONNX Runtime Web עם Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [OpenVino צ'אט](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [רב-דגם - Phi-3-mini אינטראקטיבי ו-OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - בניית מעטפת ושימוש ב-Phi-3 עם MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [אופטימיזציית מודל - איך לאופטימיזציה למודל Phi-3-min עבור ONNX Runtime Web באמצעות Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [אפליקציית WinUI3 עם Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - - [אפליקציית דוגמאות הערות מבוססת AI מרובה מודלים WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [כיוונון עדין ואינטגרציה של מודלים מותאמים אישית של Phi-3 עם Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [כיוונון עדין ואינטגרציה של מודלים מותאמים אישית של Phi-3 עם Prompt flow ב-Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [הערכת מודל Phi-3 / Phi-3.5 מכוונן עדין ב-Microsoft Foundry תוך התמקדות בעקרונות AI אחראיים של מיקרוסופט](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [דוגמת חיזוי שפה Phi-3.5-mini-instruct (סינית/אנגלית)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + -[אפליקציית הערות עם AI מבוסס דגם רב WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [כיוונון מדויק ושילוב דגמי Phi-3 מותאמים עם Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [כיוונון מדויק ושילוב דגמי Phi-3 מותאמים עם Prompt flow ב-Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [הערכת מודל Phi-3 / Phi-3.5 מכוונן ב-Microsoft Foundry תוך התמקדות בעקרונות ה-AI האחראי של מיקרוסופט](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [דוגמת ניבוי שפה Phi-3.5-mini-instruct (סינית/אנגלית)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [צ'אטבוט RAG עם Phi-3.5-Instruct WebGPU](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [שימוש ב-GPU של Windows ליצירת פתרון Prompt flow עם Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [שימוש ב-Microsoft Phi-3.5 tflite ליצירת אפליקציית Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [דוגמת שאלות ותשובות .NET המשתמשת במודל מקומי ONNX Phi-3 עם Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [אפליקציית שיחת קונסולה .NET עם Semantic Kernel ו-Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [שימוש ב-Phi-3.5 tflite של Microsoft ליצירת אפליקציית אנדרואיד](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [דוגמת שאלות ותשובות ב-.NET המשתמשת במודל ONNX Phi-3 מקומי באמצעות Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [אפליקציית שורת פקודה .NET עם Semantic Kernel ו-Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - דוגמאות מבוססות קוד Azure AI Inference SDK + - דוגמאות קוד מבוססות Azure AI Inference SDK - דוגמאות Phi-4 - - [📓] [יצירת קוד פרויקט באמצעות Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [📓] [יצירת קוד לפרויקט באמצעות Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - דוגמאות Phi-3 / 3.5 - - [בנה את שיחת קופילוט Visual Studio Code שלך עם משפחת Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [צור סוכן שיחת קופילוט Visual Studio Code מותאם אישית עם Phi-3.5 על ידי מודלים של GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [בנה את עצמך צ'אט GitHub Copilot עבור Visual Studio Code עם משפחת Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [צור סוכן צ'אט Copilot עבור Visual Studio Code עם Phi-3.5 על ידי דגמי GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - דוגמאות חשיבה מתקדמת - דוגמאות Phi-4 - [📓] [דוגמאות חשיבה Phi-4-mini או Phi-4](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [כיוונון עדין של Phi-4-mini-reasoning עם Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [כיוונון עדין של Phi-4-mini-reasoning עם Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning עם מודלים של GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning עם מודלים של Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [📓] [כיוונון מדויק לחשיבת Phi-4-mini באמצעות Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [כיוונון מדויק לחשיבת Phi-4-mini באמצעות Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [חשיבת Phi-4-mini עם דגמי GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [חשיבת Phi-4-mini עם דגמי Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - הדגמות - - [דוגמאות Phi-4-mini מתארחות ב-Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [דוגמאות Phi-4-multimodal מתארחות ב-Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - דוגמאות ראייה + - [הדגמות Phi-4-mini מאוחסנות ב-Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [הדגמות Phi-4-multimodal מאוחסנות ב-Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - דוגמאות הראייה - דוגמאות Phi-4 - - [📓] [שימוש ב-Phi-4-multimodal לקריאת תמונות ויצירת קוד](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [📓] [שימוש ב-Phi-4-multimodal לקריאת תמונות וליצירת קוד](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - דוגמאות Phi-3 / 3.5 - - [📓][Phi-3-vision תמונת טקסט לטקסט](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][טקסט מתמונה לטקסט Phi-3-vision](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][הטמעת CLIP של Phi-3-vision](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [הדגמה: מיחזור Phi-3](https://github.com/jennifermarsman/PhiRecycling/) - [עוזר שפה חזותית Phi-3-vision עם Phi3-Vision ו-OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision דוגמה רב-מסגרת או רב-תמונה](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [מודל מקומי של Phi-3 Vision ONNX באמצעות Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [מודל מקומי של Phi-3 Vision ONNX מבוסס תפריט באמצעות Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][דוגמת מרובים של Φhi-3.5 Vision או תמונות מרובות](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [מודל ONNX מקומי של Phi-3 Vision באמצעות Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [מודל ONNX מקומי של Phi-3 Vision מבוסס תפריט באמצעות Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - דוגמאות ראייה-וחשיבה + - דוגמאות חשיבה והראייה - Phi-4-Reasoning-Vision-15B - - [📓] [שימוש ב-Phi-4-Reasoning-Vision-15B לזיהוי הליכה מחוץ למעבר חציה](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [שימוש ב-Phi-4-Reasoning-Vision-15B לזיהוי הליכה מסוכנת על כביש](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [שימוש ב-Phi-4-Reasoning-Vision-15B במתמטיקה](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [שימוש ב-Phi-4-Reasoning-Vision-15B לזיהוי UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [שימוש ב-Phi-4-Reasoning-Vision-15B לזיהוי ממשק משתמש (UI)](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - דוגמאות מתמטיקה - - דוגמאות Phi-4-Mini-Flash-Reasoning-Instruct [דמו מתמטיקה עם Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - דוגמאות Phi-4-Mini-Flash-Reasoning-Instruct [הדגמת מתמטיקה עם Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - דוגמאות אודיו - דוגמאות Phi-4 - - [📓] [חילוץ תמלולים מאודיו באמצעות Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [חילוץ תמלולים של אודיו באמצעות Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [דוגמת אודיו Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [דוגמת תרגום דיבור Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [אפליקציית קונסולה .NET המשתמשת ב-Phi-4-multimodal לאנליזה של קובץ אודיו ויצירת תמלול](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [אפליקציית קונסול .NET המשתמשת ב-Phi-4-multimodal לניתוח קובץ אודיו ויצירת תמלול](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - דוגמאות MOE - דוגמאות Phi-3 / 3.5 - - [📓] [דוגמת מדיה חברתית של מודלים Phi-3.5 Mixture of Experts (MoEs)](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [בניית צינור הפקה מוגבר על ידי אחזור (RAG) עם NVIDIA NIM Phi-3 MOE, Azure AI Search ו-LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - + - [📓] [דוגמאות דגמי מומחים מעורבים (MoEs) של Phi-3.5 לסושיאל מדיה](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [בניית צינור הפקה עם אוגמנטציה לשחזור (RAG) באמצעות NVIDIA NIM Phi-3 MOE, Azure AI Search, ו-LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - - דוגמאות קריאת פונקציות - דוגמאות Phi-4 🆕 - - [📓] [שימוש בקריאת פונקציה עם Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [שימוש בקריאת פונקציה ליצירת סוכנים מרובים עם Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [שימוש בקריאת פונקציה עם Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [שימוש בקריאת פונקציה עם ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - דוגמאות עירוב רב-מודלי + - [📓] [שימוש בקריאת פונקציות עם Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [שימוש בקריאת פונקציות ליצירת סוכני ריבוי עם Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [שימוש בקריאת פונקציות עם Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [שימוש בקריאת פונקציות עם ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - דוגמאות מיזוג מולטימודאלי - דוגמאות Phi-4 🆕 - - [📓] [שימוש ב-Phi-4-multimodal כעיתונאי טכנולוגיה](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [אפליקציית קונסולה .NET המשתמשת ב-Phi-4-multimodal לניתוח תמונות](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- כיוונון עדין של דגמי Phi - - [תסריטי כיוונון עדין](./md/03.FineTuning/FineTuning_Scenarios.md) - - [כיוונון עדין מול RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [כיוונון עדין: להפוך את Phi-3 למומחה תעשייתי](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [כיוונון עדין של Phi-3 עם ערכת כלים ל-AI ל-VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [כיוונון עדין של Phi-3 עם שירות למידת מכונה של Azure](./md/03.FineTuning/Introduce_AzureML.md) - - [כיוונון עדין של Phi-3 עם Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [כיוונון עדין של Phi-3 עם QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [כיוונון עדין של Phi-3 עם Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [כיוונון עדין של Phi-3 עם Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [כיוונון עדין עם Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [כיוונון עדין עם מעבדה מעשית Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [כיוונון עדין של Phi-3-vision עם Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [📓] [שימוש ב-Phi-4-multimodal כעיתונאי טכנולוגיה](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [אפליקציית קונסול .NET המשתמשת ב-Phi-4-multimodal לניתוח תמונות](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- כיוונון מדויק לדגמי Phi + - [תרחישי כיוונון מדויק](./md/03.FineTuning/FineTuning_Scenarios.md) + - [כיוונון מדויק לעומת RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [כיוונון מדויק: הפוך את Phi-3 למומחה תעשייתי](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [כיוונון מדויק ל-Phi-3 עם כלי AI עבור VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [כיוונון מדויק ל-Phi-3 עם Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [כיוונון מדויק ל-Phi-3 עם Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [כיוונון מדויק ל-Phi-3 עם QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [כיוונון מדויק ל-Phi-3 עם Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [כיוונון מדויק ל-Phi-3 עם Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [כיוונון מדויק עם Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [כיוונון מדויק עם Microsoft Olive במחברת מעשית](./md/03.FineTuning/olive-lab/readme.md) + - [כיוונון מדויק ל-Phi-3-vision עם Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [כיוונון עדין של Phi-3 עם מסגרת Apple MLX](./md/03.FineTuning/FineTuning_MLX.md) - [כיוונון עדין של Phi-3-vision (תמיכה רשמית)](./md/03.FineTuning/FineTuning_Vision.md) - - [כוונון עדין של Phi-3 עם Kaito AKS, מיכלי Azure (תמיכה רשמית)](./md/03.FineTuning/FineTuning_Kaito.md) - - [כוונון עדין של Phi-3 ו-3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + - [כיוונון עדין של Phi-3 עם Kaito AKS, מכולות Azure (תמיכה רשמית)](./md/03.FineTuning/FineTuning_Kaito.md) + - [כיוונון עדין של Phi-3 ו-Phi-3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) -- סדנת מעשית - - [חקירת מודלים חדשניים: LLMs, SLMs, פיתוח מקומי ועוד](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [שחרור הפוטנציאל של NLP: כוונון עדין עם Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) +- מעבדת עשה-זאת-בעצמך + - [חקירת מודלים פורצי דרך: LLM, SLM, פיתוח מקומי ועוד](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [שחרור הפוטנציאל של NLP: כיוונון עדין עם Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - מאמרים אקדמיים ופרסומים - - [Textbooks Are All You Need II: דוח טכני של phi-1.5](https://arxiv.org/abs/2309.05463) - - [דוח טכני של Phi-3: מודל שפה בעל יכולת גבוהה במכשירך הנייד](https://arxiv.org/abs/2404.14219) - - [דוח טכני של Phi-4](https://arxiv.org/abs/2412.08905) - - [דוח טכני של Phi-4-Mini: מודלים רב-מודאליים קטנים אך חזקים באמצעות תערובת של LoRAs](https://arxiv.org/abs/2503.01743) - - [אופטימיזציה של מודלי שפה קטנים לפונקציית קריאה ברכב](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) כוונון עדין של PHI-3 למענה על שאלות רב-ברירה: מתודולוגיה, תוצאות ואתגרים](https://arxiv.org/abs/2501.01588) - - [דוח טכני של Phi-4-הגיוני](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [דוח טכני של Phi-4-mini-הגיוני](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Textbooks Are All You Need II: דו"ח טכני של phi-1.5](https://arxiv.org/abs/2309.05463) + - [דו"ח טכני של Phi-3: מודל שפה מתקדם במיוחד במכשירך](https://arxiv.org/abs/2404.14219) + - [דו"ח טכני של Phi-4](https://arxiv.org/abs/2412.08905) + - [דו"ח טכני של Phi-4-Mini: מודלים רב-ממדיים קומפקטיים אך רבי עוצמה באמצעות תערובת LoRAs](https://arxiv.org/abs/2503.01743) + - [אופטימיזציה של מודלים קטנים לקריאה פונקציונלית ברכב](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) כיוונון עדין של PHI-3 למענה על שאלות רב-ברירתיות: מתודולוגיה, תוצאות, ואתגרים](https://arxiv.org/abs/2501.01588) + - [דו"ח טכני של Phi-4- reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [דו"ח טכני של Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## שימוש במודלי Phi -### Phi ב-Microsoft Foundry +### Phi על Microsoft Foundry -אתה יכול ללמוד כיצד להשתמש ב-Microsoft Phi וכיצד לבנות פתרונות מקצה לקצה (E2E) במגוון ההתקנים שלך. כדי לחוות את Phi בעצמך, התחל לשחק עם המודלים ולבצע התאמות אישיות ל-Phi עבור התרחישים שלך באמצעות [קטלוג מודלי AI של Microsoft Foundry Azure](https://aka.ms/phi3-azure-ai). ניתן ללמוד עוד ב-התחלה מהירה עם [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +תוכלו ללמוד כיצד להשתמש ב-Phi של מיקרוסופט וכיצד לבנות פתרונות מקצה לקצה במכשירי החומרה השונים שלכם. כדי להתנסות ב-Phi בעצמכם, התחילו לשחק עם המודלים ולהתאים אותם לתרחישים שלכם באמצעות [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) תוכלו ללמוד עוד ב-התחלת עבודה עם [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) **מגרש משחקים** -לכל מודל יש מגרש משחקים ייעודי לבדיקת המודל ב-[Azure AI Playground](https://aka.ms/try-phi3). +לכל מודל יש מגרש משחקים ייעודי לבחינת המודל [Azure AI Playground](https://aka.ms/try-phi3). -### Phi במודלי GitHub +### Phi על מודלים ב-GitHub -אתה יכול ללמוד כיצד להשתמש ב-Microsoft Phi וכיצד לבנות פתרונות מקצה לקצה במגוון ההתקנים שלך. כדי לחוות את Phi בעצמך, התחל לשחק עם המודל ולבצע התאמות אישיות ל-Phi עבור התרחישים שלך באמצעות [קטלוג מודלים של GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). ניתן ללמוד עוד ב-התחלה מהירה עם [קטלוג מודלים של GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) +תוכלו ללמוד כיצד להשתמש ב-Phi של מיקרוסופט וכיצד לבנות פתרונות מקצה לקצה במכשירי החומרה השונים שלכם. כדי להתנסות ב-Phi בעצמכם, התחילו לשחק עם המודל ולהתאים את Phi לתרחישים שלכם באמצעות [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) תוכלו ללמוד עוד ב-התחלת עבודה עם [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) **מגרש משחקים** -לכל מודל יש מגרש משחקים ייעודי [לבדיקת המודל](/md/02.QuickStart/GitHubModel_QuickStart.md). +לכל מודל יש [מגרש משחקים ייעודי לבדיקת המודל](/md/02.QuickStart/GitHubModel_QuickStart.md). -### Phi ב-Hugging Face +### Phi על Hugging Face -ניתן גם למצוא את המודל ב-[Hugging Face](https://huggingface.co/microsoft) +תוכלו למצוא גם את המודל ב-[Hugging Face](https://huggingface.co/microsoft) **מגרש משחקים** - [מגרש המשחקים של Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) + [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 קורסים נוספים + ## 🎒 קורסים נוספים -הצוות שלנו מייצר קורסים נוספים! בדוק: +הצוות שלנו מפיק קורסים נוספים! בדקו: ### LangChain @@ -264,11 +267,11 @@ --- -### סדרת AI יצירתי -[![AI יצירתי למתחילים](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI יצירתי (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![AI יצירתי (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![AI יצירתי (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### סדרת AI גנרטיבי +[![AI גנרטיבי למתחילים](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI גנרטיבי (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![AI גנרטיבי (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![AI גנרטיבי (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- @@ -276,49 +279,51 @@ [![למידת מכונה למתחילים](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![מדעי הנתונים למתחילים](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI למתחילים](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![אבטחת סייבר למתחילים](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![פיתוח ווב למתחילים](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![סייברסקיוריטי למתחילים](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![פיתוח אתרי אינטרנט למתחילים](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![IoT למתחילים](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![פיתוח XR למתחילים](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### סדרת Copilot -[![Copilot לתכנות זוגי עם AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot ל-C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![הרפתקאות Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +### סדרת קופיילוט +[![קופיילוט לתכנות משותף עם AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![קופיילוט ל-C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![הרפתקאות קופיילוט](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## AI אחראי +## AI אחראית + +מיקרוסופט מחויבת לסייע ללקוחותיה להשתמש במוצרי ה-AI שלנו באחריות, לשתף את הלקחים שלנו ולבנות שותפויות מבוססות אמון באמצעות כלים כמו הערות שקיפות והערכות השפעה. משאבים רבים אלו זמינים ב-[https://aka.ms/RAI](https://aka.ms/RAI). +הגישה של מיקרוסופט ל-AI אחראית מבוססת על עקרונות ה-AI שלנו: הוגנות, אמינות ובטיחות, פרטיות וביטחון, הכללה, שקיפות ואחריות. + +מודלים להיקף רחב של שפה טבעית, תמונה ודיבור - כמו אלו שבדוגמה זו - עלולים להתנהג בצורה לא הוגנת, לא אמינה או פוגענית, ולגרום נזקים. אנא עיינו ב-[הערת השקיפות של שירות Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) כדי להתעדכן בסיכונים והמגבלות. -מיקרוסופט מחויבת לסייע ללקוחותינו להשתמש במוצרי ה-AI שלנו באחריות, לשתף את הלמידות שלנו ולבנות שותפויות מבוססות אמון באמצעות כלים כמו הערות שקיפות והערכות השפעה. משאבים רבים ניתן למצוא ב-[https://aka.ms/RAI](https://aka.ms/RAI). -הגישה של מיקרוסופט ל-AI אחראי מבוססת על עקרונות ה-AI שלנו שהם הגינות, אמינות ובטיחות, פרטיות ואבטחה, הכללה, שקיפות ולקיחת אחריות. -מודלים רחבי היקף של שפה טבעית, תמונה ודיבור - כמו אלה המשמשים בדוגמה זו - עלולים להתנהג בדרכים שאינן הוגנות, לא אמינות, או פוגעניות, ולגרום לנזקים. יש לעיין ב-[הערת השקיפות של שירות Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) כדי להתעדכן לגבי סיכונים ומגבלות. -הגישה המומלצת להקטנת סיכונים אלו היא לכלול במערכת שלך מערכת בטיחות שיכולה לזהות ולמנוע התנהגות מזיקה. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) מספק שכבת הגנה עצמאית, המסוגלת לזהות תוכן מזיק שנוצר על ידי משתמשים או בינה מלאכותית ביישומים ובשירותים. Azure AI Content Safety כוללת ממשקי API לטקסט ולתמונה המאפשרים לך לזהות חומר מזיק. בתוך Microsoft Foundry, שירות ה-Content Safety מאפשר לך לצפות, לחקור ולנסות דוגמאות קוד לזיהוי תוכן מזיק במגוון מודלים. תיעוד [ההתחלה המהירה הבא](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ינחה אותך כיצד לבצע בקשות לשירות. +הגישה המומלצת לצמצום סיכונים אלה היא לכלול מערכת בטיחות בארכיטקטורה שלך שיכולה לזהות ולמנוע התנהגות מזיקה. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) מספק שכבת הגנה עצמאית, היכולה לזהות תוכן מזיק שנוצר על ידי משתמשים ו-AI באפליקציות ושירותים. Azure AI Content Safety כוללת ממשקי API לטקסט ותמונה שמאפשרים לך לזהות חומר מזיק. בתוך Microsoft Foundry, שירות Content Safety מאפשר לך לצפות, לחקור ולנסות קוד לדוגמה לזיהוי תוכן מזיק במגוון מודאליות. המסמך [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) הבא מדריך אותך כיצד לבצע בקשות לשירות. -היבט נוסף שיש לקחת בחשבון הוא ביצועי היישום הכוללים. ביישומים רב-מודאליים ורב-מודלים, אנו רואים בביצועים מענה לציפיות שלך ושל המשתמשים, לרבות אי-יצירת תוצאות מזיקות. חשוב להעריך את ביצועי היישום הכולל באמצעות [מדרגי ביצועים, איכות, סיכון ובטיחות](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). יש לך גם את האפשרות ליצור ולהעריך באמצעות [מדרגים מותאמים אישית](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +היבט נוסף שיש לקחת בחשבון הוא ביצועי האפליקציה הכוללים. עם אפליקציות רב-מודאליות ורב-מודליות, אנו מתייחסים לביצועים כהתנהגות המערכת כפי שאתה והמשתמשים שלך מצפים, כולל אי יצירת תוצאות מזיקות. חשוב להעריך את ביצועי האפליקציה הכוללת שלך באמצעות [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). יש לך גם את היכולת ליצור ולהעריך עם [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -אתה יכול להעריך את יישום הבינה המלאכותית שלך בסביבת הפיתוח שלך באמצעות [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). בהתחשב במערכת נתוני בדיקה או מטרה, יצירות ה-AI הגנרטיביות שלך נמדדות כמותית באמצעות מדרגים מובנים או מדרגים מותאמים אישית שבחרת. בכדי להתחיל עם Azure AI Evaluation SDK להערכת המערכת שלך, תוכל לעקוב אחרי [מדריך ההתחלה המהירה](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). לאחר ביצוע ריצת הערכה, תוכל [להמחיש את התוצאות ב-Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +תוכל להעריך את אפליקציית ה-AI שלך בסביבת הפיתוח שלך באמצעות [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). בהתבסס על מערך נתונים לבחינה או יעד, יצירות ה-AI הגנרטיביות שלך נמדדות כמותית בעזרת מעריכי ביצועים מובנים או מעריכים מותאמים אישית לפי בחירתך. כדי להתחיל עם Azure AI Evaluation SDK להערכת המערכת שלך, תוכל לעקוב אחר [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). לאחר ביצוע ריצת הערכה, תוכל [להציג את התוצאות ב-Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## סימני מסחר +## סמני מסחר -פרויקט זה עשוי להכיל סימני מסחר או לוגואים של פרויקטים, מוצרים או שירותים. שימוש מורשה בסמלי מסחר או בלוגואים של Microsoft כפוף ל-[הנחיות למותג וסימני מסחר של Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -שימוש בסמלי מסחר או בלוגואים של Microsoft בגרסאות שערוכו של פרויקט זה חייב לא לגרום לבלבול או לרמז על חסות של Microsoft. כל שימוש בסמלי מסחר או בלוגואים של צדדים שלישיים כפוף למדיניות אותם צדדים. +פרויקט זה עשוי להכיל סמני מסחר או לוגואים של פרויקטים, מוצרים או שירותים. שימוש מורשה בסמני המסחר או בלוגואים של מיקרוסופט כפוף ומחויב לעמוד ב-[Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +שימוש בסמני מסחר או לוגואים של מיקרוסופט בגרסאות שתוקנו של פרויקט זה אסור שיגרום לבלבול או יכוון לרמז על חסות של מיקרוסופט. כל שימוש בסמני מסחר או לוגואים של צדדים שלישיים כפוף למדיניות של אותם צדדים שלישיים. ## קבלת עזרה -אם נתקעת או יש לך שאלות לגבי בניית יישומי בינה מלאכותית, הצטרף: +אם נתקעת או יש לך שאלות בנוגע לבניית אפליקציות AI, הצטרף ל: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -אם יש לך משוב מוצר או שגיאות בזמן הבנייה, בקר: +אם יש לך משוב על המוצר או שגיאות בזמן הבנייה, בקר ב: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**כתב זכויות**: -מסמך זה תורגם באמצעות שירות תרגום מבוסס בינה מלאכותית [Co-op Translator](https://github.com/Azure/co-op-translator). למרות שאנו שואפים לדיוק, יש לקחת בחשבון כי תרגומים אוטומטיים עלולים להכיל שגיאות או אי-דיוקים. יש להתייחס למסמך המקורי בשפתו כמקור הסמכותי. עבור מידע קריטי, מומלץ להשתמש בתרגום מקצועי על ידי אדם. איננו אחראים לכל הבנה שגויה או פרשנות מוטעית הנובעת משימוש בתרגום זה. +**כתב ויתור**: +מסמך זה תורגם באמצעות שירות תרגום אוטומטי [Co-op Translator](https://github.com/Azure/co-op-translator). למרות שאנו שואפים לדיוק, יש לקחת בחשבון שתרגומים אוטומטיים עלולים להכיל שגיאות או אי-דיוקים. יש להחשיב את המסמך המקורי בשפתו הטבעית כמקור הסמכות. למידע קריטי מומלץ להשתמש בתרגום מקצועי על ידי מתרגם אדם. אנו לא אחראים לכל אי-הבנה או פירוש שגוי הנובע מהשימוש בתרגום זה. \ No newline at end of file diff --git a/translations/he/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/he/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..369fad272 --- /dev/null +++ b/translations/he/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 רטריבל היברידי RAG עם Microsoft phi-4-mini ו-SQLite FTS5 (מודל שפה קטן בלי ענן)\n", + "\n", + "> **מחבר:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **מיקוד:** מודלים לשוניים קטנים (SLMs), SQLite FTS5 BM25, אמבדינג צפיפות, מיזוג דירוג הופכי (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. מוטיבציה: הדילמה של זיכרון מילות מפתח ב-SLM מקומי\n", + "ארכיטקטורות RAG סטנדרטיות התלויות רק באמבדינגים של וקטורים צפופים לרוב נכשלות בשליפת טוקנים מספריים מדויקים (כגון `2,340,000 TL`, קודי חוזים, מספרי חשבונות). \n", + "לעומת זאת, החיפוש הלכסיקלי הספאצי (BM25) מפספס מילים נרדפות סמנטיות ושאלות מפורפרזות.\n", + "\n", + "מדריך זה מראה כיצד לממש **מנוע רטריבל היברידי מהיר בזיכרון** שמשלב:\n", + "1. **וקטורים צפופים** (דמיון קוסינוס)\n", + "2. **חיפוש לקסיקלי דליל** (SQLite FTS5 BM25)\n", + "3. **מיזוג דירוג הופכי (RRF, $k=60$)**\n", + "4. **יצירת ציטוטים מבוססי מקור (`[1]`, `[2]`)** עם Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. סכימת SQLite כפולה (וקטורים צפופים + טבלת FTS5 וירטואלית BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. מדד קליטת וביצוע דוגמאות\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. ניסוח הנחיה מבוססת עבור Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**כתב ויתור**:\nמסמך זה תורגם באמצעות שירות תרגום אוטומטי [Co-op Translator](https://github.com/Azure/co-op-translator). למרות שאנו שואפים לדיוק, יש לקחת בחשבון שתרגומים אוטומטיים עלולים להכיל שגיאות או אי-דיוקים. יש להחשיב את המסמך המקורי בשפתו הטבעית כמקור הסמכות. למידע קריטי מומלץ להשתמש בתרגום מקצועי על ידי מתרגם אדם. אנו לא אחראים לכל אי-הבנה או פירוש שגוי הנובע מהשימוש בתרגום זה.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/hi/.co-op-translator.json b/translations/hi/.co-op-translator.json index d4cfb1e9d..5a1123322 100644 --- a/translations/hi/.co-op-translator.json +++ b/translations/hi/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "hi" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:16:49+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:29:15+00:00", "source_file": "README.md", "language_code": "hi" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "hi" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:30:09+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "hi" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:13+00:00", diff --git a/translations/hi/README.md b/translations/hi/README.md index 9afd84aa1..7e548ecfc 100644 --- a/translations/hi/README.md +++ b/translations/hi/README.md @@ -1,12 +1,12 @@ -# Phi कुकबुक: Microsoft के Phi मॉडलों के साथ व्यावहारिक उदाहरण +# फ़ाई कुकबुक: माइक्रोसॉफ्ट के फ़ाई मॉडलों के साथ हैंड्स-ऑन उदाहरण -[![GitHub Codespaces में नमूने खोलें और उपयोग करें](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![डेव कंटेनरों में खोलें](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![GitHub Codespaces में नमूनों को खोलें और उपयोग करें](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Dev Containers में खोलें](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub योगदानकर्ता](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub मुद्दे](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub पुल-रिक्वेस्ट](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs स्वागत है](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub पुल अनुरोध](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PR स्वागत है](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![GitHub वॉचर्स](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub फोर्क्स](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) @@ -14,29 +14,29 @@ [![Microsoft Foundry डिस्कॉर्ड](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi Microsoft द्वारा विकसित मुक्त स्रोत AI मॉडलों की एक श्रृंखला है। +फ़ाई माइक्रोसॉफ्ट द्वारा विकसित खुले स्रोत AI मॉडल्स की एक श्रृंखला है। -Phi वर्तमान में सबसे शक्तिशाली और किफायती छोटे भाषा मॉडल (SLM) है, जिसमें बहुभाषी, तर्कशक्ति, टेक्स्ट/चैट जनरेशन, कोडिंग, छवियाँ, ऑडियो और अन्य परिदृश्यों में बहुत अच्छे बेंचमार्क हैं। +फ़ाई वर्तमान में सबसे शक्तिशाली और किफायती स्मॉल लैंग्वेज मॉडल (SLM) है, जिसमें मल्टी-लैंग्वेज, तर्क, टेक्स्ट/चैट जनरेशन, कोडिंग, इमेज, ऑडियो और अन्य परिस्थितियों में बहुत अच्छे मानक हैं। -आप Phi को क्लाउड या एज उपकरणों पर तैनात कर सकते हैं, और सीमित कंप्यूटिंग शक्ति के साथ आसानी से जनरेटिव AI अनुप्रयोग बना सकते हैं। +आप फ़ाई को क्लाउड या एज उपकरणों पर तैनात कर सकते हैं, और सीमित कंप्यूटिंग शक्ति के साथ आसानी से जनरेटिव AI एप्लिकेशन बना सकते हैं। इन संसाधनों का उपयोग शुरू करने के लिए इन चरणों का पालन करें: -1. **संग्रहालय को फोर्क करें**: क्लिक करें [![GitHub फोर्क्स](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **संग्रहालय क्लोन करें**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Microsoft AI डिस्कॉर्ड समुदाय से जुड़ें और विशेषज्ञों तथा अन्य डेवलपर्स से मिलें**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +1. **रिपॉजिटरी को फोर्क करें**: क्लिक करें [![GitHub फोर्क्स](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **रिपॉजिटरी क्लोन करें**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Microsoft AI Discord समुदाय में शामिल हों और विशेषज्ञों और सह-डेवलपर्स से मिलें**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/hi/cover.eb18d1b9605d754b.webp) ### 🌐 बहुभाषी समर्थन -#### GitHub Action द्वारा समर्थित (स्वचालित और हमेशा अप-टू-डेट) +#### GitHub एक्शन के माध्यम से समर्थित (स्वचालित एवं हमेशा अपडेट) -[अंग्रेज़ी](../ar/README.md) | [बंगाली](../bn/README.md) | [बुल्गेरियाई](../bg/README.md) | [बर्मीज़ (म्यांमार)](../my/README.md) | [चीनी (सरलीकृत)](../zh-CN/README.md) | [चीनी (पारंपरिक, हॉन्ग कॉन्ग)](../zh-HK/README.md) | [चीनी (पारंपरिक, मकाओ)](../zh-MO/README.md) | [चीनी (पारंपरिक, ताइवान)](../zh-TW/README.md) | [क्रोएशियाई](../hr/README.md) | [चेक](../cs/README.md) | [डेनिश](../da/README.md) | [डच](../nl/README.md) | [एस्टोनियाई](../et/README.md) | [फ़िनिश](../fi/README.md) | [फ्रेंच](../fr/README.md) | [जर्मन](../de/README.md) | [ग्रीक](../el/README.md) | [हिब्रू](../he/README.md) | [हिन्दी](./README.md) | [हंगेरियन](../hu/README.md) | [इंडोनेशियाई](../id/README.md) | [इटालियन](../it/README.md) | [जापानी](../ja/README.md) | [कन्नड़](../kn/README.md) | [ख़मेर](../km/README.md) | [कोरियाई](../ko/README.md) | [लिथुआनियाई](../lt/README.md) | [मलय](../ms/README.md) | [मलयालम](../ml/README.md) | [मराठी](../mr/README.md) | [नेपाली](../ne/README.md) | [नाइजीरियाई पिजिन](../pcm/README.md) | [नॉर्वेजियन](../no/README.md) | [फारसी (पर्शियन)](../fa/README.md) | [पोलिश](../pl/README.md) | [पुर्तगाली (ब्राज़ील)](../pt-BR/README.md) | [पुर्तगाली (पुर्तगाल)](../pt-PT/README.md) | [पंजाबी (गुरमुखी)](../pa/README.md) | [रोमानियाई](../ro/README.md) | [रूसी](../ru/README.md) | [सर्बियाई (सिरिलिक)](../sr/README.md) | [स्लोवाक](../sk/README.md) | [स्लोवेनियाई](../sl/README.md) | [स्पेनिश](../es/README.md) | [स्वाहिली](../sw/README.md) | [स्वीडिश](../sv/README.md) | [टागालोग (फिलीपीनी)](../tl/README.md) | [तमिल](../ta/README.md) | [तेलुगू](../te/README.md) | [थाई](../th/README.md) | [तुर्की](../tr/README.md) | [यूक्रेनी](../uk/README.md) | [उर्दू](../ur/README.md) | [वियतनामी](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](./README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) > **स्थानीय रूप से क्लोन करना पसंद है?** > -> इस संग्रहालय में 50+ भाषा अनुवाद शामिल हैं जो डाउनलोड आकार को काफी बढ़ा देते हैं। अनुवादों के बिना क्लोन करने के लिए sparse checkout का उपयोग करें: +> इस रिपॉजिटरी में 50+ भाषा अनुवाद शामिल हैं जो डाउनलोड आकार को काफी बढ़ा देते हैं। बिना अनुवाद के क्लोन करने के लिए, स्पार्स चेक आउट का उपयोग करें: > > **Bash / macOS / Linux:** > ```bash @@ -52,202 +52,205 @@ Phi वर्तमान में सबसे शक्तिशाली औ > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> यह आपको कोर्स पूरा करने के लिए आवश्यक सब कुछ बहुत तेज़ डाउनलोड के साथ देता है। +> यह आपको कोर्स पूरा करने के लिए आवश्यक सब कुछ तेज़ डाउनलोड के साथ देता है। -## विषय सूची +## सामग्री सूची - परिचय - - [Phi परिवार में आपका स्वागत है](./md/01.Introduction/01/01.PhiFamily.md) - - [आपका पर्यावरण सेट अप करना](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [फ़ाई परिवार में आपका स्वागत है](./md/01.Introduction/01/01.PhiFamily.md) + - [अपने पर्यावरण की स्थापना](./md/01.Introduction/01/01.EnvironmentSetup.md) - [मुख्य तकनीकों को समझना](./md/01.Introduction/01/01.Understandingtech.md) - - [Phi मॉडलों के लिए AI सुरक्षा](./md/01.Introduction/01/01.AISafety.md) - - [Phi हार्डवेयर समर्थन](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi मॉडल और प्लेटफार्मों पर उपलब्धता](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Guidance-ai और Phi का उपयोग करना](./md/01.Introduction/01/01.Guidance.md) - - [GitHub मार्केटप्लेस मॉडल्स](https://github.com/marketplace/models) + - [फ़ाई मॉडलों के लिए AI सुरक्षा](./md/01.Introduction/01/01.AISafety.md) + - [फ़ाई हार्डवेयर समर्थन](./md/01.Introduction/01/01.Hardwaresupport.md) + - [फ़ाई मॉडल और प्लेटफ़ॉर्म पर उपलब्धता](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Guidance-ai और फ़ाई का उपयोग](./md/01.Introduction/01/01.Guidance.md) + - [GitHub मार्केटप्लेस मॉडल](https://github.com/marketplace/models) - [Azure AI मॉडल कैटलॉग](https://ai.azure.com) -- विभिन्न वातावरण में Phi का अनुमान +- विभिन्न वातावरणों में फ़ाई का अनुमान - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub मॉडल](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry मॉडल कैटलॉग](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI टूलकिट VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry स्थानीय](./md/01.Introduction/02/07.FoundryLocal.md) - -- Phi परिवार का अनुमान - - [iOS में Phi का अनुमान](./md/01.Introduction/03/iOS_Inference.md) - - [Android में Phi का अनुमान](./md/01.Introduction/03/Android_Inference.md) - - [Jetson में Phi का अनुमान](./md/01.Introduction/03/Jetson_Inference.md) - - [AI PC में Phi का अनुमान](./md/01.Introduction/03/AIPC_Inference.md) - - [Apple MLX फ्रेमवर्क के साथ Phi का अनुमान](./md/01.Introduction/03/MLX_Inference.md) - - [स्थानीय सर्वर में Phi का अनुमान](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI टूलकिट का उपयोग कर रिमोट सर्वर में Phi का अनुमान](./md/01.Introduction/03/Remote_Interence.md) - - [Rust के साथ Phi का अनुमान](./md/01.Introduction/03/Rust_Inference.md) - - [स्थानीय में Phi--विजन का अनुमान](./md/01.Introduction/03/Vision_Inference.md) - - [Kaito AKS, Azure कंटेनरों (आधिकारिक समर्थन) के साथ Phi का अनुमान](./md/01.Introduction/03/Kaito_Inference.md) -- [Phi परिवार का क्वांटिफिकेशन](./md/01.Introduction/04/QuantifyingPhi.md) - - [llama.cpp का उपयोग करके Phi-3.5 / 4 का क्वांटाइजेशन](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime के लिए जनरेटिव AI एक्सटेंशन्स का उपयोग करके Phi-3.5 / 4 का क्वांटाइजेशन](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Intel OpenVINO का उपयोग करके Phi-3.5 / 4 का क्वांटाइजेशन](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Apple MLX फ्रेमवर्क का उपयोग करके Phi-3.5 / 4 का क्वांटाइजेशन](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Phi का मूल्यांकन + - [Foundry लोकल](./md/01.Introduction/02/07.FoundryLocal.md) + +- फ़ाई परिवार का अनुमान + - [iOS में फ़ाई का अनुमान](./md/01.Introduction/03/iOS_Inference.md) + - [Android में फ़ाई का अनुमान](./md/01.Introduction/03/Android_Inference.md) + - [Jetson में फ़ाई का अनुमान](./md/01.Introduction/03/Jetson_Inference.md) + - [AI PC में फ़ाई का अनुमान](./md/01.Introduction/03/AIPC_Inference.md) + - [Apple MLX फ्रेमवर्क के साथ फ़ाई का अनुमान](./md/01.Introduction/03/MLX_Inference.md) + - [लोकल सर्वर में फ़ाई का अनुमान](./md/01.Introduction/03/Local_Server_Inference.md) + - [AI टूलकिट का उपयोग करके रिमोट सर्वर में फ़ाई का अनुमान](./md/01.Introduction/03/Remote_Interence.md) + - [Rust के साथ फ़ाई का अनुमान](./md/01.Introduction/03/Rust_Inference.md) + - [लोकल में फ़ाई-विज़न का अनुमान](./md/01.Introduction/03/Vision_Inference.md) + - [Kaito AKS, Azure कंटेनर (आधिकारिक समर्थन) के साथ फ़ाई का अनुमान](./md/01.Introduction/03/Kaito_Inference.md) +- [फ़ाई परिवार का क्वांटिफ़िकेशन](./md/01.Introduction/04/QuantifyingPhi.md) + - [llama.cpp का उपयोग करके Phi-3.5 / 4 का क्वांटाइज़ेशन](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntime के लिए जनरेटिव AI एक्सटेंशन का उपयोग करके Phi-3.5 / 4 का क्वांटाइज़ेशन](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Intel OpenVINO का उपयोग करके Phi-3.5 / 4 का क्वांटाइज़ेशन](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Apple MLX फ्रेमवर्क का उपयोग करके Phi-3.5 / 4 का क्वांटाइज़ेशन](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- फ़ाई का मूल्यांकन - [जिम्मेदार AI](./md/01.Introduction/05/ResponsibleAI.md) - - [मूल्यांकन के लिए Microsoft Foundry](./md/01.Introduction/05/AIFoundry.md) - - [मूल्यांकन के लिए Promptflow का उपयोग](./md/01.Introduction/05/Promptflow.md) + - [माइक्रोसॉफ्ट फाउंड्री के लिए मूल्यांकन](./md/01.Introduction/05/AIFoundry.md) + - [मूल्यांकन के लिए प्रॉम्प्टफ़्लो का उपयोग](./md/01.Introduction/05/Promptflow.md) -- Azure AI सर्च के साथ RAG - - [Azure AI सर्च के साथ Phi-4-mini और Phi-4-multimodal (RAG) कैसे उपयोग करें](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- Azure AI खोज के साथ RAG + - [Azure AI खोज के साथ Phi-4-mini और Phi-4-multimodal(RAG) का उपयोग कैसे करें](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [SQLite FTS5 और phi-4-mini के साथ Zero-Cloud लोकल हाइब्रिड RAG](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi एप्लिकेशन विकास नमूने +- फ़ाई एप्लिकेशन विकास के नमूने - टेक्स्ट और चैट एप्लिकेशन - - Phi-4 नमूने + - फ़ाई-4 नमूने - [📓] [Phi-4-mini ONNX मॉडल के साथ चैट करें](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 स्थानीय ONNX मॉडल .NET के साथ चैट करें](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Sementic Kernel का उपयोग कर Phi-4 ONNX के साथ .NET कंसोल ऐप में चैट](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - [Phi-4 लोकल ONNX मॉडल .NET के साथ चैट करें](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Sementic Kernel का उपयोग करके Phi-4 ONNX के साथ .NET कंसोल ऐप चैट](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 नमूने - - [Phi3, ONNX Runtime Web और WebGPU का उपयोग करके ब्राउज़र में स्थानीय चैटबोट](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Phi3, ONNX Runtime Web और WebGPU का उपयोग करके ब्राउज़र में स्थानीय चैटबॉट](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino चैट](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [मल्टी मॉडल - इंटरैक्टिव Phi-3-mini और OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - एक रैपर बनाना और Phi-3 को MLFlow के साथ उपयोग करना](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [मॉडल अनुकूलन - ONNX Runtime Web के लिए Phi-3-min मॉडल को Olive के साथ कैसे अनुकूलित करें](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [Phi-3 mini-4k-instruct-onnx के साथ WinUI3 ऐप](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 मल्टी मॉडल एआई पावर्ड नोट्स ऐप सैंपल](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [कस्टम Phi-3 मॉडल्स को फाइन-ट्यून और इंटीग्रेट करें Prompt flow के साथ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Microsoft Foundry में Prompt flow के साथ कस्टम Phi-3 मॉडल्स को फाइन-ट्यून और इंटीग्रेट करें](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Microsoft के जिम्मेदार AI सिद्धांतों पर ध्यान केंद्रित करते हुए Microsoft Foundry में फाइन-ट्यून किए गए Phi-3 / Phi-3.5 मॉडल का मूल्यांकन करें](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct भाषा भविष्यवाणी सैंपल (चीनी/अंग्रेज़ी)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [मल्टी मॉडल - इंटरएक्टिव Phi-3-mini और OpenAI व्हिस्पर](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Phi-3 के साथ रैपर बनाना और MLFlow का उपयोग करना](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [मॉडल ऑप्टिमाइज़ेशन - Olive के साथ ONNX Runtime Web के लिए Phi-3-min मॉडल को कैसे ऑप्टिमाइज़ करें](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 ऐप Phi-3 mini-4k-instruct-onnx के साथ](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 मल्टी मॉडल AI संचालित नोट्स ऐप नमूना](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [कस्टम Phi-3 मॉडल को फाइन-ट्यून और एकीकृत करें Prompt flow के साथ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [माइक्रोसॉफ्ट फाउंड्री में Prompt flow के साथ कस्टम Phi-3 मॉडल को फाइन-ट्यून और एकीकृत करें](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [माइक्रोसॉफ्ट फाउंड्री में माइक्रोसॉफ्ट के जिम्मेदार AI सिद्धांतों पर ध्यान केंद्रित करते हुए Phi-3 / Phi-3.5 मॉडल का मूल्यांकन करें](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct भाषा भविष्यवाणी नमूना (चीनी/अंग्रेज़ी)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG चैटबॉट](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU का उपयोग करके Prompt flow समाधान बनाना Phi-3.5-Instruct ONNX के साथ](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Microsoft Phi-3.5 tflite का उपयोग करके Android ऐप बनाना](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [स्थानीय ONNX Phi-3 मॉडल का उपयोग करते हुए Q&A .NET उदाहरण Microsoft.ML.OnnxRuntime के साथ](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Semantic Kernel और Phi-3 के साथ कंसोल चैट .NET ऐप](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI Inference SDK कोड आधारित सैंपल्स - - Phi-4 सैंपल्स - - [📓] [Phi-4-multimodal का उपयोग करके प्रोजेक्ट कोड जेनरेट करें](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 सैंपल्स - - [Microsoft Phi-3 परिवार के साथ अपना Visual Studio Code GitHub Copilot Chat बनाएं](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub मॉडल्स के साथ Phi-3.5 द्वारा अपना Visual Studio Code चैट कोपिलट एजेंट बनाएं](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - एडवांस्ड रीजनिंग सैंपल्स - - Phi-4 सैंपल्स - - [📓] [Phi-4-mini-रीज़निंग या Phi-4-रीज़निंग सैंपल्स](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Microsoft Olive के साथ Phi-4-mini-रीज़निंग का फाइन-ट्यूनिंग](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Apple MLX के साथ Phi-4-mini-रीज़निंग का फाइन-ट्यूनिंग](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [GitHub मॉडल्स के साथ Phi-4-mini-रीज़निंग](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Microsoft Foundry मॉडल्स के साथ Phi-4-mini-रीज़निंग](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [Phi-3.5-Instruct ONNX के साथ Windows GPU का उपयोग कर Prompt flow समाधान बनाना](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [माइक्रोसॉफ्ट Phi-3.5 tflite का उपयोग करके Android ऐप बनाना](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Q&A .NET उदाहरण स्थानीय ONNX Phi-3 मॉडल का उपयोग करते हुए Microsoft.ML.OnnxRuntime के साथ](../../md/04.HOL/dotnet/src/LabsPhi301) + - [कंसोल चैट .NET ऐप Semantic Kernel और Phi-3 के साथ](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI Inference SDK कोड आधारित नमूने + - Phi-4 नमूने + - [📓] [Phi-4-multimodal का उपयोग करके प्रोजेक्ट कोड जनरेट करें](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Phi-3 / 3.5 नमूने + - [Microsoft Phi-3 फैमिली के साथ अपना खुद का Visual Studio Code GitHub Copilot चैट बनाएं](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub मॉडल के साथ Phi-3.5 द्वारा अपना खुद का Visual Studio Code चैट कॉपिलॉट एजेंट बनाएं](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - एडवांस्ड रीज़निंग नमूने + - Phi-4 नमूने + - [📓] [Phi-4-mini-रीजनिंग या Phi-4-रीजनिंग नमूने](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [माइक्रोसॉफ्ट Olive के साथ Phi-4-mini-रीजनिंग का फाइन-ट्यूनिंग](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLX के साथ Phi-4-mini-रीजनिंग का फाइन-ट्यूनिंग](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [GitHub मॉडल के साथ Phi-4-mini-रीजनिंग](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Microsoft Foundry मॉडल के साथ Phi-4-mini-रीजनिंग](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - डेमोस - - [Phi-4-mini डेमोस जो Hugging Face Spaces पर होस्ट किए गए हैं](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal डेमोस जो Hugginge Face Spaces पर होस्ट किए गए हैं](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - विज़न सैंपल्स - - Phi-4 सैंपल्स - - [📓] [Phi-4-multimodal का उपयोग करके इमेज पढ़ें और कोड जनरेट करें](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 सैंपल्स - - [📓][Phi-3-विजन-इमेज टेक्स्ट से टेक्स्ट](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-विजन-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-विजन CLIP एम्बेडिंग](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [डेमो: Phi-3 रीसायक्लिंग](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-विजन - विज़ुअल भाषा सहायक - Phi3-विजन और OpenVINO के साथ](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Hugging Face Spaces पर होस्ट किए गए Phi-4-mini डेमोस](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Hugginge Face Spaces पर होस्ट किए गए Phi-4-multimodal डेमोस](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - विज़न नमूने + - Phi-4 नमूने + - [📓] [छवियाँ पढ़ने और कोड जेनरेट करने के लिए Phi-4-multimodal का उपयोग करें](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 नमूने + - [📓][Phi-3-vision-इमेज टेक्स्ट से टेक्स्ट](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vision CLIP एम्बेडिंग](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [डेमो: Phi-3 री-साइक्लिंग](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - विज़ुअल भाषा सहायक - Phi3-विज़न और OpenVINO के साथ](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 विज़न Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 विज़न OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 विज़न मल्टी-फ्रेम या मल्टी-इमेज सैंपल](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Microsoft.ML.OnnxRuntime .NET का उपयोग करके Phi-3 विज़न लोकल ONNX मॉडल](../../md/04.HOL/dotnet/src/LabsPhi303) - - [मेन्यू आधारित Phi-3 विज़न लोकल ONNX मॉडल Microsoft.ML.OnnxRuntime .NET के साथ](../../md/04.HOL/dotnet/src/LabsPhi304) - - - रीजनिंग-विजन सैंपल्स - - Phi-4-रीजनिंग-विजन-15B - - [📓] [Phi-4-रीजनिंग-विजन-15B का उपयोग करते हुए जायकिंग डिटेक्ट करना](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Phi-4-रीजनिंग-विजन-15B का उपयोग करते हुए गणना करना](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-रीजनिंग-विजन-15B का उपयोग करते हुए UI डिटेक्ट करना](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - मैथ सैंपल्स - - Phi-4-मिनी-फ्लैश-रीजनिंग-इंस्ट्रक्ट सैंपल्स [Phi-4-मिनी-फ्लैश-रीजनिंग-इंस्ट्रक्ट के साथ मैथ डेमो](./md/02.Application/09.Math/MathDemo.ipynb) - - - ऑडियो सैंपल्स - - Phi-4 सैंपल्स - - [📓] [Phi-4-multimodal का उपयोग करके ऑडियो ट्रांस्क्रिप्ट निकालना](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal ऑडियो सैंपल](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal भाषण अनुवाद सैंपल](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET कंसोल एप्लिकेशन जो Phi-4-multimodal ऑडियो का उपयोग करके ऑडियो फ़ाइल का विश्लेषण करता है और ट्रांस्क्रिप्ट जेनरेट करता है](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - MOE सैंपल्स - - Phi-3 / 3.5 सैंपल्स - - [📓] [Phi-3.5 Mixture of Experts Models (MoEs) सोशल मीडिया सैंपल](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search, और LlamaIndex के साथ एक Retrieval-Augmented Generation (RAG) पाइपलाइन बनाना](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓][Phi-3.5 विज़न मल्टी-फ्रेम या मल्टी-इमेज नमूना](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 विज़न लोकल ONNX मॉडल Microsoft.ML.OnnxRuntime .NET का उपयोग करके](../../md/04.HOL/dotnet/src/LabsPhi303) + - [मेनू आधारित Phi-3 विज़न लोकल ONNX मॉडल Microsoft.ML.OnnxRuntime .NET का उपयोग करके](../../md/04.HOL/dotnet/src/LabsPhi304) + + - रीज़निंग-विज़न नमूने + - Phi-4-रीज़निंग-विज़न-15B + - [📓] [Jaywalking का पता लगाने के लिए Phi-4-रीज़निंग-विज़न-15B का उपयोग](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [गणित के लिए Phi-4-रीज़निंग-विज़न-15B का उपयोग](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [UI का पता लगाने के लिए Phi-4-रीज़निंग-विज़न-15B का उपयोग](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - गणित नमूने + - Phi-4-मिनी-फ्लैश-रीज़निंग-इंस्ट्रक्ट नमूने [Phi-4-मिनी-फ्लैश-रीज़निंग-इंस्ट्रक्ट के साथ गणित डेमो](./md/02.Application/09.Math/MathDemo.ipynb) + + - ऑडियो नमूने + - Phi-4 नमूने + - [📓] [Phi-4-multimodal का उपयोग करके ऑडियो ट्रांसक्रिप्ट निकालना](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal ऑडियो नमूना](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal स्पीच अनुवाद नमूना](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET कंसोल एप्लिकेशन जो Phi-4-multimodal का उपयोग करके ऑडियो फ़ाइल का विश्लेषण करता है और ट्रांसक्रिप्ट जनरेट करता है](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE नमूने + - Phi-3 / 3.5 नमूने + - [📓] [Phi-3.5 एक्सपर्ट्स मॉडल (MoEs) सोशल मीडिया नमूना](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search, और LlamaIndex के साथ Retrieval-Augmented Generation (RAG) पाइपलाइन बनाना](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - फंक्शन कॉलिंग सैंपल्स - - Phi-4 सैंपल्स 🆕 - - [📓] [Phi-4-mini के साथ फंक्शन कॉलिंग का उपयोग](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Phi-4-mini के साथ मल्टी-एजेंट्स बनाने के लिए फंक्शन कॉलिंग का उपयोग](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Ollama के साथ फंक्शन कॉलिंग का उपयोग](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [ONNX के साथ फंक्शन कॉलिंग का उपयोग](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - मल्टीमॉडल मिक्सिंग सैंपल्स - - Phi-4 सैंपल्स 🆕 - - [📓] [Phi-4-multimodal का उपयोग एक टेक्नोलॉजी पत्रकार के रूप में](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET कंसोल एप्लिकेशन जो Phi-4-multimodal का उपयोग करके इमेज का विश्लेषण करता है](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Phi फाइन-ट्यूनिंग सैंपल्स + - फंक्शन कॉलिंग नमूने + - Phi-4 नमूने 🆕 + - [📓] [Phi-4-mini के साथ फंक्शन कॉलिंग का उपयोग करना](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Phi-4-mini के साथ मल्टी-एजेंट बनाने के लिए फंक्शन कॉलिंग का उपयोग करना](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Ollama के साथ फंक्शन कॉलिंग का उपयोग करना](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [ONNX के साथ फंक्शन कॉलिंग का उपयोग करना](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - मल्टीमॉडल मिक्सिंग नमूने + - Phi-4 नमूने 🆕 + - [📓] [Phi-4-multimodal को तकनीकी पत्रकार के रूप में उपयोग करना](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET कंसोल एप्लिकेशन जो Phi-4-multimodal का उपयोग करके छवियों का विश्लेषण करता है](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Phi फाइन-ट्यूनिंग नमूने - [फाइन-ट्यूनिंग परिदृश्य](./md/03.FineTuning/FineTuning_Scenarios.md) - [फाइन-ट्यूनिंग बनाम RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Phi-3 को एक उद्योग विशेषज्ञ बनने दें फाइन-ट्यूनिंग](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [VS कोड के लिए AI टूलकिट के साथ Phi-3 का फाइन-ट्यूनिंग](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Azure मशीन लर्निंग सेवा के साथ Phi-3 का फाइन-ट्यूनिंग](./md/03.FineTuning/Introduce_AzureML.md) - - [Lora के साथ Phi-3 का फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora के साथ Phi-3 का फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_Qlora.md) - - [Microsoft Foundry के साथ Phi-3 का फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Azure ML CLI/SDK के साथ Phi-3 का फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Microsoft Olive के साथ फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive Hands-On Lab के साथ फाइन-ट्यूनिंग](./md/03.FineTuning/olive-lab/readme.md) - - [Weights and Bias के साथ Phi-3-vision का फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Phi-3 को उद्योग विशेषज्ञ बनाएं फाइन-ट्यूनिंग](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [VS कोड के लिए AI टूलकिट के साथ Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure मशीन लर्निंग सेवा के साथ Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/Introduce_AzureML.md) + - [Lora के साथ Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_Lora.md) + - [QLora के साथ Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_Qlora.md) + - [माइक्रोसॉफ्ट फाउंड्री के साथ Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Azure ML CLI/SDK के साथ Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_MLSDK.md) + - [माइक्रोसॉफ्ट Olive के साथ फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [माइक्रोसॉफ्ट Olive हैंड्स-ऑन लैब के साथ फाइन-ट्यूनिंग](./md/03.FineTuning/olive-lab/readme.md) + - [Weights and Bias के साथ Phi-3-vision फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Apple MLX Framework के साथ Phi-3 का फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_MLX.md) - [Phi-3-vision का फाइन-ट्यूनिंग (आधिकारिक समर्थन)](./md/03.FineTuning/FineTuning_Vision.md) - - [काइटो AKS के साथ Phi-3 का फाइन-ट्यूनिंग, Azure कंटेनर (आधिकारिक समर्थन)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Phi-3 और 3.5 विजन का फाइन-ट्यूनिंग](https://github.com/2U1/Phi3-Vision-Finetune) + - [Kaito AKS, Azure Containers के साथ Phi-3 का फाइन-ट्यूनिंग (आधिकारिक समर्थन)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Phi-3 और 3.5 Vision का फाइन-ट्यूनिंग](https://github.com/2U1/Phi3-Vision-Finetune) - हैंड्स ऑन लैब - - [प्रगतिशील मॉडलों का अन्वेषण: LLMs, SLMs, स्थानीय विकास और भी बहुत कुछ](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP की क्षमता का अनलॉकिंग: Microsoft Olive के साथ फाइन-ट्यूनिंग](https://github.com/azure/Ignite_FineTuning_workshop) + - [अत्याधुनिक मॉडलों का अन्वेषण: LLMs, SLMs, स्थानीय विकास और अधिक](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP क्षमता को अनलॉक करना: Microsoft Olive के साथ फाइन-ट्यूनिंग](https://github.com/azure/Ignite_FineTuning_workshop) - शैक्षणिक शोध पत्र और प्रकाशन - [Textbooks Are All You Need II: phi-1.5 तकनीकी रिपोर्ट](https://arxiv.org/abs/2309.05463) - - [Phi-3 तकनीकी रिपोर्ट: एक अत्यंत सक्षम भाषा मॉडल आपके फोन पर स्थानीय](https://arxiv.org/abs/2404.14219) + - [Phi-3 तकनीकी रिपोर्ट: आपके फोन पर एक अत्यंत सक्षम भाषा मॉडल](https://arxiv.org/abs/2404.14219) - [Phi-4 तकनीकी रिपोर्ट](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini तकनीकी रिपोर्ट: मिश्रण-ऑफ़-LoRAs के माध्यम से कॉम्पैक्ट फिर भी शक्तिशाली मल्टीमॉडल भाषा मॉडल](https://arxiv.org/abs/2503.01743) - - [वाहन के अंदर फ़ंक्शन-कॉलिंग के लिए छोटे भाषा मॉडलों का अनुकूलन](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) बहुविकल्पीय प्रश्न उत्तर देने के लिए PHI-3 का फाइन-ट्यूनिंग: कार्यप्रणाली, परिणाम और चुनौतियां](https://arxiv.org/abs/2501.01588) + - [Phi-4-Mini तकनीकी रिपोर्ट: मिश्रण-ऑफ़-LoRAs के माध्यम से कॉम्पैक्ट फिर भी शक्तिशाली मल्टिमॉडल भाषा मॉडल](https://arxiv.org/abs/2503.01743) + - [वाहन में फ़ंक्शन-कॉलिंग के लिए छोटे भाषा मॉडलों का अनुकूलन](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) बहुविकल्पी प्रश्नोत्तरी के लिए PHI-3 का फाइन-ट्यूनिंग: पद्धति, परिणाम, और चुनौतियाँ](https://arxiv.org/abs/2501.01588) - [Phi-4-तर्क तकनीकी रिपोर्ट](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Phi-4-मिनी-तर्क तकनीकी रिपोर्ट](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Phi मॉडल का उपयोग करना +## Phi मॉडलों का उपयोग ### Microsoft Foundry पर Phi -आप Microsoft Phi का उपयोग कैसे करें और अपने विभिन्न हार्डवेयर उपकरणों में E2E समाधान कैसे बनाएं, यह सीख सकते हैं। Phi का अनुभव करने के लिए, मॉडल के साथ खेलने और अपने परिदृश्यों के लिए Phi को अनुकूलित करने से शुरू करें [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) का उपयोग कर सकते हैं। आप [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) के साथ शुरू करना सीख सकते हैं। +आप सीख सकते हैं कि Microsoft Phi का उपयोग कैसे करें और अपने विभिन्न हार्डवेयर उपकरणों में E2E समाधान कैसे बनाएं। खुद Phi का अनुभव करने के लिए, मॉडल के साथ खेलना शुरू करें और अपने परिदृश्यों के लिए Phi को अनुकूलित करें [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) का उपयोग करके। आप अधिक जान सकते हैं [Microsoft Foundry के साथ शुरू करना](/md/02.QuickStart/AzureAIFoundry_QuickStart.md)। -**प्लेस्टेशन** -प्रत्येक मॉडल के पास उसके परीक्षण के लिए समर्पित प्लेग्राउंड है [Azure AI Playground](https://aka.ms/try-phi3)। +**प्लेलैंड** +प्रत्येक मॉडल का परीक्षण करने के लिए एक समर्पित प्लेलैंड है [Azure AI Playground](https://aka.ms/try-phi3). -### GitHub Models पर Phi +### गिटहब मॉडल पर Phi -आप Microsoft Phi का उपयोग कैसे करें और अपने विभिन्न हार्डवेयर उपकरणों में E2E समाधान कैसे बनाएं, यह सीख सकते हैं। Phi का अनुभव करने के लिए, मॉडल के साथ खेलने और अपने परिदृश्यों के लिए Phi को अनुकूलित करने से शुरू करें [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) का उपयोग कर सकते हैं। आप [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) के साथ शुरू करना सीख सकते हैं। +आप सीख सकते हैं कि Microsoft Phi का उपयोग कैसे करें और अपने विभिन्न हार्डवेयर उपकरणों में E2E समाधान कैसे बनाएं। खुद Phi का अनुभव करने के लिए, मॉडल के साथ खेलना शुरू करें और अपने परिदृश्यों के लिए Phi को अनुकूलित करें [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) का उपयोग करके। आप अधिक जान सकते हैं [GitHub Model Catalog के साथ शुरू करना](/md/02.QuickStart/GitHubModel_QuickStart.md)। -**प्लेस्टेशन** -प्रत्येक मॉडल के लिए समर्पित परीक्षण [प्लेग्राउंड](/md/02.QuickStart/GitHubModel_QuickStart.md) उपलब्ध है। +**प्लेलैंड** +प्रत्येक मॉडल का परीक्षण करने के लिए एक समर्पित [प्लेलैंड](/md/02.QuickStart/GitHubModel_QuickStart.md) है। ### Hugging Face पर Phi -आप यह मॉडल [Hugging Face](https://huggingface.co/microsoft) पर भी पा सकते हैं। +आप मॉडल को [Hugging Face](https://huggingface.co/microsoft) पर भी पा सकते हैं। -**प्लेस्टेशन** -[Hugging Chat प्लेग्राउंड](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**प्लेलैंड** + [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 अन्य कोर्स + ## 🎒 अन्य पाठ्यक्रम -हमारी टीम अन्य कोर्स भी बनाती है! देखें: +हमारी टीम अन्य पाठ्यक्रम भी बनाती है! देखें: ### LangChain @@ -265,54 +268,56 @@ Phi वर्तमान में सबसे शक्तिशाली औ --- ### जनरेटिव AI श्रृंखला -[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +[![जनरेटिव AI शुरुआती के लिए](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![जनरेटिव AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![जनरेटिव AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![जनरेटिव AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- ### कोर लर्निंग -[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![साइबरसुरक्षा शुरुआती के लिए](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web Dev शुरुआती के लिए](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT शुरुआती के लिए](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR Development शुरुआती के लिए](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![शुरुआती के लिए ML](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![शुरुआती के लिए डेटा साइंस](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![शुरुआती के लिए AI](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![शुरुआती के लिए साइबरसिक्योरिटी](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![शुरुआती के लिए वेब डेवलपमेंट](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![शुरुआती के लिए IoT](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![शुरुआती के लिए XR डेवलपमेंट](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### कॉपिलट श्रृंखला -[![AI जोड़ीदार प्रोग्रामिंग के लिए कॉपिलट](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![C#/.NET के लिए कॉपिलट](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![कॉपिलट एडवेंचर](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +### कोपाइलट श्रृंखला +[![AI साथ प्रोग्रामिंग के लिए कोपाइलट](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![C#/.NET के लिए कोपाइलट](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![कोपाइलट एडवेंचर](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## ज़िम्मेदार AI +## उत्तरदायी AI + +Microsoft हमारे ग्राहकों को हमारे AI उत्पादों को जिम्मेदारी से उपयोग करने में मदद करने, हमारे अनुभव साझा करने, और ट्रांसपेरेंसी नोट्स और प्रभाव आकलनों जैसे उपकरणों के माध्यम से विश्वास-आधारित साझेदारी बनाने के लिए प्रतिबद्ध है। ये संसाधन [https://aka.ms/RAI](https://aka.ms/RAI) पर पाए जा सकते हैं। +Microsoft का उत्तरदायी AI के लिए दृष्टिकोण हमारे AI के सिद्धांतों पर आधारित है: निष्पक्षता, विश्वसनीयता और सुरक्षा, गोपनीयता और सुरक्षा, समावेशन, पारदर्शिता, और जवाबदेही। + +बड़े पैमाने पर प्राकृतिक भाषा, छवि और भाषण मॉडल - जैसे इस उदाहरण में उपयोग किए गए - संभावित रूप से गैर-निष्पक्ष, अविश्वसनीय, या अपमानजनक व्यवहार कर सकते हैं, जो नुकसान पहुंचा सकते हैं। कृपया जोखिमों और सीमाओं के बारे में जानकारी के लिए [Azure OpenAI सेवा ट्रांसपेरेंसी नोट](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) देखें। -Microsoft हमारे ग्राहकों को हमारे AI उत्पादों का जिम्मेदारी से उपयोग करने में सहायता करने के लिए प्रतिबद्ध है, हमारे अनुभव साझा करता है, और पारदर्शिता नोट्स और प्रभाव आकलनों जैसे उपकरणों के माध्यम से भरोसेमंद साझेदारी बनाता है। इन संसाधनों में से कई [https://aka.ms/RAI](https://aka.ms/RAI) पर पाए जा सकते हैं। -Microsoft का जिम्मेदार AI के लिए दृष्टिकोण हमारे AI सिद्धांतों पर आधारित है: निष्पक्षता, विश्वसनीयता और सुरक्षा, गोपनीयता और सुरक्षा, समावेशन, पारदर्शिता, और जवाबदेही। -विस्तृत पैमाने पर प्राकृतिक भाषा, छवि, और भाषण मॉडल - जैसे इस नमूने में उपयोग किए गए - संभावित रूप से अनुचित, अविश्वसनीय, या अपमानजनक व्यवहार कर सकते हैं, जिससे हानि हो सकती है। कृपया जोखिमों और सीमाओं के बारे में सूचित रहने के लिए [Azure OpenAI सेवा पारदर्शिता नोट](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) देखें। -इन जोखिमों को कम करने के लिए अनुशंसित तरीका यह है कि आपकी आर्किटेक्चर में एक सुरक्षा प्रणाली शामिल की जाए जो हानिकारक व्यवहार का पता लगा सके और उसे रोक सके। [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) एक स्वतंत्र सुरक्षा परत प्रदान करता है, जो अनुप्रयोगों और सेवाओं में हानिकारक उपयोगकर्ता-जनित और AI-जनित सामग्री का पता लगाने में सक्षम है। Azure AI Content Safety में टेक्स्ट और इमेज API शामिल हैं जो आपको हानिकारक सामग्री का पता लगाने की अनुमति देते हैं। Microsoft Foundry के भीतर, Content Safety सेवा आपको विभिन्न माध्यमों में हानिकारक सामग्री का पता लगाने के लिए नमूना कोड देखने, एक्सप्लोर करने और आज़माने की सुविधा प्रदान करती है। निम्नलिखित [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) आपको सेवा को अनुरोध भेजने में मार्गदर्शन करती है। +इन जोखिमों को कम करने के लिए अनुशंसित तरीका आपके आर्किटेक्चर में एक सुरक्षा प्रणाली शामिल करना है जो हानिकारक व्यवहार का पता लगा सके और उसे रोक सके। [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) एक स्वतंत्र सुरक्षा परत प्रदान करता है, जो अनुप्रयोगों और सेवाओं में हानिकारक उपयोगकर्ता-जनित और AI-जनित सामग्री का पता लगाने में सक्षम है। Azure AI Content Safety में टेक्स्ट और छवि API शामिल हैं जो आपको हानिकारक सामग्री का पता लगाने की अनुमति देते हैं। Microsoft Foundry के भीतर, Content Safety सेवा आपको विभिन्न माध्यमों में हानिकारक कंटेंट का पता लगाने के लिए नमूना कोड देखने, एक्स्प्लोर करने और आज़माने की अनुमति देती है। निम्नलिखित [क्विकस्टार्ट डॉक्यूमेंटेशन](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) आपको सेवा से अनुरोध करने के लिए मार्गदर्शन करता है। -एक अन्य पहलू जिसे ध्यान में रखना आवश्यक है, वह है समग्र आवेदन प्रदर्शन। मल्टी-मॉडल और मल्टी-मॉडल्स एप्लिकेशन के साथ, हम प्रदर्शन को इस रूप में देखते हैं कि प्रणाली आपकी और आपके उपयोगकर्ताओं की अपेक्षा के अनुसार कार्य करती है, जिसमें हानिकारक आउटपुट उत्पन्न न करना भी शामिल है। आपके पूरे आवेदन के प्रदर्शन का मूल्यांकन करना महत्वपूर्ण है, इसके लिए [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) का उपयोग करें। आपके पास [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) बनाने और उनका मूल्यांकन करने की भी क्षमता है। +एक अन्य पहलू जिसे ध्यान में रखना है वह है समग्र एप्लिकेशन प्रदर्शन। मल्टी-मोडल और मल्टी-मॉडल अनुप्रयोगों के साथ, प्रदर्शन का मतलब है कि सिस्टम आपकी और आपके उपयोगकर्ताओं की उम्मीद के अनुसार कार्य करता है, जिसमें हानिकारक आउटपुट न उत्पन्न करना भी शामिल है। अपने समग्र एप्लिकेशन के प्रदर्शन का आकलन करना महत्वपूर्ण है इसके लिए [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) का उपयोग करें। आपके पास [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) के साथ बनाने और मूल्यांकन करने की भी क्षमता है। -आप अपने विकास पर्यावरण में [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) का उपयोग करके अपने AI आवेदन का मूल्यांकन कर सकते हैं। चाहे आपके पास एक परीक्षण डेटासेट हो या कोई लक्ष्य, आपके जनरेटिव AI आवेदन के जनरेशन को बिल्ट-इन एवालुएटर्स या आपकी पसंद के कस्टम इवालुएटर्स के साथ मात्रात्मक रूप से मापा जाता है। अपने सिस्टम का मूल्यांकन करने के लिए azure ai evaluation sdk के साथ आरंभ करने के लिए, आप [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) का पालन कर सकते हैं। एक बार जब आप एक मूल्यांकन रन निष्पादित करते हैं, तो आप [Microsoft Foundry में परिणामों का विज़ुअलीकरण कर सकते हैं](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)। +आप अपने विकास पर्यावरण में [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) का उपयोग करके अपने AI एप्लिकेशन का मूल्यांकन कर सकते हैं। चाहे आपके पास परीक्षण डेटा सेट हो या लक्ष्य, आपके जनरेटिव AI एप्लिकेशन के निर्माणों को बिल्ट-इन मूल्यांकनकर्ताओं या आपकी पसंद के कस्टम मूल्यांकनकर्ताओं के साथ मात्रात्मक रूप से मापा जाता है। अपने सिस्टम का मूल्यांकन करने के लिए azure ai evaluation sdk के साथ शुरुआत करने के लिए, आप [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) का पालन कर सकते हैं। एक बार जब आप मूल्यांकन रन निष्पादित कर लेते हैं, तो आप [Microsoft Foundry में परिणामों का विज़ुअलाइज़ेशन](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) कर सकते हैं। ## ट्रेडमार्क -यह परियोजना परियोजनाओं, उत्पादों, या सेवाओं के ट्रेडमार्क या लोगो शामिल कर सकती है। माइक्रोसॉफ्ट ट्रेडमार्क या लोगो के अधिकृत उपयोग को [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) का पालन करना आवश्यक है। -माइक्रोसॉफ्ट ट्रेडमार्क या लोगो का इस परियोजना के परिवर्तित संस्करणों में उपयोग भ्रम पैदा नहीं करना चाहिए या माइक्रोसॉफ्ट प्रायोजन का संकेत नहीं देना चाहिए। किसी भी तृतीय-पक्ष ट्रेडमार्क या लोगो का उपयोग तृतीय-पक्ष की नीतियों के अनुसार होगा। +इस परियोजना में किसी प्रोजेक्ट, उत्पाद, या सेवा के ट्रेडमार्क या लोगो हो सकते हैं। Microsoft ट्रेडमार्क या लोगो के अधिकृत उपयोग के लिए [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) का पालन करना आवश्यक है। +इस परियोजना के संशोधित संस्करणों में Microsoft ट्रेडमार्क या लोगो का उपयोग भ्रमित नहीं करना चाहिए या Microsoft प्रायोजन का संकेत नहीं देना चाहिए। किसी तीसरे पक्ष के ट्रेडमार्क या लोगो का उपयोग उन तीसरे पक्ष की नीतियों के अधीन है। ## सहायता प्राप्त करना -यदि आप फंस जाते हैं या AI ऐप्स बनाने के संबंध में कोई प्रश्न है, तो जुड़ें: +यदि आप फंस जाते हैं या AI ऐप्स बनाने के बारे में कोई प्रश्न है, तो जुड़ें: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -यदि आपके पास उत्पाद प्रतिक्रिया है या निर्माण के दौरान त्रुटियाँ होती हैं तो देखें: +यदि आपके पास उत्पाद प्रतिक्रिया है या निर्माण के दौरान कोई त्रुटि आई है तो जाएं: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) @@ -320,5 +325,5 @@ Microsoft का जिम्मेदार AI के लिए दृष्ट **अस्वीकरण**: -इस दस्तावेज़ का अनुवाद एआई अनुवाद सेवा [Co-op Translator](https://github.com/Azure/co-op-translator) का उपयोग करके किया गया है। जबकि हम सटीकता के लिए प्रयासरत हैं, कृपया ध्यान रखें कि स्वचालित अनुवादों में त्रुटियाँ या गलतियाँ हो सकती हैं। मूल दस्तावेज़ उसकी स्वदेशी भाषा में ही आधिकारिक स्रोत माना जाना चाहिए। महत्वपूर्ण जानकारी के लिए, पेशेवर मानव अनुवाद की सिफारिश की जाती है। इस अनुवाद के उपयोग से उत्पन्न किसी भी गलतफहमी या गलत व्याख्या के लिए हम जिम्मेदार नहीं हैं। +इस दस्तावेज़ का अनुवाद AI अनुवाद सेवा [Co-op Translator](https://github.com/Azure/co-op-translator) का उपयोग करके किया गया है। जबकि हम सटीकता के लिए प्रयास करते हैं, कृपया ध्यान दें कि स्वचालित अनुवादों में त्रुटियाँ या अशुद्धियाँ हो सकती हैं। मूल दस्तावेज़ अपनी मूल भाषा में ही प्रामाणिक स्रोत माना जाना चाहिए। महत्वपूर्ण जानकारी के लिए, पेशेवर मानव अनुवाद की सिफारिश की जाती है। इस अनुवाद के उपयोग से उत्पन्न किसी भी गलतफहमी या गलत व्याख्या के लिए हम उत्तरदायी नहीं हैं। \ No newline at end of file diff --git a/translations/hi/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/hi/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..171c6c1d2 --- /dev/null +++ b/translations/hi/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Microsoft phi-4-mini & SQLite FTS5 के साथ हाइब्रिड RAG (ज़ीरो-क्लाउड SLM)\n", + "\n", + "> **लेखक:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **ध्यान केंद्रित:** छोटे भाषा मॉडल (SLMs), SQLite FTS5 BM25, डेंस एम्बेडिंग्स, रेसिप्रोकल रैंक फ्यूजन (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. प्रेरणा: स्थानीय SLMs में कीवर्ड रिकॉल दुविधा\n", + "केवल डेंस वेक्टर एम्बेडिंग्स पर आधारित मानक RAG आर्किटेक्चर अक्सर सटीक संख्यात्मक टोकन (जैसे, `2,340,000 TL`, अनुबंध कोड, खाता नंबर) पुनः प्राप्त करने में विफल रहते हैं। \n", + "इसके विपरीत, विरल लेक्सिकल खोज (BM25) सार्थक पर्यायवाची और परिभाषित प्रश्नों को चूक जाती है।\n", + "\n", + "यह कुकबुक प्रदर्शित करता है कि कैसे एक **उच्च-गति, इन-मेमोरी हाइब्रिड रिट्रीवल इंजन** को लागू किया जाए जिसमें सम्मिलित हैं:\n", + "1. **डेंस वेक्टर** (कोसाइन समानता)\n", + "2. **विरल लेक्सिकल खोज** (SQLite FTS5 BM25)\n", + "3. **रेसिप्रोकल रैंक फ्यूजन (RRF, $k=60$)**\n", + "4. Microsoft `phi-4-mini` के साथ **ग्राउंडेड उद्धरण उत्पादन (`[1]`, `[2]`)**।\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. डुअल SQLite स्कीमा (सघन वेक्टर + वर्चुअल FTS5 BM25 टेबल)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. नमूना इनगेस्टन और निष्पादन बेंचमार्क\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Microsoft phi-4-mini के लिए आधारभूत प्रॉम्प्ट फॉर्मूलेशन\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**अस्वीकरण**:\nइस दस्तावेज़ का अनुवाद AI अनुवाद सेवा [Co-op Translator](https://github.com/Azure/co-op-translator) का उपयोग करके किया गया है। जबकि हम सटीकता के लिए प्रयास करते हैं, कृपया ध्यान दें कि स्वचालित अनुवादों में त्रुटियाँ या अशुद्धियाँ हो सकती हैं। मूल दस्तावेज़ अपनी मूल भाषा में ही प्रामाणिक स्रोत माना जाना चाहिए। महत्वपूर्ण जानकारी के लिए, पेशेवर मानव अनुवाद की सिफारिश की जाती है। इस अनुवाद के उपयोग से उत्पन्न किसी भी गलतफहमी या गलत व्याख्या के लिए हम उत्तरदायी नहीं हैं।\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/hr/.co-op-translator.json b/translations/hr/.co-op-translator.json index 470843102..eb6321b8a 100644 --- a/translations/hr/.co-op-translator.json +++ b/translations/hr/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "hr" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:15:33+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:55:14+00:00", "source_file": "README.md", "language_code": "hr" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "hr" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:58:35+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "hr" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:32:20+00:00", diff --git a/translations/hr/README.md b/translations/hr/README.md index 793970ec5..690734a36 100644 --- a/translations/hr/README.md +++ b/translations/hr/README.md @@ -1,42 +1,42 @@ -# Phi Cookbook: Primjeri iz prakse s Microsoftovim Phi modelima +# Phi Kuharica: Praktični Primjeri s Microsoftovim Phi Modelima -[![Otvori i koristi uzorke u GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Otvori i koristi primjere u GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Otvori u Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub suradnici](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub issues](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub pull-requests](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub problemi](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub pull-zahtjevi](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Dobrodošli](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![GitHub pratitelji](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub forkovi](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub zvjezdice](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub zvijezde](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi je niz AI modela otvorenog koda koje je razvila tvrtka Microsoft. +Phi je niz open source AI modela razvijenih od strane Microsofta. -Phi je trenutno najsnažniji i isplativ mali jezični model (SLM), s vrlo dobrim rezultatima u više jezika, rezoniranju, generiranju teksta/četa, kodiranju, slikama, zvuku i drugim scenarijima. +Phi je trenutno najsnažniji i najisplativiji mali jezični model (SLM), s vrlo dobrim referentnim rezultatima u višjezičnosti, rezoniranju, generiranju teksta/čavrljanju, kodiranju, slikama, zvuku i drugim scenarijima. -Možete implementirati Phi u oblaku ili na rubnim uređajima, i lako izgraditi generativne AI aplikacije s ograničenom računalnom snagom. +Phi možete postaviti u oblak ili na edge uređaje, a lako možete graditi generativne AI aplikacije s ograničenom računalnom snagom. -Slijedite ove korake da biste započeli korištenje ovog resursa: +Slijedite ove korake da biste započeli s korištenjem ovih resursa: 1. **Forkajte repozitorij**: Kliknite [![GitHub forkovi](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Klonirajte repozitorij**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Pridružite se Microsoft AI Discord zajednici i upoznajte stručnjake i kolege programere**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +2. **Kloni repozitorij**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Pridružite se Microsoft AI Discord zajednici i upoznajte stručnjake i kolege developere**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/hr/cover.eb18d1b9605d754b.webp) -### 🌐 Podrška za više jezika +### 🌐 Višjezična podrška -#### Podržano putem GitHub akcije (Automatski i uvijek ažurirano) +#### Podržano putem GitHub Akcije (Automatski i uvijek ažurirano) -[Arapski](../ar/README.md) | [Bengalski](../bn/README.md) | [Bugarski](../bg/README.md) | [Burmêski (Myanmar)](../my/README.md) | [Kineski (pojednostavljeni)](../zh-CN/README.md) | [Kineski (tradicionalni, Hong Kong)](../zh-HK/README.md) | [Kineski (tradicionalni, Macau)](../zh-MO/README.md) | [Kineski (tradicionalni, Taiwan)](../zh-TW/README.md) | [Hrvatski](./README.md) | [Češki](../cs/README.md) | [Danski](../da/README.md) | [Nizozemski](../nl/README.md) | [Estonski](../et/README.md) | [Finski](../fi/README.md) | [Francuski](../fr/README.md) | [Njemački](../de/README.md) | [Grčki](../el/README.md) | [Hebrejski](../he/README.md) | [Hindi](../hi/README.md) | [Mađarski](../hu/README.md) | [Indonezijski](../id/README.md) | [Talijanski](../it/README.md) | [Japanski](../ja/README.md) | [Kanadški](../kn/README.md) | [Khmer](../km/README.md) | [Korejski](../ko/README.md) | [Litavski](../lt/README.md) | [Malajski](../ms/README.md) | [Malajalamski](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalski](../ne/README.md) | [Nigerijski Pidžin](../pcm/README.md) | [Norveški](../no/README.md) | [Perzijski (Farsi)](../fa/README.md) | [Poljski](../pl/README.md) | [Portugalski (Brazil)](../pt-BR/README.md) | [Portugalski (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumunjski](../ro/README.md) | [Ruski](../ru/README.md) | [Srpski (ćirilica)](../sr/README.md) | [Slovački](../sk/README.md) | [Slovenski](../sl/README.md) | [Španjolski](../es/README.md) | [Svahili](../sw/README.md) | [Švedski](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamilski](../ta/README.md) | [Telugu](../te/README.md) | [Tajlandski](../th/README.md) | [Turski](../tr/README.md) | [Ukrajinski](../uk/README.md) | [Urdu](../ur/README.md) | [Vijetnamski](../vi/README.md) +[Arapski](../ar/README.md) | [Bengalski](../bn/README.md) | [Bugarski](../bg/README.md) | [Burmanski (Mjanmar)](../my/README.md) | [Kineski (Pojednostavljeni)](../zh-CN/README.md) | [Kineski (Tradicionalni, Hong Kong)](../zh-HK/README.md) | [Kineski (Tradicionalni, Makao)](../zh-MO/README.md) | [Kineski (Tradicionalni, Tajvan)](../zh-TW/README.md) | [Hrvatski](./README.md) | [Češki](../cs/README.md) | [Danski](../da/README.md) | [Nizozemski](../nl/README.md) | [Estonski](../et/README.md) | [Finski](../fi/README.md) | [Francuski](../fr/README.md) | [Njemački](../de/README.md) | [Grčki](../el/README.md) | [Hebrejski](../he/README.md) | [Hindi](../hi/README.md) | [Mađarski](../hu/README.md) | [Indonezijski](../id/README.md) | [Talijanski](../it/README.md) | [Japanski](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korejski](../ko/README.md) | [Litvanski](../lt/README.md) | [Malajski](../ms/README.md) | [Malajalamski](../ml/README.md) | [Marathijski](../mr/README.md) | [Nepalski](../ne/README.md) | [Nigerijski Pidžin](../pcm/README.md) | [Norveški](../no/README.md) | [Perzijski (Farsi)](../fa/README.md) | [Poljski](../pl/README.md) | [Portugalski (Brazil)](../pt-BR/README.md) | [Portugalski (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumunjski](../ro/README.md) | [Ruski](../ru/README.md) | [Srpski (ćirilica)](../sr/README.md) | [Slovački](../sk/README.md) | [Slovenski](../sl/README.md) | [Španjolski](../es/README.md) | [Svahili](../sw/README.md) | [Švedski](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamilski](../ta/README.md) | [Telugu](../te/README.md) | [Tajlandski](../th/README.md) | [Turski](../tr/README.md) | [Ukrajinski](../uk/README.md) | [Urdu](../ur/README.md) | [Vijetnamski](../vi/README.md) -> **Radije klonirati lokalno?** +> **Radije lokalno klonirati?** > -> Ovaj repozitorij uključuje 50+ prijevoda jezika što značajno povećava veličinu preuzimanja. Za kloniranje bez prijevoda, upotrijebite sparse checkout: +> Ovaj repozitorij sadrži 50+ prijevoda jezika što znatno povećava veličinu preuzimanja. Da biste klonirali bez prijevoda, koristite sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,273 +52,278 @@ Slijedite ove korake da biste započeli korištenje ovog resursa: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> To vam daje sve što vam treba za završetak tečaja uz mnogo brže preuzimanje. +> Ovo vam daje sve što vam treba za dovršetak tečaja s puno bržim preuzimanjem. ## Sadržaj - Uvod - - [Dobrodošli u Phi obitelj](./md/01.Introduction/01/01.PhiFamily.md) + - [Dobrodošli u Phi Obitelj](./md/01.Introduction/01/01.PhiFamily.md) - [Postavljanje vašeg okruženja](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Razumijevanje ključnih tehnologija](./md/01.Introduction/01/01.Understandingtech.md) - - [Sigurnost AI za Phi modele](./md/01.Introduction/01/01.AISafety.md) - - [Podrška za Phi hardver](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi modeli i dostupnost na različitim platformama](./md/01.Introduction/01/01.Edgeandcloud.md) + - [AI Sigurnost za Phi modele](./md/01.Introduction/01/01.AISafety.md) + - [Phi hardverska podrška](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Phi modeli i dostupnost na platformama](./md/01.Introduction/01/01.Edgeandcloud.md) - [Korištenje Guidance-ai i Phi](./md/01.Introduction/01/01.Guidance.md) - [GitHub Marketplace modeli](https://github.com/marketplace/models) - [Azure AI katalog modela](https://ai.azure.com) -- Inferencija Phi u različitim okruženjima +- Izvođenje Phi u različitim okruženjima - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub modeli](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry katalog modela](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry lokalno](./md/01.Introduction/02/07.FoundryLocal.md) - -- Inferencija Phi obitelji - - [Inferencija Phi na iOS-u](./md/01.Introduction/03/iOS_Inference.md) - - [Inferencija Phi na Androidu](./md/01.Introduction/03/Android_Inference.md) - - [Inferencija Phi na Jetsonu](./md/01.Introduction/03/Jetson_Inference.md) - - [Inferencija Phi na AI PC-u](./md/01.Introduction/03/AIPC_Inference.md) - - [Inferencija Phi s Apple MLX Frameworkom](./md/01.Introduction/03/MLX_Inference.md) - - [Inferencija Phi na lokalnom poslužitelju](./md/01.Introduction/03/Local_Server_Inference.md) - - [Inferencija Phi na udaljenom poslužitelju koristeći AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Inferencija Phi s Rust-om](./md/01.Introduction/03/Rust_Inference.md) - - [Inferencija Phi--Vizija lokalno](./md/01.Introduction/03/Vision_Inference.md) - - [Inferencija Phi s Kaito AKS, Azure Containers (službena podrška)](./md/01.Introduction/03/Kaito_Inference.md) -- [Kvantificiranje Phi obitelji](./md/01.Introduction/04/QuantifyingPhi.md) - - [Kvantiziranje Phi-3.5 / 4 koristeći llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Kvantiziranje Phi-3.5 / 4 koristeći Generative AI ekstenzije za onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Kvantiziranje Phi-3.5 / 4 koristeći Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Kvantiziranje Phi-3.5 / 4 koristeći Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) + +- Izvođenje Phi obitelji + - [Izvođenje Phi na iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Izvođenje Phi na Android](./md/01.Introduction/03/Android_Inference.md) + - [Izvođenje Phi na Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Izvođenje Phi na AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Izvođenje Phi s Apple MLX frameworkom](./md/01.Introduction/03/MLX_Inference.md) + - [Izvođenje Phi na lokalnom serveru](./md/01.Introduction/03/Local_Server_Inference.md) + - [Izvođenje Phi na udaljenom serveru koristeći AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Izvođenje Phi s Rustom](./md/01.Introduction/03/Rust_Inference.md) + - [Izvođenje Phi – Vizija lokalno](./md/01.Introduction/03/Vision_Inference.md) + - [Izvođenje Phi s Kaito AKS, Azure Containers (službena podrška)](./md/01.Introduction/03/Kaito_Inference.md) +- [Quantificiranje Phi obitelji](./md/01.Introduction/04/QuantifyingPhi.md) + - [Quantiziranje Phi-3.5 / 4 koristeći llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Quantiziranje Phi-3.5 / 4 koristeći Generative AI ekstenzije za onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Quantiziranje Phi-3.5 / 4 koristeći Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Quantiziranje Phi-3.5 / 4 koristeći Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Evaluacija Phi - [Odgovorni AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry za evaluaciju](./md/01.Introduction/05/AIFoundry.md) - [Korištenje Promptflow za evaluaciju](./md/01.Introduction/05/Promptflow.md) -- RAG s Azure AI pretraživanjem - - [Kako koristiti Phi-4-mini i Phi-4-multimodal (RAG) s Azure AI pretraživanjem](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- RAG s Azure AI pretragom + - [Kako koristiti Phi-4-mini i Phi-4-multimodal(RAG) s Azure AI pretragom](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Nulti oblak lokalni hibridni RAG sa SQLite FTS5 i phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Primjeri razvoja Phi aplikacija +- Uzorci razvoja Phi aplikacija - Tekstualne i chat aplikacije - Phi-4 primjeri - - [📓] [Čat sa Phi-4-mini ONNX modelom](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Čat s Phi-4 lokalnim ONNX modelom .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Čat .NET konzolna aplikacija s Phi-4 ONNX koristeći Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 primjeri + - [📓] [Ćaskaj sa Phi-4-mini ONNX modelom](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Ćaskaj s Phi-4 lokalnim ONNX modelom .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Ćaskaj .NET konzolna aplikacija s Phi-4 ONNX koristeći Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Primjeri - [Lokalni chatbot u pregledniku koristeći Phi3, ONNX Runtime Web i WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Višestruki model - Interaktivni Phi-3-mini i OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [Višestruki modeli - Interaktivni Phi-3-mini i OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Izrada omotača i korištenje Phi-3 s MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Optimizacija modela - Kako optimizirati model Phi-3-min za ONNX Runtime Web s Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Optimizacija modela - Kako optimizirati Phi-3-min model za ONNX Runtime Web s Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [WinUI3 aplikacija s Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Višestruki model AI pokretana aplikacija za bilješke primjeri](https://github.com/microsoft/ai-powered-notes-winui3-sample) + -[WinUI3 aplikacija s Višestrukim modelima i AI potpomognutim bilješkama](https://github.com/microsoft/ai-powered-notes-winui3-sample) - [Fino podešavanje i integracija prilagođenih Phi-3 modela s Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - [Fino podešavanje i integracija prilagođenih Phi-3 modela s Prompt flow u Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Evaluacija fino podešenih Phi-3 / Phi-3.5 modela u Microsoft Foundry s fokusom na Microsoftova načela Odgovornog AI](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct uzorak za predviđanje jezika (kineski/engleski)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Korištenje Windows GPU-a za stvaranje rješenja Prompt flow s Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Evaluacija fino podešenog Phi-3 / Phi-3.5 modela u Microsoft Foundry s naglaskom na Microsoftove principe odgovornog AI](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct primjer predviđanja jezika (kineski/engleski)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Korištenje Windows GPU za stvaranje Prompt flow rješenja s Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Korištenje Microsoft Phi-3.5 tflite za stvaranje Android aplikacije](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Pitanja i odgovori .NET primjer koristeći lokalni ONNX Phi-3 model koristeći Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Konzolna chat .NET aplikacija sa Semantic Kernel i Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Q&A .NET primjer koristeći lokalni ONNX Phi-3 model s Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Console chat .NET aplikacija sa Semantic Kernel i Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK primjeri bazirani na kodu - - Phi-4 primjeri + - Azure AI Inference SDK Primjeri Koda + - Phi-4 Primjeri - [📓] [Generiranje koda projekta koristeći Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 primjeri + - Phi-3 / 3.5 Primjeri - [Izgradite vlastiti Visual Studio Code GitHub Copilot Chat s Microsoft Phi-3 obitelji](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Stvorite vlastitog Visual Studio Code Chat Copilot agenta s Phi-3.5 putem GitHub modela](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - Napredni uzorci rezoniranja - - Phi-4 primjeri - - [📓] [Phi-4-mini-reasoning ili Phi-4-reasoning primjeri](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Fino podešavanje Phi-4-mini-reasoning s Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Fino podešavanje Phi-4-mini-reasoning s Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning s GitHub modelima](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning s Microsoft Foundry modelima](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demos - - [Phi-4-mini demonstracije hostane na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demonstracije hostane na Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Primjeri za vid - - Phi-4 primjeri - - [📓] [Koristite Phi-4-multimodal za čitanje slika i generiranje koda](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 primjeri - - [📓][Phi-3-vision-Slika tekst u tekst](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Kreirajte vlastitog Visual Studio Code Chat Copilot agenta s Phi-3.5 koristeći GitHub modele](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - Primjeri Naprednog Rasuđivanja + - Phi-4 Primjeri + - [📓] [Phi-4-mini-razumijevanje ili Phi-4-razumijevanje primjeri](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Fino podešavanje Phi-4-mini-razumijevanja s Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Fino podešavanje Phi-4-mini-razumijevanja s Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-razumijevanje s GitHub modelima](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-razumijevanje s Microsoft Foundry modelima](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Demo primjeri + - [Phi-4-mini demo primjeri na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal demo primjeri na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Vision Primjeri + - Phi-4 Primjeri + - [📓] [Koristi Phi-4-multimodal za čitanje slika i generiranje koda](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 Primjeri + - [📓][Phi-3-vision-pretvaranje slike u tekst](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP ugradnja](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 recikliranje](https://github.com/jennifermarsman/PhiRecycling/) + - [📓][Phi-3-vision CLIP ugrađivanje](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [DEMO: Phi-3 Recikliranje](https://github.com/jennifermarsman/PhiRecycling/) - [Phi-3-vision - vizualni jezični asistent - s Phi3-Vision i OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision uzorak s više okvira ili više slika](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [📓][Phi-3.5 Vision višeslike ili višeslojni primjer](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - [Phi-3 Vision lokalni ONNX model koristeći Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - [Izbornik bazirani Phi-3 Vision lokalni ONNX model koristeći Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - Uzorci rezoniranja i vida - - Phi-4-Reasoning-Vision-15B - - [📓] [Korištenje Phi-4-Reasoning-Vision-15B za detekciju nezakonitog prelaska ceste](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Korištenje Phi-4-Reasoning-Vision-15B za matematiku](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Korištenje Phi-4-Reasoning-Vision-15B za detekciju UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - Primjeri Rasuđivanja i Vida + - Phi-4-Rasuđivanje-Vid-15B + - [📓] [Korištenje Phi-4-Rasuđivanje-Vid-15B za otkrivanje nepropisnog prelaska ceste](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Korištenje Phi-4-Rasuđivanje-Vid-15B za matematiku](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Korištenje Phi-4-Rasuđivanje-Vid-15B za otkrivanje korisničkog sučelja](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - Matematički primjeri - - Phi-4-Mini-Flash-Reasoning-Instruct primjeri [Matematička demonstracija s Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Matematički Primjeri + - Phi-4-Mini-Flash-Rasuđivanje-Upute Primjeri [Matematički Demo s Phi-4-Mini-Flash-Rasuđivanje-Instrukt](./md/02.Application/09.Math/MathDemo.ipynb) - - Audio primjeri - - Phi-4 primjeri - - [📓] [Izdvajanje audio transkripata koristeći Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal audio primjer](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal primjer govornog prijevoda](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - Audio Primjeri + - Phi-4 Primjeri + - [📓] [Ekstrahiranje audio transkripata koristeći Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal Audio Primjer](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal Primjer prijevoda govora](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - [.NET konzolna aplikacija koristeći Phi-4-multimodal audio za analizu audio datoteke i generiranje transkripta](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - MOE primjeri - - Phi-3 / 3.5 primjeri - - [📓] [Phi-3.5 Mixture of Experts Models (MoEs) primjer na društvenim mrežama](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Izgradnja Retrieval-Augmented Generation (RAG) pipeline-a s NVIDIA NIM Phi-3 MOE, Azure AI Search i LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - MOE Primjeri + - Phi-3 / 3.5 Primjeri + - [📓] [Phi-3.5 Mješavina stručnjaka modela (MoEs) Primjer za društvene mreže](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Izgradnja Retrieval-Augmented Generation (RAG) cjevovoda s NVIDIA NIM Phi-3 MOE, Azure AI Search i LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Primjeri poziva funkcija - - Phi-4 primjeri 🆕 - - [📓] [Korištenje poziva funkcija s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Korištenje poziva funkcija za stvaranje multi-agenta s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Korištenje poziva funkcija s Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Korištenje poziva funkcija s ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Primjeri miješanja multimodala - - Phi-4 primjeri 🆕 + - Primjeri Pozivanja Funkcija + - Phi-4 Primjeri 🆕 + - [📓] [Korištenje pozivanja funkcija s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Korištenje pozivanja funkcija za stvaranje više agenata s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Korištenje pozivanja funkcija s Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Korištenje pozivanja funkcija s ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Primjeri Miješanja Multimodala + - Phi-4 Primjeri 🆕 - [📓] [Korištenje Phi-4-multimodal kao tehnološki novinar](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [.NET konzolna aplikacija koristeći Phi-4-multimodal za analizu slika](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Fino podešavanje Phi primjera +- Fino podešavanje Phi Primjera - [Scenariji fino podešavanja](./md/03.FineTuning/FineTuning_Scenarios.md) - [Fino podešavanje vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Fino podešavanje neka Phi-3 postane industrijski stručnjak](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Dopusti Phi-3 da postane industrijski stručnjak](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - [Fino podešavanje Phi-3 s AI Toolkit za VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Fino podešavanje Phi-3 s Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Fino podešavanje Phi-3 s Azure Machine Learning servisom](./md/03.FineTuning/Introduce_AzureML.md) - [Fino podešavanje Phi-3 s Lora](./md/03.FineTuning/FineTuning_Lora.md) - [Fino podešavanje Phi-3 s QLora](./md/03.FineTuning/FineTuning_Qlora.md) - [Fino podešavanje Phi-3 s Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - [Fino podešavanje Phi-3 s Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - [Fino podešavanje s Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Fino podešavanje s Microsoft Olive praktična radionica](./md/03.FineTuning/olive-lab/readme.md) + - [Fino podešavanje s Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - [Fino podešavanje Phi-3-vision s Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Fino podešavanje Phi-3 s Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Fino podešavanje Phi-3-vision (službena podrška)](./md/03.FineTuning/FineTuning_Vision.md) - - [Fino podešavanje Phi-3 s Kaito AKS, Azure kontejnerima (službena podrška)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Fino podešavanje Phi-3 i 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + + - [Fine-tuning Phi-3 s Apple MLX Frameworkom](./md/03.FineTuning/FineTuning_MLX.md) + - [Fine-tuning Phi-3-vision (službena podrška)](./md/03.FineTuning/FineTuning_Vision.md) + - [Fine-Tuning Phi-3 s Kaito AKS, Azure Containers (službena podrška)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Fine-Tuning Phi-3 i 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Praktična radionica - [Istraživanje najnovijih modela: LLM-ovi, SLM-ovi, lokalni razvoj i više](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Otključavanje potencijala NLP-a: fino podešavanje s Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Otključavanje NLP potencijala: Fine-Tuning s Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - Akademski istraživački radovi i publikacije - - [Udžbenici su sve što vam treba II: tehnički izvještaj phi-1.5](https://arxiv.org/abs/2309.05463) - - [Phi-3 tehnički izvještaj: vrlo sposoban jezični model lokalno na vašem telefonu](https://arxiv.org/abs/2404.14219) + - [Tekstovi su sve što trebate II: tehnički izvještaj phi-1.5](https://arxiv.org/abs/2309.05463) + - [Phi-3 tehnički izvještaj: visoko sposoban jezični model lokalno na vašem telefonu](https://arxiv.org/abs/2404.14219) - [Phi-4 tehnički izvještaj](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini tehnički izvještaj: kompaktni, ali moćni multimodalni jezični modeli putem mješavine LoRA-e](https://arxiv.org/abs/2503.01743) - - [Optimizacija malih jezičnih modela za pozivanje funkcija u vozilu](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Fino podešavanje PHI-3 za odgovore na pitanja s višestrukim izborom: metodologija, rezultati i izazovi](https://arxiv.org/abs/2501.01588) - - [Phi-4-reasoning tehnički izvještaj](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini-reasoning tehnički izvještaj](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Phi-4-Mini tehnički izvještaj: kompaktni, ali moćni multimedijalni jezični modeli putem mješavine LoRA-e](https://arxiv.org/abs/2503.01743) + - [Optimizacija malih jezičnih modela za funkcijsko pozivanje u vozilu](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Fine-Tuning PHI-3 za višestruke izbore: metodologija, rezultati i izazovi](https://arxiv.org/abs/2501.01588) + - [Phi-4-razmišljanje tehnički izvještaj](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-razmišljanje tehnički izvještaj](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Korištenje Phi modela ### Phi na Microsoft Foundry -Možete naučiti kako koristiti Microsoft Phi i kako izgraditi E2E rješenja na različitim hardverskim uređajima. Za vlastito isprobavanje Phi-a, započnite s igranjem s modelima i prilagođavanjem Phi za vaše scenarije koristeći [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) više možete saznati u Uvod u [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Možete naučiti kako koristiti Microsoft Phi i kako izgraditi E2E rješenja na vašim različitim hardverskim uređajima. Da iskusite Phi sami, započnite igranjem s modelima i prilagođavanjem Phi za vaše scenarije koristeći [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai), više možete saznati u Uvod u [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Igralište** -Svaki model ima posebno igralište za testiranje modela [Azure AI Playground](https://aka.ms/try-phi3). +**Igralište** +Svaki model ima vlastito igralište za testiranje modela [Azure AI Playground](https://aka.ms/try-phi3). ### Phi na GitHub modelima -Možete naučiti kako koristiti Microsoft Phi i kako izgraditi E2E rješenja na različitim hardverskim uređajima. Za vlastito isprobavanje Phi-a, započnite s igranjem s modelom i prilagođavanjem Phi za vaše scenarije koristeći [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) više možete saznati u Uvod u [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Možete naučiti kako koristiti Microsoft Phi i kako izgraditi E2E rješenja na vašim različitim hardverskim uređajima. Da iskusite Phi sami, započnite igranjem s modelom i prilagođavanjem Phi za vaše scenarije koristeći [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo), više možete saznati u Uvod u [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Igralište** -Svaki model ima posebno [igralište za testiranje modela](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Igralište** +Svaki model ima posvećeno [igralište za testiranje modela](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi na Hugging Face -Model također možete pronaći na [Hugging Face](https://huggingface.co/microsoft) +Model možete također pronaći na [Hugging Face](https://huggingface.co/microsoft) -**Igralište** +**Igralište** [Hugging Chat igralište](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Ostali tečajevi + ## 🎒 Ostali tečajevi Naš tim proizvodi i druge tečajeve! Pogledajte: -### LangChain -[![LangChain4j za početnike](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js za početnike](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain za početnike](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +### LangChain +[![LangChain4j za početnike](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js za početnike](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain za početnike](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agent -[![AZD za početnike](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI za početnike](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP za početnike](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI agenti za početnike](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agent +[![AZD za početnike](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI za početnike](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP za početnike](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agent za početnike](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Generativni AI serijal -[![Generativni AI za početnike](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generativni AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generativni AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generativni AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### Serija Generative AI +[![Generative AI za početnike](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Osnovno učenje -[![ML za početnike](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science za početnike](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI za početnike](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity za početnike](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web razvoj za početnike](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT za početnike](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) + +### Jezgra učenja +[![ML za početnike](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science za početnike](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI za početnike](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersecurity za početnike](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web razvoj za početnike](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT za početnike](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![XR razvoj za početnike](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Copilot serijal -[![Copilot za AI upareno programiranje](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot za C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot avantura](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + +### Serija Copilot +[![Copilot za AI u parnom programiranju](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot za C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot avantura](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## Odgovorni AI -Microsoft se obvezuje pomoći našim korisnicima u odgovornom korištenju AI proizvoda, dijeljenju naših saznanja i izgradnji partnerstava zasnovanih na povjerenju putem alata poput Transparency Notes i Impact Assessments. Mnogi od tih resursa dostupni su na [https://aka.ms/RAI](https://aka.ms/RAI). -Microsoftov pristup odgovornom AI temelji se na našim AI principima pravednosti, pouzdanosti i sigurnosti, privatnosti i sigurnosti, uključivosti, transparentnosti i odgovornosti. +Microsoft je predan pomaganju našim korisnicima da odgovorno koriste naše AI proizvode, dijeleći naša iskustva i gradeći partnerske odnose temeljene na povjerenju kroz alate poput Transparency Notes i Impact Assessments. Mnogi od ovih resursa mogu se pronaći na [https://aka.ms/RAI](https://aka.ms/RAI). +Pristup Microsofta odgovornom AI temelji se na našim AI principima pravednosti, pouzdanosti i sigurnosti, privatnosti i sigurnosti, uključivosti, transparentnosti i odgovornosti. + +Veliki prirodni jezični, slikovni i govorni modeli - kao oni korišteni u ovom uzorku - mogu potencijalno imati ponašanja koja su nepravedna, nepouzdana ili uvredljiva, što može prouzročiti štetu. Molimo konzultirajte [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) kako biste bili informirani o rizicima i ograničenjima. + -Veliki modeli za prirodni jezik, sliku i govor - poput onih korištenih u ovom primjeru - mogu potencijalno imati ponašanja koja su nepravedna, nepouzdana ili uvredljiva, što može uzrokovati štetu. Molimo pogledajte [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) kako biste bili informirani o rizicima i ograničenjima. -Preporučeni pristup za ublažavanje ovih rizika je uključivanje sigurnosnog sustava u vašu arhitekturu koji može otkriti i spriječiti štetno ponašanje. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) pruža neovisni sloj zaštite, sposoban za otkrivanje štetnog sadržaja kojeg generiraju korisnici i AI u aplikacijama i servisima. Azure AI Content Safety uključuje tekstualne i slikovne API-je koji vam omogućuju otkrivanje štetnog materijala. Unutar Microsoft Foundry, servis Content Safety omogućuje vam pregled, istraživanje i isprobavanje primjera koda za otkrivanje štetnog sadržaja kroz različite modalitete. Sljedeća [brza uputa](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) vodi vas kroz slanje zahtjeva servisu. +Preporučeni pristup za ublažavanje ovih rizika je uključivanje sigurnosnog sustava u vašu arhitekturu koji može otkriti i spriječiti štetno ponašanje. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) pruža neovisni zaštitni sloj, sposoban otkriti štetni sadržaj koji generiraju korisnici i AI u aplikacijama i servisima. Azure AI Content Safety uključuje tekstualne i slikovne API-je koji vam omogućavaju otkrivanje štetnog materijala. U okviru Microsoft Foundry-a, servis Content Safety omogućuje vam pregled, istraživanje i isprobavanje primjera koda za otkrivanje štetnog sadržaja kroz različite modalitete. Sljedeća [quickstart dokumentacija](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) vodi vas kroz slanje zahtjeva servisu. -Drugi aspekt koji treba uzeti u obzir je opća izvedba aplikacije. Kod multimodalnih i multimodelskih aplikacija smatramo da izvedba znači da sustav radi kako vi i vaši korisnici očekujete, uključujući i ne-generiranje štetnih izlaza. Važno je procijeniti izvedbu vaše ukupne aplikacije koristeći [Evaluatore izvedbe i kvalitete te rizika i sigurnosti](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Također imate mogućnost kreirati i ocjenjivati s [prilagođenim evaluatorima](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Drugi aspekt koji treba uzeti u obzir je ukupna izvedba aplikacije. Kod aplikacija s više modaliteta i modela, izvedba se smatra time da sustav radi kako vi i vaši korisnici očekujete, uključujući neproizvođenje štetnih rezultata. Važno je procijeniti izvedbu vaše ukupne aplikacije koristeći [evaluatore za izvedbu i kvalitetu te rizik i sigurnost](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Također imate mogućnost kreiranja i evaluacije s [prilagođenim evaluatorima](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Svoju AI aplikaciju možete evaluirati u vašem razvojnom okruženju koristeći [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Na osnovi testnog skupa podataka ili cilja, generacije vaše generativne AI aplikacije kvantitativno se mjere ugrađenim ili prilagođenim evaluatorima po vašem izboru. Za početak s azure ai evaluation sdk za evaluaciju vašeg sustava, možete pratiti [brz vodič](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Nakon što provedete evaluaciju, možete [vizualizirati rezultate u Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Svoju AI aplikaciju možete evaluirati u razvojnom okruženju koristeći [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Imajući testni skup podataka ili cilj, generacije vaše generativne AI aplikacije kvantitativno se mjere ugrađenim evaluatorima ili prilagođenim evaluatorima po vašem izboru. Za početak s azure ai evaluation sdk za evaluaciju vašeg sustava, možete slijediti [quickstart vodič](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Nakon što izvršite evaluacijski rad, možete [vizualizirati rezultate u Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## Zaštitni znakovi +## Zaštićeni znakovi -Ovaj projekt može sadržavati zaštitne znakove ili logotipe za projekte, proizvode ili usluge. Ovlaštena upotreba Microsoftovih zaštitnih znakova ili logotipa podliježe i mora slijediti [Microsoftove smjernice za zaštitne znakove i brend](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Upotreba Microsoftovih zaštitnih znakova ili logotipa u modificiranim verzijama ovog projekta ne smije izazvati zbunjenost niti implicirati sponzorstvo Microsofta. Bilo koja upotreba zaštitnih znakova ili logotipa trećih strana podliježe pravilima tih trećih strana. +Ovaj projekt može sadržavati zaštitne znakove ili logotipe za projekte, proizvode ili usluge. Ovlaštena upotreba Microsoftovih zaštitnih znakova ili logotipa podliježe i mora slijediti [Microsoftova pravila za zaštitne znakove i brend](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Upotreba Microsoftovih zaštitnih znakova ili logotipa u izmijenjenim verzijama ovog projekta ne smije izazvati zabunu niti sugerirati Microsoftovu sponzorstvo. Svaka upotreba zaštitnih znakova ili logotipa trećih strana podliježe pravilima tih trećih strana. ## Dobivanje pomoći -Ako zapnete ili imate pitanja o izradi AI aplikacija, pridružite se: +Ako zapnete ili imate pitanja o izgradnji AI aplikacija, pridružite se: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Ako imate povratne informacije o proizvodu ili greške tijekom izrade, posjetite: +Ako imate povratne informacije o proizvodu ili greške prilikom izrade, posjetite: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Odricanje od odgovornosti**: -Ovaj dokument je preveden pomoću AI usluge za prijevod [Co-op Translator](https://github.com/Azure/co-op-translator). Iako težimo točnosti, imajte na umu da automatski prijevodi mogu sadržavati pogreške ili netočnosti. Izvorni dokument na izvornom jeziku treba se smatrati autoritativnim izvorom. Za ključne informacije preporučuje se profesionalni ljudski prijevod. Nismo odgovorni za bilo kakva nerazumijevanja ili pogrešne interpretacije koje proizlaze iz upotrebe ovog prijevoda. +**Napomena**: +Ovaj dokument je preveden korištenjem AI prevoditeljskog servisa [Co-op Translator](https://github.com/Azure/co-op-translator). Iako težimo točnosti, imajte na umu da automatski prijevodi mogu sadržavati greške ili netočnosti. Izvorni dokument na izvornom jeziku treba smatrati autoritativnim izvorom. Za važne informacije preporuča se profesionalni ljudski prijevod. Nismo odgovorni za bilo kakva nesporazumevanja ili pogrešne interpretacije koje proizlaze iz korištenja ovog prijevoda. \ No newline at end of file diff --git a/translations/hr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/hr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..c442ebc2c --- /dev/null +++ b/translations/hr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hibridni RAG s Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Autor:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Fokus:** Mali jezični modeli (SLM), SQLite FTS5 BM25, gusta ugradnja, Reciprokalna Rank Fuzija (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivacija: Dilema prisjećanja ključnih riječi u lokalnim SLM-ovima\n", + "Standardne RAG arhitekture koje se oslanjaju isključivo na guste vektorske ugradnje često ne uspijevaju dohvatiti točne numeričke tokene (npr. `2,340,000 TL`, šifre ugovora, brojeve računa). \n", + "Suprotno tome, rijetko leksičko pretraživanje (BM25) propušta semantičke sinonime i parafrazirana pitanja.\n", + "\n", + "Ovaj vodič pokazuje kako implementirati **brzi, u memoriji hibridni pretraživač** koji kombinira:\n", + "1. **Guste vektore** (kosinusna sličnost)\n", + "2. **Rijetko leksičko pretraživanje** (SQLite FTS5 BM25)\n", + "3. **Reciprokalnu Rank fuziju (RRF, $k=60$)**\n", + "4. **Utemeljenu generaciju citata (`[1]`, `[2]`)** s Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Dvostruka SQLite Shema (Gusti Vektori + Virtualna FTS5 BM25 Tablica)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Benchmark unosa i izvođenja uzorka\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Formulacija utemeljenog upita za Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Napomena**:\nOvaj dokument je preveden korištenjem AI prevoditeljskog servisa [Co-op Translator](https://github.com/Azure/co-op-translator). Iako težimo točnosti, imajte na umu da automatski prijevodi mogu sadržavati greške ili netočnosti. Izvorni dokument na izvornom jeziku treba smatrati autoritativnim izvorom. Za važne informacije preporuča se profesionalni ljudski prijevod. Nismo odgovorni za bilo kakva nesporazumevanja ili pogrešne interpretacije koje proizlaze iz korištenja ovog prijevoda.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/hu/.co-op-translator.json b/translations/hu/.co-op-translator.json index 4cfc0d89a..9d6cc0fa7 100644 --- a/translations/hu/.co-op-translator.json +++ b/translations/hu/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "hu" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:25:17+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:39:02+00:00", "source_file": "README.md", "language_code": "hu" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "hu" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:41:58+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "hu" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:53+00:00", diff --git a/translations/hu/README.md b/translations/hu/README.md index 260a65b98..8c414b2c4 100644 --- a/translations/hu/README.md +++ b/translations/hu/README.md @@ -1,42 +1,42 @@ # Phi Cookbook: Gyakorlati példák a Microsoft Phi modelljeivel -[![Open and use the samples in GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Open in Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Nyisd meg és használd a példákat GitHub Codespaces-ben](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Megnyitás Dev Containers-ben](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub contributors](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub issues](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub pull-requests](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub közreműködők](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub hibák](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub pull-kérelmek](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PR-k szívesen látottak](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub watchers](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub stars](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub megfigyelők](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub forkok](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub csillagok](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -A Phi a Microsoft által fejlesztett nyílt forráskódú AI modellek sorozata. +A Phi a Microsoft által fejlesztett nyílt forráskódú mesterséges intelligencia modellek sorozata. -A Phi jelenleg a leghatékonyabb és legköltséghatékonyabb kis nyelvi modell (SLM), amely kiváló eredményeket ért el többnyelvűség, következtetés, szöveg/csevegés generálás, kódolás, képek, hang és egyéb szcenáriók terén. +A Phi jelenleg a legerősebb és költséghatékonyabb kis nyelvi modell (SLM), nagyon jó eredményekkel többnyelvűség, érvelés, szöveg/csevegés generálás, kódolás, képek, hang és egyéb alkalmazási területeken. -Telepítheted a Phit felhőbe vagy élő eszközökre, és könnyedén építhetsz generatív AI alkalmazásokat korlátozott számítási kapacitással. +A Phit felhőbe vagy élő eszközökre telepítheted, és könnyedén építhetsz generatív AI alkalmazásokat korlátozott számítási kapacitással. -Kövesd az alábbi lépéseket az erőforrás használatának megkezdéséhez: -1. **Forkold le a tárat**: Kattints ide [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +Kövesd az alábbi lépéseket a kezdéshez: +1. **Forkold a tárat**: Kattints [![GitHub forkok](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Klónozd a tárat**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Csatlakozz a Microsoft AI Discord közösséghez, és találkozz szakértőkkel és fejlesztőtársaiddal**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Csatlakozz a Microsoft AI Discord közösséghez, és találkozz szakértőkkel és fejlesztőtársakkal**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/hu/cover.eb18d1b9605d754b.webp) +![borító](../../translated_images/hu/cover.eb18d1b9605d754b.webp) ### 🌐 Többnyelvű támogatás -#### GitHub Action segítségével támogatott (Automatikus és mindig naprakész) +#### GitHub Action (Automatizált és mindig naprakész) által támogatott -[Arab](../ar/README.md) | [Bengáli](../bn/README.md) | [Bolgár](../bg/README.md) | [Burmai (Mianmar)](../my/README.md) | [Kínai (egyszerűsített)](../zh-CN/README.md) | [Kínai (hagyományos, Hongkong)](../zh-HK/README.md) | [Kínai (hagyományos, Makaó)](../zh-MO/README.md) | [Kínai (hagyományos, Tajvan)](../zh-TW/README.md) | [Horvát](../hr/README.md) | [Cseh](../cs/README.md) | [Dán](../da/README.md) | [Holland](../nl/README.md) | [Észt](../et/README.md) | [Finn](../fi/README.md) | [Francia](../fr/README.md) | [Német](../de/README.md) | [Görög](../el/README.md) | [Héber](../he/README.md) | [Hindi](../hi/README.md) | [Magyar](./README.md) | [Indonéz](../id/README.md) | [Olasz](../it/README.md) | [Japán](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreai](../ko/README.md) | [Litván](../lt/README.md) | [Maláj](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepáli](../ne/README.md) | [Nigériai pidgin](../pcm/README.md) | [Norvég](../no/README.md) | [Perzsa (Fárszi)](../fa/README.md) | [Lengyel](../pl/README.md) | [Portugál (Brazília)](../pt-BR/README.md) | [Portugál (Portugália)](../pt-PT/README.md) | [Pandzsábi (Gurmukhi)](../pa/README.md) | [Román](../ro/README.md) | [Orosz](../ru/README.md) | [Szerb (cirill)](../sr/README.md) | [Szlovák](../sk/README.md) | [Szlovén](../sl/README.md) | [Spanyol](../es/README.md) | [Szuahéli](../sw/README.md) | [Svéd](../sv/README.md) | [Tagalog (Filippínó)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thailand](../th/README.md) | [Török](../tr/README.md) | [Ukrán](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnami](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgár](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Kínai (Egyszerűsített)](../zh-CN/README.md) | [Kínai (Hagyományos, Hongkong)](../zh-HK/README.md) | [Kínai (Hagyományos, Makaó)](../zh-MO/README.md) | [Kínai (Hagyományos, Taiwan)](../zh-TW/README.md) | [Horvát](../hr/README.md) | [Cseh](../cs/README.md) | [Dán](../da/README.md) | [Holland](../nl/README.md) | [Észt](../et/README.md) | [Finn](../fi/README.md) | [Francia](../fr/README.md) | [Német](../de/README.md) | [Görög](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Magyar](./README.md) | [Indonéz](../id/README.md) | [Olasz](../it/README.md) | [Japán](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreai](../ko/README.md) | [Litván](../lt/README.md) | [Maláj](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepáli](../ne/README.md) | [Nigeriai Pidgin](../pcm/README.md) | [Norvég](../no/README.md) | [Perzsa (Farsi)](../fa/README.md) | [Lengyel](../pl/README.md) | [Portugál (Brazília)](../pt-BR/README.md) | [Portugál (Portugália)](../pt-PT/README.md) | [Pandzsábi (Gurmukhi)](../pa/README.md) | [Román](../ro/README.md) | [Orosz](../ru/README.md) | [Szerb (Cirill)](../sr/README.md) | [Szlovák](../sk/README.md) | [Szlovén](../sl/README.md) | [Spanyol](../es/README.md) | [Szuahéli](../sw/README.md) | [Svéd](../sv/README.md) | [Tagalog (Filippínó)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Török](../tr/README.md) | [Ukrán](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnami](../vi/README.md) -> **Szeretnéd helyben klónozni?** +> **Inkább helyileg klónoznád?** > -> Ez a tár több mint 50 nyelvi fordítást tartalmaz, ami jelentősen megnöveli a letöltési méretet. Ha fordítások nélkül szeretnéd klónozni, használj sparse checkout-ot: +> Ez a tár 50+ nyelvi fordítást tartalmaz, ami jelentősen megnöveli a letöltési méretet. Fordítások nélkül klónozáshoz használd a sparse checkout-ot: > > **Bash / macOS / Linux:** > ```bash @@ -52,202 +52,205 @@ Kövesd az alábbi lépéseket az erőforrás használatának megkezdéséhez: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Ez minden szükséges dolgot biztosít a kurzus befejezéséhez sokkal gyorsabb letöltéssel. +> Ez mindent biztosít, amire a kurzus teljesítéséhez szükséged van, sokkal gyorsabb letöltéssel. ## Tartalomjegyzék - Bevezetés - [Üdvözlünk a Phi családban](./md/01.Introduction/01/01.PhiFamily.md) - - [Környezeti beállítás](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Kulcsfontosságú technológiák megértése](./md/01.Introduction/01/01.Understandingtech.md) - - [AI biztonság a Phi modellekhez](./md/01.Introduction/01/01.AISafety.md) - - [Phi hardvertámogatás](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi modellek és elérhetőség platformokon](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Guidance-ai és Phi használata](./md/01.Introduction/01/01.Guidance.md) - - [GitHub piactéri modellek](https://github.com/marketplace/models) + - [A környezet beállítása](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [A kulcstechnológiák megértése](./md/01.Introduction/01/01.Understandingtech.md) + - [Mesterséges intelligencia biztonsága a Phi modellek esetén](./md/01.Introduction/01/01.AISafety.md) + - [Phi hardver támogatás](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Phi modellek és elérhetőség a platformokon](./md/01.Introduction/01/01.Edgeandcloud.md) + - [A Guidance-ai és Phinak a használata](./md/01.Introduction/01/01.Guidance.md) + - [GitHub Marketplace modellek](https://github.com/marketplace/models) - [Azure AI modell katalógus](https://ai.azure.com) -- Phi futtatása különböző környezetekben +- Phi használata különböző környezetekben - [Hugging Face](./md/01.Introduction/02/01.HF.md) - [GitHub modellek](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry modell katalógus](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Microsoft Foundry Modell Katalógus](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry helyi](./md/01.Introduction/02/07.FoundryLocal.md) -- Phi család futtatása - - [Phi futtatása iOS-en](./md/01.Introduction/03/iOS_Inference.md) - - [Phi futtatása Androidon](./md/01.Introduction/03/Android_Inference.md) - - [Phi futtatása Jetsonon](./md/01.Introduction/03/Jetson_Inference.md) - - [Phi futtatása AI PC-n](./md/01.Introduction/03/AIPC_Inference.md) - - [Phi futtatása Apple MLX Framework-kel](./md/01.Introduction/03/MLX_Inference.md) - - [Phi futtatása helyi szerveren](./md/01.Introduction/03/Local_Server_Inference.md) - - [Phi futtatása távoli szerveren AI Toolkit használatával](./md/01.Introduction/03/Remote_Interence.md) - - [Phi futtatása Rust-tal](./md/01.Introduction/03/Rust_Inference.md) - - [Phi Vision futtatása helyben](./md/01.Introduction/03/Vision_Inference.md) - - [Phi futtatása Kaito AKS, Azure Containers (hivatalos támogatás)](./md/01.Introduction/03/Kaito_Inference.md) -- [Phi család kvantálása](./md/01.Introduction/04/QuantifyingPhi.md) - - [Phi-3.5 / 4 kvantálása llama.cpp használatával](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Phi-3.5 / 4 kvantálása Generative AI bővítményekkel onnxruntime-hoz](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Phi-3.5 / 4 kvantálása Intel OpenVINO használatával](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Phi-3.5 / 4 kvantálása Apple MLX Framework-kel](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Phi értékelése +- Phi család használata + - [Phi használata iOS-en](./md/01.Introduction/03/iOS_Inference.md) + - [Phi használata Androidon](./md/01.Introduction/03/Android_Inference.md) + - [Phi használata Jetsonon](./md/01.Introduction/03/Jetson_Inference.md) + - [Phi használata AI PC-n](./md/01.Introduction/03/AIPC_Inference.md) + - [Phi használata Apple MLX keretrendszerrel](./md/01.Introduction/03/MLX_Inference.md) + - [Phi használata helyi szerveren](./md/01.Introduction/03/Local_Server_Inference.md) + - [Phi használata távoli szerveren AI Toolkit-kel](./md/01.Introduction/03/Remote_Interence.md) + - [Phi használata Rusttal](./md/01.Introduction/03/Rust_Inference.md) + - [Phi látás használata helyben](./md/01.Introduction/03/Vision_Inference.md) + - [Phi használata Kaito AKS, Azure Containers-szel (hivatalos támogatás)](./md/01.Introduction/03/Kaito_Inference.md) +- [Phi család mennyiségi jellemzése](./md/01.Introduction/04/QuantifyingPhi.md) + - [Phi-3.5 / 4 kvantálása llama.cpp-vel](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Phi-3.5 / 4 kvantálása generatív AI bővítményekkel onnxruntime-hoz](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Phi-3.5 / 4 kvantálása Intel OpenVINO-val](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Phi-3.5 / 4 kvantálása Apple MLX keretrendszerrel](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- Phi értékelése - [Felelős AI](./md/01.Introduction/05/ResponsibleAI.md) - - [Microsoft Foundry értékeléshez](./md/01.Introduction/05/AIFoundry.md) + - [Microsoft Foundry az értékeléshez](./md/01.Introduction/05/AIFoundry.md) - [Promptflow használata értékeléshez](./md/01.Introduction/05/Promptflow.md) -- RAG az Azure AI Kereséssel - - [Hogyan használd a Phi-4-mini és Phi-4-multimodal (RAG) modelleket Azure AI Kereséssel](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- RAG Azure AI Search segítségével + - [Hogyan használd Phi-4-mini és Phi-4-multimodal(RAG)-t Azure AI Search-csal](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud helyi hibrid RAG SQLite FTS5-tel és phi-4-minivel](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Phi alkalmazásfejlesztési példák - - Szöveg és csevegő alkalmazások - - Phi-4 példák - - [📓] [Csevegés Phi-4-mini ONNX modellel](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - Szöveg és Csevegés alkalmazások + - Phi-4 példák + - [📓] [Csevegj a Phi-4-mini ONNX modellel](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Csevegés Phi-4 helyi ONNX modellel .NET-ben](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Csevegő .NET konzolalkalmazás Phi-4 ONNX-szel a Semantic Kernel használatával](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 példák - - [Helyi csevegő a böngészőben Phi3, ONNX Runtime Web és WebGPU használatával](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Csevegés .NET konzol alkalmazás Phi-4 ONNX használatával Sementic Kernellel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Példák + - [Helyi chatbot a böngészőben Phi3, ONNX Runtime Web és WebGPU használatával](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [Több modell - Interaktív Phi-3-mini és OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Wrapper készítése és Phi-3 használata MLFlow-val](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Modelloptimalizálás - Hogyan optimalizáljuk a Phi-3-mini modellt ONNX Runtime Web használatra Olive-dal](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 alkalmazás Phi-3 mini-4k-instruct-onnx modellel](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Többmodell AI által vezérelt jegyzet alkalmazás példa](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [MLFlow - Wrapper készítése és Phi-3 használata az MLFlow-val](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Modell optimalizálás - Hogyan optimalizáljuk a Phi-3-min modellt ONNX Runtime Webhez az Olive segítségével](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 alkalmazás Phi-3 mini-4k-instruct-onnx-szal](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 Többmodellű AI által vezérelt jegyzet alkalmazás példa](https://github.com/microsoft/ai-powered-notes-winui3-sample) - [Egyedi Phi-3 modellek finomhangolása és integrálása Prompt flow-val](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - [Egyedi Phi-3 modellek finomhangolása és integrálása Prompt flow-val a Microsoft Foundry-ban](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [A finomhangolt Phi-3 / Phi-3.5 modell értékelése a Microsoft Foundry-ban, fókuszálva a Microsoft Felelős MI elveire](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [A finomhangolt Phi-3 / Phi-3.5 modell értékelése a Microsoft Foundry-ban a Microsoft felelős MI elveire fókuszálva](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Phi-3.5-mini-instruct nyelvi előrejelző minta (kínai/angol)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU használata Prompt flow megoldás létrehozásához Phi-3.5-Instruct ONNX-szel](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Phi-3.5-Instruct WebGPU RAG chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Windows GPU használata Prompt flow megoldás létrehozásához Phi-3.5-Instruct ONNX-szal](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Microsoft Phi-3.5 tflite használata Android alkalmazás létrehozásához](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - [Kérdések és válaszok .NET példa helyi ONNX Phi-3 modellel a Microsoft.ML.OnnxRuntime használatával](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Konzol chat .NET alkalmazás Semantic Kernel és Phi-3 modellel](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI Inference SDK Kód Alapú Minták - - Phi-4 Minták - - [📓] [Projektkód generálása Phi-4-multimodal használatával](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 Minták - - [Építsd meg saját Visual Studio Code GitHub Copilot Chat-edet a Microsoft Phi-3 családdal](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Hozd létre saját Visual Studio Code Chat Copilot ügynököd Phi-3.5-tel GitHub modellekkel](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - Haladó Érvelési Minták - - Phi-4 Minták - - [📓] [Phi-4-mini-érvelés vagy Phi-4-érvelés minták](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Phi-4-mini-érvelés finomhangolása Microsoft Olive-dal](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-érvelés finomhangolása Apple MLX-szel](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [Konzol chat .NET alkalmazás Semantic Kernel és Phi-3 használatával](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI Inference SDK Kód alapú példák + - Phi-4 Példák + - [📓] [Projektkód generálása Phi-4-multimodálissal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Phi-3 / 3.5 Példák + - [Saját Visual Studio Code GitHub Copilot Chat építése a Microsoft Phi-3 családdal](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Hozzon létre saját Visual Studio Code Chat Copilot ügynököt Phi-3.5-tel a GitHub modellek segítségével](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - Haladó érvelési példák + - Phi-4 Példák + - [📓] [Phi-4-mini-érvelési vagy Phi-4-érvelési példák](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Phi-4-mini-érvelés finomhangolása Microsoft Olive segítségével](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-érvelés finomhangolása Apple MLX használatával](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-érvelés GitHub modellekkel](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-érvelés Microsoft Foundry modellekkel](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demók - [Phi-4-mini demók a Hugging Face Spaces-en](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demók a Hugging Face Spaces-en](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Látás Minták - - Phi-4 Minták - - [📓] [Használja a Phi-4-multimodal-t képek olvasására és kód generálására](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 Minták - - [📓][Phi-3-látás-Képszöveg átalakítása szöveggé](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-látás-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-látás CLIP beágyazás](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 újrahasznosítás](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-látás - Vizualis nyelvi asszisztens - Phi3-Latással és OpenVINO-val](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Látás Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Látás OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Látás többkeretes vagy többképes minta](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Látás Helyi ONNX Modell a Microsoft.ML.OnnxRuntime .NET segítségével](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menü alapú Phi-3 Látás Helyi ONNX Modell a Microsoft.ML.OnnxRuntime .NET segítségével](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Érvelés-Látás Minták - - Phi-4-Érvelés-Látás-15B - - [📓] [Phi-4-Érvelés-Látás-15B használata tilos átkelés észlelésére](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Phi-4-Érvelés-Látás-15B használata matematikához](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-Érvelés-Látás-15B használata UI észlelésére](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - Matematikai Minták - - Phi-4-Mini-Flash-Érvelés-Utasítás Minták [Matematika Demo Phi-4-Mini-Flash-Érvelés-Utasítással](./md/02.Application/09.Math/MathDemo.ipynb) - - - Hang Minták - - Phi-4 Minták - - [📓] [Hangátiratok kinyerése Phi-4-multimodal segítségével](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal Hangminta](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal Beszédfordítási minta](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET konzolalkalmazás Phi-4-multimodallal hangfájl elemzésére és átirat generálására](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - MOE Minták - - Phi-3 / 3.5 Minták - - [📓] [Phi-3.5 Szakértői keverék modellek (MoEs) közösségi média minta](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Visszakeresés-alapú generálási (RAG) pipeline építése NVIDIA NIM Phi-3 MOE-val, Azure AI Kereséssel és LlamaIndex-szel](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [Phi-4-multimodális demók a Hugginge Face Spaces-en](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Vízió példák + - Phi-4 Példák + - [📓] [Phi-4-multimodális képek olvasása és kód generálása](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 Példák + - [📓][Phi-3-vízió Kép szöveghez szöveg online végpont](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vízió-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vízió CLIP beágyazás](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [DEMÓ: Phi-3 Újrahasznosítás](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vízió - vizuális nyelvi asszisztens Phi3-Vision és OpenVINO segítségével](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision többkeretes vagy többképes minta](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision helyi ONNX modell Microsoft.ML.OnnxRuntime .NET használatával](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Menü alapú Phi-3 Vision helyi ONNX modell Microsoft.ML.OnnxRuntime .NET használatával](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Érvelés-Vízió példák + - Phi-4-Érvelés-Vízió-15B + - [📓] [Phi-4-Érvelés-Vízió-15B használata tilos átkelés érzékelésére](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-Érvelés-Vízió-15B használata matematikához](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Phi-4-Érvelés-Vízió-15B használata UI érzékelésére](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - Matematikai példák + - Phi-4-Mini-Flash-Érvelés-Irányított példák [Matematikai demo Phi-4-Mini-Flash-Érvelés-Irányítottal](./md/02.Application/09.Math/MathDemo.ipynb) + + - Hang példák + - Phi-4 Példák + - [📓] [Hangátiratok kinyerése Phi-4-multimodálissal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodális hang minta](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodális beszédfordítás minta](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET konzol alkalmazás Phi-4-multimodális hang fájl elemzésére és átirat készítésére](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE példák + - Phi-3 / 3.5 Példák + - [📓] [Phi-3.5 Szakértők keveréke modellek (MoE-k) közösségi média minta](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Lekérés-alapú generálási (RAG) folyamat létrehozása NVIDIA NIM Phi-3 MOE, Azure AI Search és LlamaIndex segítségével](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Funkcióhívás Minták - - Phi-4 Minták 🆕 - - [📓] [Funkcióhívás használata Phi-4-mini-vel](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Funkcióhívás használata multi-ügynökök létrehozásához Phi-4-minivel](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Funkcióhívás használata Ollama-val](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Funkcióhívás használata ONNX-szel](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Többmodalitás keverése Minták - - Phi-4 Minták 🆕 - - [📓] [Phi-4-multimodal használata technológiai újságíróként](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET konzolalkalmazás Phi-4-multimodal képelemzéshez](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Phi finomhangolása minták + - Függvényhívás példák + - Phi-4 Példák 🆕 + - [📓] [Függvényhívás használata Phi-4-minivel](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Függvényhívás használata több ügynök létrehozásához Phi-4-minivel](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Függvényhívás használata Ollamával](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Függvényhívás használata ONNX-szal](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Multimodális keverési példák + - Phi-4 Példák 🆕 + - [📓] [Phi-4-multimodális használata technológiai újságíróként](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET konzol alkalmazás képek elemzésére Phi-4-multimodális használatával](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Phi finomhangolás - [Finomhangolási forgatókönyvek](./md/03.FineTuning/FineTuning_Scenarios.md) - [Finomhangolás vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Finomhangolás: Engedd, hogy Phi-3 iparági szakértővé váljon](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Phi-3 finomhangolása AI Toolkitkel a VS Code-hoz](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Phi-3 finomhangolása Azure Machine Learning Service-szel](./md/03.FineTuning/Introduce_AzureML.md) - - [Phi-3 finomhangolása Lora-val](./md/03.FineTuning/FineTuning_Lora.md) - - [Phi-3 finomhangolása QLora-val](./md/03.FineTuning/FineTuning_Qlora.md) + - [Finomhangolás: Hagyjuk, hogy Phi-3 ipari szakértővé váljon](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Phi-3 finomhangolása AI Toolkit for VS Code használatával](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Phi-3 finomhangolása Azure Machine Learning Service használatával](./md/03.FineTuning/Introduce_AzureML.md) + - [Phi-3 finomhangolása Lora használatával](./md/03.FineTuning/FineTuning_Lora.md) + - [Phi-3 finomhangolása QLora használatával](./md/03.FineTuning/FineTuning_Qlora.md) - [Phi-3 finomhangolása Microsoft Foundry-val](./md/03.FineTuning/FineTuning_AIFoundry.md) - [Phi-3 finomhangolása Azure ML CLI/SDK-val](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Finomhangolás Microsoft Olive-dal](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive gyakorlati labor finomhangolással](./md/03.FineTuning/olive-lab/readme.md) - - [Phi-3-látás finomhangolása Weights and Biases-szal](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Phi-3 finomhangolása Apple MLX Frameworkkel](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-látás finomhangolása (hivatalos támogatás)](./md/03.FineTuning/FineTuning_Vision.md) - - [Phi-3 finomhangolása Kaito AKS-szel, Azure konténerekkel (hivatalos támogatás)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Finomhangolás Microsoft Olive használatával](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Finomhangolás Microsoft Olive gyakorlati laborral](./md/03.FineTuning/olive-lab/readme.md) + - [Phi-3-vízió finomhangolása Weights and Bias segítségével](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Phi-3 finomhangolása az Apple MLX keretrendszerrel](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision finomhangolása (hivatalos támogatás)](./md/03.FineTuning/FineTuning_Vision.md) + - [Phi-3 finomhangolása Kaito AKS, Azure konténerekkel (hivatalos támogatás)](./md/03.FineTuning/FineTuning_Kaito.md) - [Phi-3 és 3.5 Vision finomhangolása](https://github.com/2U1/Phi3-Vision-Finetune) - Gyakorlati labor - - [Élvonalbeli modellek felfedezése: LLM-ek, SLM-ek, helyi fejlesztés és még sok más](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Az NLP potenciáljának kiaknázása: Finomhangolás a Microsoft Olive-dal](https://github.com/azure/Ignite_FineTuning_workshop) + - [Élvonalbeli modellek felfedezése: LLM-ek, SLM-ek, helyi fejlesztés és mások](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Az NLP potenciáljának feltárása: finomhangolás a Microsoft Olive-dzsal](https://github.com/azure/Ignite_FineTuning_workshop) -- Tudományos kutatási cikkek és publikációk +- Tudományos kutatási cikkek és kiadványok - [Textbooks Are All You Need II: phi-1.5 műszaki jelentés](https://arxiv.org/abs/2309.05463) - - [Phi-3 műszaki jelentés: Nagyon képzett nyelvi modell helyileg a telefonodon](https://arxiv.org/abs/2404.14219) + - [Phi-3 műszaki jelentés: Nagyon képzett nyelvi modell helyben a telefonodon](https://arxiv.org/abs/2404.14219) - [Phi-4 műszaki jelentés](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini műszaki jelentés: Kompakt, mégis erős multimodális nyelvi modellek Mixture-of-LoRAs segítségével](https://arxiv.org/abs/2503.01743) - - [Kis nyelvi modellek optimalizálása járműben végzett funkcionalitás-híváshoz](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) PHI-3 finomhangolása feleletválasztós kérdésekre: Módszertan, eredmények és kihívások](https://arxiv.org/abs/2501.01588) - - [Phi-4-reasoning műszaki jelentés](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini-reasoning műszaki jelentés](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Phi-4-Mini műszaki jelentés: Kompakt, mégis erőteljes multimodális nyelvi modellek Mixture-of-LoRAs segítségével](https://arxiv.org/abs/2503.01743) + - [Kis nyelvi modellek optimalizálása járműfunkció-hívásokra](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) PHI-3 finomhangolása feleletválasztós kérdésekhez: módszertan, eredmények és kihívások](https://arxiv.org/abs/2501.01588) + - [Phi-4-érvelés műszaki jelentés](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-érvelés műszaki jelentés](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Phi modellek használata ### Phi a Microsoft Foundry-n -Megtanulhatod, hogyan használd a Microsoft Phi-t, és hogyan építs végponttól végpontig (E2E) megoldásokat különböző hardvereszközeiden. Ahhoz, hogy te magad is kipróbáld a Phi-t, kezdj el játszani a modellekkel és testre szabni a Phi-t a saját eseteidhez a [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) segítségével. További információkat találsz a [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) kezdő lépéseiről. +Megtanulhatod, hogyan használd a Microsoft Phi-t, és hogyan építs végponttól végpontig (E2E) megoldásokat különböző hardvereszközeiden. Ha saját magad szeretnéd kipróbálni a Phit, kezdd azzal, hogy játszol a modellekkel és testre szabod a Phi-t a saját eseteidhez a [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) segítségével. Többet megtudhatsz a Kezdés a [Microsoft Foundry-ral](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) című dokumentumból. -**Játszótér** -Minden modellhez külön játszótér tartozik a modell tesztelésére: [Azure AI Playground](https://aka.ms/try-phi3). +**Játéktér** +Minden modellhez tartozik egy dedikált játéktér, ahol tesztelheted a modellt [Azure AI Playground](https://aka.ms/try-phi3). ### Phi a GitHub modelleken -Megtanulhatod, hogyan használd a Microsoft Phi-t, és hogyan építs végponttól végpontig (E2E) megoldásokat különböző hardvereszközeiden. Ahhoz, hogy te magad is kipróbáld a Phi-t, kezdj el játszani a modellel és testre szabni a Phi-t a saját eseteidhez a [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) segítségével. További információkat találsz a [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) kezdő lépéseiről. +Megtanulhatod, hogyan használd a Microsoft Phi-t, és hogyan építs E2E megoldásokat különböző hardvereszközeiden. Ha saját magad szeretnéd kipróbálni a Phit, kezdd azzal, hogy játszol a modellel és testre szabod a Phi-t a saját eseteidhez a [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) segítségével. Többet megtudhatsz a Kezdés a [GitHub Model Catalog]-val (/md/02.QuickStart/GitHubModel_QuickStart.md) című dokumentumból. -**Játszótér** -Minden modellhez külön [játszótér a modell tesztelésére](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Játéktér** +Minden modellhez tartozik egy dedikált [játéktér a modell teszteléséhez](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi a Hugging Face-en -A modellt megtalálhatod a [Hugging Face-en](https://huggingface.co/microsoft). +A modellt megtalálhatod a [Hugging Face-en](https://huggingface.co/microsoft) -**Játszótér** - [Hugging Chat játszótér](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Játéktér** + [Hugging Chat játéktér](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) ## 🎒 Egyéb kurzusok -Csapatunk más kurzusokat is kínál! Tekintsd meg: +Csapatunk további kurzusokat is készít! Nézd meg: ### LangChain @@ -256,11 +259,11 @@ Csapatunk más kurzusokat is kínál! Tekintsd meg: [![LangChain kezdőknek](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Ügynökök +### Azure / Edge / MCP / Agents [![AZD kezdőknek](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI kezdőknek](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP kezdőknek](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI ügynökök kezdőknek](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents kezdőknek](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- @@ -272,12 +275,12 @@ Csapatunk más kurzusokat is kínál! Tekintsd meg: --- -### Alapvető tanulás -[![Gépi tanulás kezdőknek](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +### Alapképzés +[![ML kezdőknek](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Adattudomány kezdőknek](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![Mesterséges intelligencia kezdőknek](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![AI kezdőknek](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![Kiberbiztonság kezdőknek](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web fejlesztés kezdőknek](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![Webfejlesztés kezdőknek](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![IoT kezdőknek](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![XR fejlesztés kezdőknek](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) @@ -289,36 +292,38 @@ Csapatunk más kurzusokat is kínál! Tekintsd meg: [![Copilot kaland](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Felelős mesterséges intelligencia +## Felelős MI + +A Microsoft elkötelezett amellett, hogy ügyfeleinket felelősen támogassa MI termékeink használatában, megossza tanulságainkat és bizalmon alapuló partnerségeket építsen átláthatósági jegyzetek és hatásvizsgálatok segítségével. Ezeknek az erőforrásoknak sok megtalálható itt: [https://aka.ms/RAI](https://aka.ms/RAI). +A Microsoft felelős MI megközelítése az igazságosság, megbízhatóság és biztonság, adatvédelem és biztonság, befogadás, átláthatóság és elszámoltathatóság MI alapelveire épül. + +Nagy méretű természetes nyelvi, képi és beszédmodellek - mint az ebben a mintában használtak - potenciálisan olyan viselkedést tanúsíthatnak, amely igazságtalan, megbízhatatlan vagy sértő, ami károkat okozhat. Kérjük, tekintsd meg az [Azure OpenAI szolgáltatás átláthatósági jegyzetét](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), hogy tájékozódj a kockázatokról és korlátokról. -A Microsoft elkötelezett amellett, hogy ügyfeleink felelősségteljesen használják mesterséges intelligencia termékeinket, megosszák tanulságainkat, és bizalmon alapuló partnerségeket építsenek olyan eszközök révén, mint az Átláthatósági jegyzetek és Hatásértékelések. Ezeknek az erőforrásoknak sok megtalálható a [https://aka.ms/RAI](https://aka.ms/RAI) címen. -A Microsoft felelős MI megközelítése az igazságosság, megbízhatóság és biztonság, adatvédelem és biztonság, befogadás, átláthatóság és elszámoltathatóság alapelvein nyugszik. -A nagy léptékű természetes nyelvű, képi és beszédfeldolgozó modellek – mint például a mintában használtak is – potenciálisan igazságtalan, megbízhatatlan vagy sértő módon viselkedhetnek, ami károkat okozhat. Kérjük, tájékozódj az [Azure OpenAI szolgáltatás Átláthatósági jegyzetéből](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), hogy informált legyél a kockázatokról és korlátokról. -Az ajánlott megközelítés ezen kockázatok mérséklésére egy olyan biztonsági rendszer beépítése az architektúrába, amely képes észlelni és megakadályozni a káros viselkedést. Az [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) egy független védelmi réteget biztosít, amely képes észlelni a káros, felhasználók által vagy AI által generált tartalmakat az alkalmazásokban és szolgáltatásokban. Az Azure AI Content Safety szöveg- és kép API-kat tartalmaz, amelyek lehetővé teszik a káros anyagok felderítését. A Microsoft Foundry-n belül a Content Safety szolgáltatás lehetőséget ad arra, hogy megtekinthesse, felfedezze és kipróbálja a káros tartalom észlelésére vonatkozó különböző modalitások mintakódjait. A következő [gyorsindítási dokumentáció](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) vezet végig a szolgáltatásnak történő kérelmek küldésén. +A kockázatok mérséklésére ajánlott megközelítés, hogy a rendszertervedbe beépítesz egy biztonsági rendszert, amely képes felismerni és megakadályozni a káros viselkedést. Az [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) egy független védelmi réteget nyújt, amely képes felismerni a káros, felhasználók vagy mesterséges intelligencia által generált tartalmakat alkalmazásokban és szolgáltatásokban. Az Azure AI Content Safety szöveges és képi API-kat tartalmaz, amelyek lehetővé teszik a káros anyagok felismerését. A Microsoft Foundry keretén belül a Content Safety szolgáltatás lehetőséget ad arra, hogy megtekinthesd, felfedezd és kipróbáld a káros tartalmak különböző módokon történő felismerését bemutató mintakódokat. A következő [gyorsindítási dokumentáció](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) végigvezet a szolgáltatáshoz intézett kérések elkészítésén. -Egy másik szempont, amit figyelembe kell venni, a teljes alkalmazás teljesítménye. Többmodalitású és többmodelles alkalmazások esetén a teljesítményünk azt jelenti, hogy a rendszer úgy működik, ahogyan Ön és felhasználói elvárják, beleértve azt is, hogy nem generál káros kimeneteket. Fontos az általános alkalmazás teljesítményének értékelése a [Teljesítmény és Minőség, valamint Kockázat és Biztonság értékelőkkel](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Továbbá lehetősége van [egyedi értékelők](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) létrehozására és használatára is. +Egy másik figyelembe veendő szempont az alkalmazás teljesítménye általánosságban. Többmodalitású és többmodelles alkalmazások esetén a teljesítmény azt jelenti, hogy a rendszer úgy működik, ahogy te és a felhasználóid várják, ideértve azt is, hogy nem generál káros kimeneteket. Fontos felmérni az alkalmazás teljesítményét az [Teljesítmény és Minőség valamint Kockázat és Biztonságértékelők](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) használatával. Ezenkívül lehetőséged van [egyéni értékelők](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) létrehozására és használatára is. -Az AI alkalmazását értékelheti a fejlesztési környezetében az [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) segítségével. Egy teszt-adatkészlet vagy cél megadása esetén a generatív AI alkalmazás generációit kvantitatívan mérik a beépített vagy az Ön által választott egyedi értékelőkkel. Az azure ai evaluation sdk-val való kezdéshez és a rendszer értékeléséhez kövesse a [gyorsindítási útmutatót](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Az értékelés futtatása után a [Microsoft Foundry-ban meg is jelenítheti az eredményeket](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Az AI alkalmazásodat a fejlesztői környezetedben is értékelheted az [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) segítségével. Egy tesztadatkészlet vagy cél megadása esetén a generatív AI alkalmazásod generációit mennyiségi szempontból mérik beépített vagy általad választott egyéni értékelők. Az azure ai evaluation sdk-val való kezdéshez és a rendszer értékeléséhez követheted a [gyorsindítási útmutatót](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Az értékelési futtatás után a [Microsoft Foundry-ban megtekintheted az eredményeket](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Védjegyek -Ez a projekt tartalmazhat védjegyeket vagy logókat projektekhez, termékekhez vagy szolgáltatásokhoz. A Microsoft védjegyek vagy logók jogosult felhasználása a [Microsoft Védjegy- és Márkaelveknek](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) megfelelően történhet, és azokat követni kell. -A Microsoft védjegyek vagy logók használata a projekt módosított változataiban nem okozhat félreértést, és nem utalhat a Microsoft támogatására. Harmadik fél védjegyeinek vagy logóinak bármilyen használata a harmadik fél szabályzata hatálya alá tartozik. +Ez a projekt tartalmazhat projektek, termékek vagy szolgáltatások védjegyeit vagy logóit. A Microsoft védjegyeinek vagy logóinak jogosult használata a [Microsoft védjegy- és márka irányelveinek](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) követelményei szerint történhet. +A Microsoft védjegyek vagy logók használata módosított projektverziókban nem okozhat félreértést vagy azt, hogy a Microsoft támogatását sugallja. Harmadik fél védjegyeinek vagy logóinak bármilyen használata a harmadik fél szabályzataihoz kötött. -## Segítségkérés +## Segítség kérése -Ha elakad, vagy kérdése van AI alkalmazások fejlesztésével kapcsolatban, csatlakozzon: +Ha elakadsz vagy kérdésed van AI alkalmazások fejlesztésével kapcsolatban, csatlakozz: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Ha termék-visszajelzése vagy hibabejelentése van fejlesztés közben, látogasson el ide: +Ha termék-visszajelzésed vagy hibák jelentkeznek a fejlesztés során, látogasd meg: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Jogi nyilatkozat**: -Ez a dokumentum az AI fordító szolgáltatás, a [Co-op Translator](https://github.com/Azure/co-op-translator) segítségével készült. Bár törekszünk a pontosságra, kérjük, vegye figyelembe, hogy az automatikus fordítások hibákat vagy pontatlanságokat tartalmazhatnak. Az eredeti dokumentum anyanyelvű változata tekintendő hiteles forrásnak. Kritikus információk esetén professzionális emberi fordítás ajánlott. Nem vállalunk felelősséget a fordítás használatából eredő félreértésekért vagy téves értelmezésekért. +**Jogi nyilatkozat**: +Ez a dokumentum az AI fordítási szolgáltatás, a [Co-op Translator](https://github.com/Azure/co-op-translator) segítségével készült. Bár az pontosságra törekszünk, kérjük, vegye figyelembe, hogy az automatikus fordítások hibákat vagy pontatlanságokat tartalmazhatnak. Az eredeti dokumentum az anyanyelvén tekintendő hiteles forrásnak. Fontos információk esetén professzionális emberi fordítást javasolunk. Nem vállalunk felelősséget semmilyen félreértésért vagy téves értelmezésért, amely ebből a fordításból ered. \ No newline at end of file diff --git a/translations/hu/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/hu/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..8554bad26 --- /dev/null +++ b/translations/hu/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hibrid RAG a Microsoft phi-4-mini és SQLite FTS5 (Zero-Cloud SLM) segítségével\n", + "\n", + "> **Szerző:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Fókusz:** Kis Nyelvi Modellek (SLM-ek), SQLite FTS5 BM25, Sűrű Beágyazások, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motiváció: A Kulcsszó Visszahívás Dilemmája a Helyi SLM-ekben\n", + "A szabványos RAG architektúrák, amelyek kizárólag sűrű vektorbeágyazásokra támaszkodnak, gyakran nem képesek pontos numerikus tokeneket visszakeresni (pl. `2,340,000 TL`, szerződéskódok, számlaszámok). \n", + "Ezzel szemben a ritkább lexikai keresés (BM25) nem találja meg a szemantikus szinonimákat és a parafrazált kérdéseket.\n", + "\n", + "Ez a szakácskönyv bemutatja, hogyan valósítható meg egy **nagy sebességű, memóriabeli hibrid keresőmotor**, amely kombinálja:\n", + "1. **Sűrű Vektorokat** (Koszinusz hasonlóság)\n", + "2. **Ritka Lexikai Keresést** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion-t (RRF, $k=60$)**\n", + "4. **Alapozott Hivatkozás Generálást (`[1]`, `[2]`)** a Microsoft `phi-4-mini` segítségével.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Kettős SQLite Séma (Sűrű Vektorok + Virtuális FTS5 BM25 tábla)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Minta bevitele és végrehajtási összehasonlítás\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Alapozott Parancsmegfogalmazás a Microsoft phi-4-mini számára\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Jogi nyilatkozat**:\nEz a dokumentum az AI fordítási szolgáltatás, a [Co-op Translator](https://github.com/Azure/co-op-translator) segítségével készült. Bár az pontosságra törekszünk, kérjük, vegye figyelembe, hogy az automatikus fordítások hibákat vagy pontatlanságokat tartalmazhatnak. Az eredeti dokumentum az anyanyelvén tekintendő hiteles forrásnak. Fontos információk esetén professzionális emberi fordítást javasolunk. Nem vállalunk felelősséget semmilyen félreértésért vagy téves értelmezésért, amely ebből a fordításból ered.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/id/.co-op-translator.json b/translations/id/.co-op-translator.json index 2ca559ded..9240a87be 100644 --- a/translations/id/.co-op-translator.json +++ b/translations/id/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "id" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:37:46+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:26:07+00:00", "source_file": "README.md", "language_code": "id" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "id" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:32:44+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "id" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:36+00:00", diff --git a/translations/id/README.md b/translations/id/README.md index d5e1f7ac5..1c118ecd1 100644 --- a/translations/id/README.md +++ b/translations/id/README.md @@ -1,29 +1,29 @@ -# Buku Resep Phi: Contoh Praktis dengan Model Phi dari Microsoft +# Phi Cookbook: Contoh Praktis dengan Model Phi dari Microsoft -[![Buka dan gunakan sampel di GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Buka dan gunakan contoh di GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Buka di Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![Kontributor GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![Masalah di GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Permintaan tarik di GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Masalah GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![Permintaan tarikan GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Selamat Datang](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![Pengamat GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![Pemantau GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![Forks GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![Bintang GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) -[![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) +[![Discord Microsoft Foundry](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi adalah serangkaian model AI open source yang dikembangkan oleh Microsoft. +Phi adalah serangkaian model AI sumber terbuka yang dikembangkan oleh Microsoft. -Phi saat ini merupakan model bahasa kecil (SLM) yang paling kuat dan hemat biaya, dengan tolok ukur yang sangat baik dalam multi-bahasa, penalaran, generasi teks/chat, pengkodean, gambar, audio, dan skenario lainnya. +Phi saat ini adalah model bahasa kecil (SLM) yang paling kuat dan hemat biaya, dengan tolok ukur yang sangat baik dalam multi-bahasa, penalaran, generasi teks/chat, pengkodean, gambar, audio, dan skenario lainnya. Anda dapat menerapkan Phi ke cloud atau perangkat edge, dan Anda dapat dengan mudah membangun aplikasi AI generatif dengan daya komputasi terbatas. Ikuti langkah-langkah ini untuk memulai menggunakan sumber daya ini: 1. **Fork Repositori**: Klik [![Forks GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Clone Repositori**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Bergabung dengan Komunitas Discord AI Microsoft dan temui para ahli serta sesama pengembang**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Bergabung dengan Komunitas AI Microsoft Discord dan temui para ahli serta pengembang lain**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/id/cover.eb18d1b9605d754b.webp) @@ -32,11 +32,11 @@ Ikuti langkah-langkah ini untuk memulai menggunakan sumber daya ini: #### Didukung melalui GitHub Action (Otomatis & Selalu Terbaru) -[Arab](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgaria](../bg/README.md) | [Burma (Myanmar)](../my/README.md) | [Cina (Sederhana)](../zh-CN/README.md) | [Cina (Tradisional, Hong Kong)](../zh-HK/README.md) | [Cina (Tradisional, Macau)](../zh-MO/README.md) | [Cina (Tradisional, Taiwan)](../zh-TW/README.md) | [Kroasia](../hr/README.md) | [Ceko](../cs/README.md) | [Denmark](../da/README.md) | [Belanda](../nl/README.md) | [Estonia](../et/README.md) | [Finlandia](../fi/README.md) | [Prancis](../fr/README.md) | [Jerman](../de/README.md) | [Yunani](../el/README.md) | [Ibrani](../he/README.md) | [Hindi](../hi/README.md) | [Hungaria](../hu/README.md) | [Indonesia](./README.md) | [Italia](../it/README.md) | [Jepang](../ja/README.md) | [Kannada](../kn/README.md) | [Kamboja](../km/README.md) | [Korea](../ko/README.md) | [Lituania](../lt/README.md) | [Melayu](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Pidgin Nigeria](../pcm/README.md) | [Norwegia](../no/README.md) | [Persia (Farsi)](../fa/README.md) | [Polandia](../pl/README.md) | [Portugis (Brasil)](../pt-BR/README.md) | [Portugis (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumania](../ro/README.md) | [Rusia](../ru/README.md) | [Serbia (Sirilik)](../sr/README.md) | [Slovakia](../sk/README.md) | [Slovenia](../sl/README.md) | [Spanyol](../es/README.md) | [Swahili](../sw/README.md) | [Swedia](../sv/README.md) | [Tagalog (Filipina)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turki](../tr/README.md) | [Ukraina](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnam](../vi/README.md) +[Arab](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgaria](../bg/README.md) | [Birma (Myanmar)](../my/README.md) | [Cina (Sederhana)](../zh-CN/README.md) | [Cina (Tradisional, Hong Kong)](../zh-HK/README.md) | [Cina (Tradisional, Macau)](../zh-MO/README.md) | [Cina (Tradisional, Taiwan)](../zh-TW/README.md) | [Kroasia](../hr/README.md) | [Ceko](../cs/README.md) | [Denmark](../da/README.md) | [Belanda](../nl/README.md) | [Estonia](../et/README.md) | [Finlandia](../fi/README.md) | [Prancis](../fr/README.md) | [Jerman](../de/README.md) | [Yunani](../el/README.md) | [Ibrani](../he/README.md) | [Hindi](../hi/README.md) | [Hungaria](../hu/README.md) | [Indonesia](./README.md) | [Italia](../it/README.md) | [Jepang](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korea](../ko/README.md) | [Lituania](../lt/README.md) | [Melayu](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Pidgin Nigeria](../pcm/README.md) | [Norwegia](../no/README.md) | [Persia (Farsi)](../fa/README.md) | [Polandia](../pl/README.md) | [Portugis (Brasil)](../pt-BR/README.md) | [Portugis (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumania](../ro/README.md) | [Rusia](../ru/README.md) | [Serbia (Sirilik)](../sr/README.md) | [Slovakia](../sk/README.md) | [Slovenia](../sl/README.md) | [Spanyol](../es/README.md) | [Swahili](../sw/README.md) | [Swedia](../sv/README.md) | [Tagalog (Filipina)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turki](../tr/README.md) | [Ukraina](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnam](../vi/README.md) -> **Lebih Suka Clone Secara Lokal?** +> **Lebih suka Clone secara Lokal?** > -> Repositori ini termasuk lebih dari 50 terjemahan bahasa yang secara signifikan meningkatkan ukuran unduhan. Untuk clone tanpa terjemahan, gunakan sparse checkout: +> Repositori ini menyertakan lebih dari 50 terjemahan bahasa yang secara signifikan meningkatkan ukuran unduhan. Untuk meng-clone tanpa terjemahan, gunakan sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -57,7 +57,7 @@ Ikuti langkah-langkah ini untuk memulai menggunakan sumber daya ini: ## Daftar Isi -- Pengantar +- Pengenalan - [Selamat datang di Keluarga Phi](./md/01.Introduction/01/01.PhiFamily.md) - [Menyiapkan lingkungan Anda](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Memahami Teknologi Kunci](./md/01.Introduction/01/01.Understandingtech.md) @@ -66,7 +66,7 @@ Ikuti langkah-langkah ini untuk memulai menggunakan sumber daya ini: - [Model Phi & Ketersediaan di berbagai platform](./md/01.Introduction/01/01.Edgeandcloud.md) - [Menggunakan Guidance-ai dan Phi](./md/01.Introduction/01/01.Guidance.md) - [Model Marketplace GitHub](https://github.com/marketplace/models) - - [Katalog Model Azure AI](https://ai.azure.com) + - [Katalog Model AI Azure](https://ai.azure.com) - Inferensi Phi di lingkungan berbeda - [Hugging face](./md/01.Introduction/02/01.HF.md) @@ -81,119 +81,121 @@ Ikuti langkah-langkah ini untuk memulai menggunakan sumber daya ini: - [Inferensi Phi di iOS](./md/01.Introduction/03/iOS_Inference.md) - [Inferensi Phi di Android](./md/01.Introduction/03/Android_Inference.md) - [Inferensi Phi di Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Inferensi Phi di AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Inferensi Phi dengan Kerangka MLX Apple](./md/01.Introduction/03/MLX_Inference.md) + - [Inferensi Phi di PC AI](./md/01.Introduction/03/AIPC_Inference.md) + - [Inferensi Phi dengan Kerangka Apple MLX](./md/01.Introduction/03/MLX_Inference.md) - [Inferensi Phi di Server Lokal](./md/01.Introduction/03/Local_Server_Inference.md) - [Inferensi Phi di Server Jarak Jauh menggunakan AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - [Inferensi Phi dengan Rust](./md/01.Introduction/03/Rust_Inference.md) - [Inferensi Phi--Vision di Lokal](./md/01.Introduction/03/Vision_Inference.md) - [Inferensi Phi dengan Kaito AKS, Azure Containers (dukungan resmi)](./md/01.Introduction/03/Kaito_Inference.md) -- [Quantifying Keluarga Phi](./md/01.Introduction/04/QuantifyingPhi.md) +- [Mengkuantisasi Keluarga Phi](./md/01.Introduction/04/QuantifyingPhi.md) - [Mengkuantisasi Phi-3.5 / 4 menggunakan llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - [Mengkuantisasi Phi-3.5 / 4 menggunakan ekstensi AI Generatif untuk onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [Mengkuantisasi Phi-3.5 / 4 menggunakan Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Mengkuantisasi Phi-3.5 / 4 menggunakan Kerangka MLX Apple](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Mengkuantisasi Phi-3.5 / 4 menggunakan Kerangka Apple MLX](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- Evaluasi Phi +- Evaluasi Phi - [AI yang Bertanggung Jawab](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry untuk Evaluasi](./md/01.Introduction/05/AIFoundry.md) - [Menggunakan Promptflow untuk Evaluasi](./md/01.Introduction/05/Promptflow.md) - RAG dengan Azure AI Search - [Cara menggunakan Phi-4-mini dan Phi-4-multimodal (RAG) dengan Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG Hibrida Lokal Tanpa Cloud dengan SQLite FTS5 dan phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Contoh pengembangan aplikasi Phi - Aplikasi Teks & Chat - Contoh Phi-4 - [📓] [Chat Dengan Model Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Chat dengan Model Phi-4 lokal ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Chat dengan Model Phi-4 Lokal ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - [Chat Aplikasi Konsol .NET dengan Phi-4 ONNX menggunakan Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Contoh Phi-3 / 3.5 + + - Phi-3 / 3.5 Sampel - [Chatbot Lokal di browser menggunakan Phi3, ONNX Runtime Web dan WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Chat OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [Multi Model - Interaktif Phi-3-mini dan OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Membangun pembungkus dan menggunakan Phi-3 dengan MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [MLFlow - Membuat pembungkus dan menggunakan Phi-3 dengan MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [Optimasi Model - Cara mengoptimalkan model Phi-3-min untuk ONNX Runtime Web dengan Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [Aplikasi WinUI3 dengan Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Contoh Aplikasi Catatan AI Multi Model WinUI3 Berdaya](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Fine-tune dan Integrasikan model Phi-3 kustom dengan Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Fine-tune dan Integrasikan model Phi-3 kustom dengan Prompt flow di Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Evaluasi Model Phi-3 / Phi-3.5 Fine-tuned di Microsoft Foundry dengan Fokus pada Prinsip AI Bertanggung Jawab Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + -[Contoh Aplikasi Catatan bertenaga AI Multi Model WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Fine-tune dan Integrasi model Phi-3 kustom dengan Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Fine-tune dan Integrasi model Phi-3 kustom dengan Prompt flow di Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Evaluasi Model Phi-3 / Phi-3.5 yang sudah di-Fine-tune di Microsoft Foundry dengan fokus pada Prinsip AI Bertanggung Jawab Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Contoh prediksi bahasa Phi-3.5-mini-instruct (Cina/Inggris)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Chatbot RAG WebGPU Phi-3.5-Instruct](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Menggunakan GPU Windows untuk membuat solusi Prompt flow dengan Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Menggunakan Microsoft Phi-3.5 tflite untuk membuat aplikasi Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Contoh Q&A .NET menggunakan model Phi-3 ONNX lokal menggunakan Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Contoh Q&A .NET menggunakan model ONNX Phi-3 lokal dengan Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - [Aplikasi chat konsol .NET dengan Semantic Kernel dan Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Contoh Kode Azure AI Inference SDK Based - - Contoh Phi-4 - - [📓] [Hasilkan kode proyek menggunakan Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Contoh Phi-3 / 3.5 - - [Bangun Visual Studio Code Chat GitHub Copilot Anda dengan Keluarga Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Buat Agen Chat GitHub Copilot Sendiri di Visual Studio Code dengan Phi-3.5 oleh Model GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - Contoh Penalaran Lanjut - - Contoh Phi-4 - - [📓] [Contoh Phi-4-mini-penalaran atau Phi-4-penalaran](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Fine-tuning Phi-4-mini-penalaran dengan Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Fine-tuning Phi-4-mini-penalaran dengan Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-penalaran dengan Model GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-penalaran dengan Model Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Sampel Kode SDK Azure AI Inference Berbasis Kode + - Sampel Phi-4 + - [📓] [Menghasilkan kode proyek menggunakan Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Sampel Phi-3 / 3.5 + - [Bangun Visual Studio Code GitHub Copilot Chat sendiri dengan Keluarga Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Buat Agen Chat Copilot Visual Studio Code sendiri dengan Phi-3.5 oleh Model GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - Sampel Penalaran Lanjutan + - Sampel Phi-4 + - [📓] [Sampel Phi-4-mini-reasoning atau Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Fine-tuning Phi-4-mini-reasoning dengan Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Fine-tuning Phi-4-mini-reasoning dengan Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-reasoning dengan Model GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-reasoning dengan Model Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demo - [Demo Phi-4-mini yang dihosting di Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Demo Phi-4-multimodal yang dihosting di Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Contoh Vision - - Contoh Phi-4 + - [Demo Phi-4-multimodal yang dihosting di Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Sampel Vision + - Sampel Phi-4 - [📓] [Gunakan Phi-4-multimodal untuk membaca gambar dan menghasilkan kode](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Contoh Phi-3 / 3.5 + - Sampel Phi-3 / 3.5 - [📓][Phi-3-vision-Teks gambar ke teks](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 Recycling](https://github.com/jennifermarsman/PhiRecycling/) + - [DEMO: Daur ulang Phi-3](https://github.com/jennifermarsman/PhiRecycling/) - [Phi-3-vision - Asisten bahasa visual - dengan Phi3-Vision dan OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Contoh Vision multi-frame atau multi-gambar Phi-3.5](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Model Lokal Phi-3 Vision ONNX menggunakan Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menu berbasis Model Lokal Phi-3 Vision ONNX menggunakan Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Contoh Phi-3.5 Vision multi-frame atau multi-gambar](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Model ONNX Lokal Phi-3 Vision menggunakan Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Model ONNX Lokal Phi-3 Vision berbasis menu menggunakan Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - Contoh Reasoning-Vision + - Sampel Reasoning-Vision - Phi-4-Reasoning-Vision-15B - [📓] [Menggunakan Phi-4-Reasoning-Vision-15B untuk mendeteksi jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [Menggunakan Phi-4-Reasoning-Vision-15B untuk matematika](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - [📓] [Menggunakan Phi-4-Reasoning-Vision-15B untuk mendeteksi UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - Contoh Matematika - - Contoh Phi-4-Mini-Flash-Reasoning-Instruct [Demo Matematika dengan Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Sampel Matematika + - Sampel Phi-4-Mini-Flash-Reasoning-Instruct [Demo Matematika dengan Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - Contoh Audio - - Contoh Phi-4 - - [📓] [Mengekstrak transkrip audio menggunakan Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Contoh Audio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Contoh Terjemahan Ucapan Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - Sampel Audio + - Sampel Phi-4 + - [📓] [Mengambil transkrip audio menggunakan Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Sampel Audio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Sampel Terjemahan Ucapan Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - [Aplikasi konsol .NET menggunakan Phi-4-multimodal Audio untuk menganalisis file audio dan menghasilkan transkrip](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - Contoh MOE - - Contoh Phi-3 / 3.5 - - [📓] [Contoh Media Sosial Model Mixture of Experts Phi-3.5 (MoEs)](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - Sampel MOE + - Sampel Phi-3 / 3.5 + - [📓] [Model Campuran Ahli Phi-3.5 (MoEs) Sampel Media Sosial](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - [📓] [Membangun Pipeline Retrieval-Augmented Generation (RAG) dengan NVIDIA NIM Phi-3 MOE, Azure AI Search, dan LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Contoh Pemanggilan Fungsi - - Contoh Phi-4 🆕 - - [📓] [Menggunakan Pemanggilan Fungsi dengan Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Menggunakan Pemanggilan Fungsi untuk membuat multi-agen dengan Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - Sampel Pemanggilan Fungsi + - Sampel Phi-4 🆕 + - [📓] [Menggunakan Pemanggilan Fungsi Dengan Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Menggunakan Pemanggilan Fungsi untuk membuat multi-agen Dengan Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Menggunakan Pemanggilan Fungsi dengan Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [Menggunakan Pemanggilan Fungsi dengan ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Contoh Pencampuran Multimodal - - Contoh Phi-4 🆕 - - [📓] [Menggunakan Phi-4-multimodal sebagai jurnalis teknologi](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - Sampel Campuran Multimodal + - Sampel Phi-4 🆕 + - [📓] [Menggunakan Phi-4-multimodal sebagai jurnalis Teknologi](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [Aplikasi konsol .NET menggunakan Phi-4-multimodal untuk menganalisis gambar](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Fine-tuning Contoh Phi - - [Skenario Fine-tuning](./md/03.FineTuning/FineTuning_Scenarios.md) +- Fine-tuning Sampel Phi + - [Skema Fine-tuning](./md/03.FineTuning/FineTuning_Scenarios.md) - [Fine-tuning vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Fine-tuning Membuat Phi-3 Menjadi Pakar Industri](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Fine-tuning Biarkan Phi-3 menjadi ahli industri](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - [Fine-tuning Phi-3 dengan AI Toolkit untuk VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [Fine-tuning Phi-3 dengan Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - [Fine-tuning Phi-3 dengan Lora](./md/03.FineTuning/FineTuning_Lora.md) @@ -201,53 +203,54 @@ Ikuti langkah-langkah ini untuk memulai menggunakan sumber daya ini: - [Fine-tuning Phi-3 dengan Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - [Fine-tuning Phi-3 dengan Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - [Fine-tuning dengan Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Fine-tuning dengan Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) + - [Lab Praktik Fine-tuning dengan Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - [Fine-tuning Phi-3-vision dengan Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Fine-tuning Phi-3 dengan Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - [Fine-tuning Phi-3-vision (dukungan resmi)](./md/03.FineTuning/FineTuning_Vision.md) - - [Fine-Tuning Phi-3 dengan Kaito AKS, Kontainer Azure (Dukungan Resmi)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Fine-Tuning Phi-3 dengan Kaito AKS , Azure Containers(Dukungan Resmi)](./md/03.FineTuning/FineTuning_Kaito.md) - [Fine-Tuning Phi-3 dan 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) -- Hands on Lab - - [Menjelajahi model terkini: LLM, SLM, pengembangan lokal dan lainnya](https://github.com/microsoft/aitour-exploring-cutting-edge-models) +- Laboratorium Praktik + - [Menjelajahi model mutakhir: LLMs, SLMs, pengembangan lokal dan lainnya](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - [Membuka Potensi NLP: Fine-Tuning dengan Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Makalah dan Publikasi Penelitian Akademik +- Makalah dan Publikasi Riset Akademik - [Textbooks Are All You Need II: laporan teknis phi-1.5](https://arxiv.org/abs/2309.05463) - - [Laporan Teknis Phi-3: Model Bahasa Berkapasitas Tinggi secara Lokal di Ponsel Anda](https://arxiv.org/abs/2404.14219) + - [Laporan Teknis Phi-3: Model Bahasa Berkinerja Tinggi Secara Lokal di Ponsel Anda](https://arxiv.org/abs/2404.14219) - [Laporan Teknis Phi-4](https://arxiv.org/abs/2412.08905) - - [Laporan Teknis Phi-4-Mini: Model Bahasa Multimodal Kompak namun Kuat melalui Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Mengoptimalkan Model Bahasa Kecil untuk Pemanggilan Fungsi Dalam Kendaraan](https://arxiv.org/abs/2501.02342) + - [Laporan Teknis Phi-4-Mini: Model Bahasa Multimodal Ringkas namun Kuat melalui Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Optimasi Model Bahasa Kecil untuk Pemanggilan Fungsi Dalam Kendaraan](https://arxiv.org/abs/2501.02342) - [(WhyPHI) Fine-Tuning PHI-3 untuk Menjawab Pertanyaan Pilihan Ganda: Metodologi, Hasil, dan Tantangan](https://arxiv.org/abs/2501.01588) - [Laporan Teknis Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Laporan Teknis Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Menggunakan Model Phi -### Phi di Microsoft Foundry +### Phi pada Microsoft Foundry -Anda dapat belajar cara menggunakan Microsoft Phi dan cara membangun solusi E2E di berbagai perangkat keras Anda. Untuk mengalami Phi secara langsung, mulailah dengan mencoba model dan menyesuaikan Phi untuk skenario Anda menggunakan [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Anda dapat mempelajari lebih lanjut pada Memulai dengan [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Anda dapat belajar cara menggunakan Microsoft Phi dan cara membangun solusi end-to-end di berbagai perangkat keras Anda. Untuk mengalami Phi secara langsung, mulai dengan bereksperimen dengan model dan menyesuaikan Phi untuk skenario Anda menggunakan [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) Anda dapat mempelajari lebih lanjut pada Memulai dengan [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Playground** -Setiap model memiliki playground khusus untuk menguji model di [Azure AI Playground](https://aka.ms/try-phi3). +**Tempat Bermain** +Setiap model memiliki tempat bermain khusus untuk menguji model [Azure AI Playground](https://aka.ms/try-phi3). -### Phi di Model GitHub +### Phi pada Model GitHub -Anda dapat belajar cara menggunakan Microsoft Phi dan cara membangun solusi E2E di berbagai perangkat keras Anda. Untuk mengalami Phi secara langsung, mulailah dengan mencoba model dan menyesuaikan Phi untuk skenario Anda menggunakan [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Anda dapat mempelajari lebih lanjut pada Memulai dengan [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Anda dapat belajar cara menggunakan Microsoft Phi dan cara membangun solusi end-to-end di berbagai perangkat keras Anda. Untuk mengalami Phi secara langsung, mulai dengan bereksperimen dengan model dan menyesuaikan Phi untuk skenario Anda menggunakan [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) Anda dapat mempelajari lebih lanjut pada Memulai dengan [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Playground** -Setiap model memiliki [playground khusus untuk menguji model](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Tempat Bermain** +Setiap model memiliki [tempat bermain khusus untuk menguji model](/md/02.QuickStart/GitHubModel_QuickStart.md). -### Phi di Hugging Face +### Phi pada Hugging Face Anda juga dapat menemukan model di [Hugging Face](https://huggingface.co/microsoft) -**Playground** -[Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Tempat Bermain** + [Tempat bermain Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Kursus Lainnya + ## 🎒 Kursus Lainnya -Tim kami memproduksi kursus lainnya! Lihat di: +Tim kami juga memproduksi kursus lain! Lihat: ### LangChain @@ -256,14 +259,14 @@ Tim kami memproduksi kursus lainnya! Lihat di: [![LangChain untuk Pemula](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agents +### Azure / Edge / MCP / Agen [![AZD untuk Pemula](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI untuk Pemula](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP untuk Pemula](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents untuk Pemula](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Agen AI untuk Pemula](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Seri AI Generatif [![AI Generatif untuk Pemula](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![AI Generatif (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) @@ -271,10 +274,10 @@ Tim kami memproduksi kursus lainnya! Lihat di: [![AI Generatif (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - + ### Pembelajaran Inti [![ML untuk Pemula](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science untuk Pemula](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![Ilmu Data untuk Pemula](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI untuk Pemula](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![Keamanan Siber untuk Pemula](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Pengembangan Web untuk Pemula](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) @@ -282,43 +285,45 @@ Tim kami memproduksi kursus lainnya! Lihat di: [![Pengembangan XR untuk Pemula](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Seri Copilot -[![Copilot untuk Pemrograman Berpasangan AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot untuk Pemrograman Pasangan AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot untuk C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Petualangan Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## AI yang Bertanggung Jawab +## AI Bertanggung Jawab + +Microsoft berkomitmen membantu pelanggan kami menggunakan produk AI kami secara bertanggung jawab, berbagi pembelajaran kami, dan membangun kemitraan berdasarkan kepercayaan melalui alat seperti Transparency Notes dan Impact Assessments. Banyak sumber daya ini dapat ditemukan di [https://aka.ms/RAI](https://aka.ms/RAI). +Pendekatan Microsoft terhadap AI yang bertanggung jawab didasarkan pada prinsip AI kami yaitu keadilan, keandalan dan keamanan, privasi dan keamanan, inklusivitas, transparansi, dan akuntabilitas. + +Model bahasa, gambar, dan suara berskala besar - seperti yang digunakan dalam contoh ini - berpotensi berperilaku dengan cara yang tidak adil, tidak dapat diandalkan, atau menyinggung, yang dapat menyebabkan kerugian. Silakan konsultasikan [catatan transparansi layanan Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) untuk mengetahui risiko dan keterbatasannya. -Microsoft berkomitmen membantu pelanggan kami menggunakan produk AI kami dengan bertanggung jawab, berbagi pembelajaran, dan membangun kemitraan berbasis kepercayaan melalui alat seperti Transparency Notes dan Impact Assessments. Banyak sumber daya ini dapat ditemukan di [https://aka.ms/RAI](https://aka.ms/RAI). -Pendekatan Microsoft terhadap AI yang bertanggung jawab didasarkan pada prinsip AI kami yaitu keadilan, keandalan dan keselamatan, privasi dan keamanan, inklusivitas, transparansi, dan akuntabilitas. -Model skala besar untuk bahasa, gambar, dan suara - seperti yang digunakan dalam contoh ini - berpotensi berperilaku dengan cara yang tidak adil, tidak dapat diandalkan, atau menyinggung, yang pada akhirnya menyebabkan kerugian. Harap tinjau [catatan transparansi layanan Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) untuk mengetahui risiko dan keterbatasannya. -Pendekatan yang direkomendasikan untuk mengurangi risiko ini adalah dengan memasukkan sistem keamanan dalam arsitektur Anda yang dapat mendeteksi dan mencegah perilaku berbahaya. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) menyediakan lapisan perlindungan independen, mampu mendeteksi konten berbahaya yang dihasilkan oleh pengguna dan AI dalam aplikasi serta layanan. Azure AI Content Safety mencakup API teks dan gambar yang memungkinkan Anda mendeteksi materi yang berbahaya. Dalam Microsoft Foundry, layanan Content Safety memungkinkan Anda melihat, menjelajahi, dan mencoba contoh kode untuk mendeteksi konten berbahaya di berbagai modalitas. Dokumentasi [quickstart berikut](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) memandu Anda untuk membuat permintaan ke layanan tersebut. +Pendekatan yang direkomendasikan untuk mengurangi risiko ini adalah dengan menyertakan sistem keamanan dalam arsitektur Anda yang dapat mendeteksi dan mencegah perilaku berbahaya. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) menyediakan lapisan perlindungan independen, mampu mendeteksi konten berbahaya yang dihasilkan oleh pengguna dan AI dalam aplikasi dan layanan. Azure AI Content Safety mencakup API teks dan gambar yang memungkinkan Anda mendeteksi materi yang berbahaya. Dalam Microsoft Foundry, layanan Content Safety memungkinkan Anda untuk melihat, mengeksplorasi, dan mencoba kode contoh untuk mendeteksi konten berbahaya di berbagai modalitas. Dokumentasi [quickstart berikut](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) membimbing Anda melalui proses membuat permintaan ke layanan tersebut. -Aspek lain yang perlu diperhatikan adalah performa keseluruhan aplikasi. Dengan aplikasi multi-modal dan multi-model, kami menganggap performa berarti sistem bekerja seperti yang Anda dan pengguna Anda harapkan, termasuk tidak menghasilkan keluaran yang berbahaya. Penting untuk menilai kinerja aplikasi keseluruhan Anda menggunakan [Penilai Performa dan Kualitas serta Risiko dan Keamanan](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Anda juga memiliki kemampuan untuk membuat dan menilai dengan [penilai kustom](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Aspek lain yang perlu diperhatikan adalah kinerja aplikasi secara keseluruhan. Dengan aplikasi multi-modal dan multi-model, kami menganggap kinerja berarti sistem bekerja sebagaimana Anda dan pengguna Anda harapkan, termasuk tidak menghasilkan keluaran yang berbahaya. Penting untuk menilai kinerja aplikasi Anda secara keseluruhan menggunakan [evaluasi Kinerja dan Kualitas serta Risiko dan Keamanan](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Anda juga memiliki kemampuan untuk membuat dan mengevaluasi dengan [evaluator khusus](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Anda dapat mengevaluasi aplikasi AI Anda di lingkungan pengembangan Anda menggunakan [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Dengan dataset pengujian atau target, generasi aplikasi AI generatif Anda diukur secara kuantitatif dengan penilai bawaan atau penilai kustom pilihan Anda. Untuk memulai dengan azure ai evaluation sdk untuk mengevaluasi sistem Anda, Anda dapat mengikuti [panduan quickstart](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Setelah Anda menjalankan evaluasi, Anda dapat [memvisualisasikan hasilnya di Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Anda dapat mengevaluasi aplikasi AI Anda di lingkungan pengembangan menggunakan [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Dengan dataset uji atau target, generasi aplikasi AI generatif Anda diukur secara kuantitatif dengan evaluator bawaan atau evaluator khusus pilihan Anda. Untuk memulai dengan Azure AI Evaluation SDK guna mengevaluasi sistem Anda, Anda dapat mengikuti [panduan quickstart](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Setelah Anda menjalankan evaluasi, Anda dapat [memvisualisasikan hasilnya di Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Merek Dagang -Proyek ini mungkin mengandung merek dagang atau logo untuk proyek, produk, atau layanan. Penggunaan resmi merek dagang atau logo Microsoft tunduk pada dan harus mengikuti [Pedoman Merek Dagang & Merek Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Penggunaan merek dagang atau logo Microsoft pada versi modifikasi proyek ini tidak boleh menyebabkan kebingungan atau mengindikasikan sponsor Microsoft. Setiap penggunaan merek dagang atau logo pihak ketiga tunduk pada kebijakan pihak ketiga tersebut. +Proyek ini mungkin berisi merek dagang atau logo untuk proyek, produk, atau layanan. Penggunaan resmi merek dagang atau logo Microsoft tunduk pada dan harus mengikuti [Panduan Merek Dagang & Merek Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Penggunaan merek dagang atau logo Microsoft dalam versi modifikasi proyek ini tidak boleh menimbulkan kebingungan atau mengimplikasikan sponsor Microsoft. Penggunaan merek dagang atau logo pihak ketiga tunduk pada kebijakan pihak ketiga tersebut. ## Mendapatkan Bantuan -Jika Anda mengalami kesulitan atau memiliki pertanyaan tentang membangun aplikasi AI, bergabunglah dengan: +Jika Anda mengalami hambatan atau memiliki pertanyaan tentang membangun aplikasi AI, bergabunglah di: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Jika Anda memiliki umpan balik produk atau menemukan kesalahan saat membangun, kunjungi: +Jika Anda memiliki masukan produk atau menemui kesalahan selama pembangunan, kunjungi: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Penafian**: -Dokumen ini telah diterjemahkan menggunakan layanan terjemahan AI [Co-op Translator](https://github.com/Azure/co-op-translator). Meskipun kami berupaya untuk akurasi, harap diingat bahwa terjemahan otomatis mungkin mengandung kesalahan atau ketidakakuratan. Dokumen asli dalam bahasa aslinya harus dianggap sebagai sumber yang sahih. Untuk informasi penting, disarankan menggunakan terjemahan profesional oleh manusia. Kami tidak bertanggung jawab atas kesalahpahaman atau kesalahan interpretasi yang timbul dari penggunaan terjemahan ini. +**Penafian**: +Dokumen ini telah diterjemahkan menggunakan layanan terjemahan AI [Co-op Translator](https://github.com/Azure/co-op-translator). Meskipun kami berupaya untuk mencapai akurasi, harap diketahui bahwa terjemahan otomatis mungkin mengandung kesalahan atau ketidakakuratan. Dokumen asli dalam bahasa aslinya harus dianggap sebagai sumber yang sah. Untuk informasi penting, disarankan menggunakan terjemahan profesional oleh manusia. Kami tidak bertanggung jawab atas kesalahpahaman atau penafsiran yang keliru yang timbul dari penggunaan terjemahan ini. \ No newline at end of file diff --git a/translations/id/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/id/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..09d8d7da3 --- /dev/null +++ b/translations/id/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrid RAG dengan Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Penulis:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Fokus:** Small Language Models (SLMs), SQLite FTS5 BM25, Dense Embeddings, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivasi: Dilema Pengambilan Kata Kunci pada SLM Lokal\n", + "Arsitektur RAG standar yang hanya mengandalkan dense vector embeddings sering gagal mengambil token numerik tepat (misalnya, `2,340,000 TL`, kode kontrak, nomor rekening). \n", + "Sebaliknya, pencarian leksikal jarang (BM25) melewatkan sinonim semantik dan pertanyaan yang diparafrase.\n", + "\n", + "Cookbook ini menunjukkan cara mengimplementasikan **mesin Pengambilan hybrid berkecepatan tinggi dalam memori** yang menggabungkan:\n", + "1. **Dense Vectors** (Kesamaan Kosinus)\n", + "2. **Pencarian Leksikal Jarang** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Generasi Sitasi Terbentuk (`[1]`, `[2]`)** dengan Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Skema SQLite Ganda (Vektor Padat + Tabel Virtual FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Sampel Pengambilan & Tolok Ukur Eksekusi\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Formulasi Prompt Berasas untuk Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Penafian**:\nDokumen ini telah diterjemahkan menggunakan layanan terjemahan AI [Co-op Translator](https://github.com/Azure/co-op-translator). Meskipun kami berupaya untuk mencapai akurasi, harap diketahui bahwa terjemahan otomatis mungkin mengandung kesalahan atau ketidakakuratan. Dokumen asli dalam bahasa aslinya harus dianggap sebagai sumber yang sah. Untuk informasi penting, disarankan menggunakan terjemahan profesional oleh manusia. Kami tidak bertanggung jawab atas kesalahpahaman atau penafsiran yang keliru yang timbul dari penggunaan terjemahan ini.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/it/.co-op-translator.json b/translations/it/.co-op-translator.json index cd83c8fda..74ce2758c 100644 --- a/translations/it/.co-op-translator.json +++ b/translations/it/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "it" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:48:38+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:00:35+00:00", "source_file": "README.md", "language_code": "it" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "it" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:06:44+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "it" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:44+00:00", diff --git a/translations/it/README.md b/translations/it/README.md index 4f8af06df..5ca827cc6 100644 --- a/translations/it/README.md +++ b/translations/it/README.md @@ -1,42 +1,42 @@ -# Ricettario Phi: Esempi Pratici con i Modelli Phi di Microsoft +# Phi Cookbook: Esempi Pratici con i Modelli Phi di Microsoft [![Apri e usa gli esempi in GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Apri in Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![Contributori su GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![Contributori GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![Problemi su GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) [![Richieste di pull su GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Benvenuti](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Benvenute](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![Watcher su GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![Osservatori GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![Fork su GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![Stelle su GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi è una serie di modelli di intelligenza artificiale open source sviluppati da Microsoft. +Phi è una serie di modelli AI open source sviluppati da Microsoft. -Phi è attualmente il modello linguistico piccolo (SLM) più potente e conveniente, con ottimi risultati in benchmark multi-lingua, ragionamento, generazione di testo/chat, programmazione, immagini, audio e altri scenari. +Attualmente Phi è il modello linguistico piccolo (SLM) più potente ed economico, con ottime prestazioni in molte lingue, ragionamento, generazione di testo/chat, programmazione, immagini, audio e altri scenari. -Puoi distribuire Phi nel cloud o su dispositivi edge, e puoi facilmente costruire applicazioni di IA generativa con potenza di calcolo limitata. +Puoi distribuire Phi nel cloud o su dispositivi edge, e puoi facilmente costruire applicazioni di AI generativa con potenza di calcolo limitata. Segui questi passaggi per iniziare a usare queste risorse: -1. **Forka il Repository**: Clicca su [![Fork su GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +1. **Forka il Repository**: Clicca [![Fork su GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Clona il Repository**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Unisciti alla Comunità Microsoft AI Discord e incontra esperti e altri sviluppatori**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Unisciti alla Community Microsoft AI su Discord e incontra esperti e sviluppatori**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![copertina](../../translated_images/it/cover.eb18d1b9605d754b.webp) ### 🌐 Supporto Multilingue -#### Supportato tramite GitHub Action (Automatizzato & Sempre Aggiornato) +#### Supportato tramite GitHub Action (Automatizzato e Sempre Aggiornato) [Arabo](../ar/README.md) | [Bengalese](../bn/README.md) | [Bulgaro](../bg/README.md) | [Birmano (Myanmar)](../my/README.md) | [Cinese (Semplificato)](../zh-CN/README.md) | [Cinese (Tradizionale, Hong Kong)](../zh-HK/README.md) | [Cinese (Tradizionale, Macao)](../zh-MO/README.md) | [Cinese (Tradizionale, Taiwan)](../zh-TW/README.md) | [Croato](../hr/README.md) | [Ceco](../cs/README.md) | [Danese](../da/README.md) | [Olandese](../nl/README.md) | [Estone](../et/README.md) | [Finlandese](../fi/README.md) | [Francese](../fr/README.md) | [Tedesco](../de/README.md) | [Greco](../el/README.md) | [Ebraico](../he/README.md) | [Hindi](../hi/README.md) | [Ungherese](../hu/README.md) | [Indonesiano](../id/README.md) | [Italiano](./README.md) | [Giapponese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Coreano](../ko/README.md) | [Lituano](../lt/README.md) | [Malese](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalese](../ne/README.md) | [Pidgin Nigeriano](../pcm/README.md) | [Norvegese](../no/README.md) | [Persiano (Farsi)](../fa/README.md) | [Polacco](../pl/README.md) | [Portoghese (Brasile)](../pt-BR/README.md) | [Portoghese (Portogallo)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumeno](../ro/README.md) | [Russo](../ru/README.md) | [Serbo (Cirillico)](../sr/README.md) | [Slovacco](../sk/README.md) | [Sloveno](../sl/README.md) | [Spagnolo](../es/README.md) | [Swahili](../sw/README.md) | [Svedese](../sv/README.md) | [Tagalog (Filippino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thailandese](../th/README.md) | [Turco](../tr/README.md) | [Ucraino](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamita](../vi/README.md) > **Preferisci clonare localmente?** > -> Questo repository include traduzioni in oltre 50 lingue che aumentano significativamente la dimensione del download. Per clonare senza traduzioni, usa il sparse checkout: +> Questo repository include più di 50 traduzioni che aumentano significativamente la dimensione del download. Per clonare senza traduzioni, usa sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -59,16 +59,16 @@ Segui questi passaggi per iniziare a usare queste risorse: - Introduzione - [Benvenuto nella Famiglia Phi](./md/01.Introduction/01/01.PhiFamily.md) - - [Configurare l'ambiente](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Configurazione dell'ambiente](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Comprendere le Tecnologie Chiave](./md/01.Introduction/01/01.Understandingtech.md) - - [Sicurezza IA per i Modelli Phi](./md/01.Introduction/01/01.AISafety.md) + - [Sicurezza AI per i Modelli Phi](./md/01.Introduction/01/01.AISafety.md) - [Supporto Hardware Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Modelli Phi & Disponibilità attraverso le piattaforme](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Modelli Phi & Disponibilità sulle piattaforme](./md/01.Introduction/01/01.Edgeandcloud.md) - [Usare Guidance-ai e Phi](./md/01.Introduction/01/01.Guidance.md) - - [Modelli GitHub Marketplace](https://github.com/marketplace/models) - - [Catalogo Modelli AI Azure](https://ai.azure.com) + - [Modelli su GitHub Marketplace](https://github.com/marketplace/models) + - [Catalogo Modelli AI di Azure](https://ai.azure.com) -- Inference Phi in diversi ambienti +- Inferenza Phi in diversi ambienti - [Hugging face](./md/01.Introduction/02/01.HF.md) - [Modelli GitHub](./md/01.Introduction/02/02.GitHubModel.md) - [Catalogo Modelli Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) @@ -77,123 +77,125 @@ Segui questi passaggi per iniziare a usare queste risorse: - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Locale](./md/01.Introduction/02/07.FoundryLocal.md) -- Famiglia Phi in inferenza +- Inferenza Phi Family - [Inferenza Phi su iOS](./md/01.Introduction/03/iOS_Inference.md) - [Inferenza Phi su Android](./md/01.Introduction/03/Android_Inference.md) - [Inferenza Phi su Jetson](./md/01.Introduction/03/Jetson_Inference.md) - [Inferenza Phi su AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Inferenza Phi con il framework Apple MLX](./md/01.Introduction/03/MLX_Inference.md) + - [Inferenza Phi con Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - [Inferenza Phi su Server Locale](./md/01.Introduction/03/Local_Server_Inference.md) - [Inferenza Phi su Server Remoto usando AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - [Inferenza Phi con Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inferenza Phi–Visione in Locale](./md/01.Introduction/03/Vision_Inference.md) - - [Inferenza Phi con Kaito AKS, contenitori Azure (supporto ufficiale)](./md/01.Introduction/03/Kaito_Inference.md) -- [Quantizzazione della Famiglia Phi](./md/01.Introduction/04/QuantifyingPhi.md) + - [Inferenza Phi--Visione su Locale](./md/01.Introduction/03/Vision_Inference.md) + - [Inferenza Phi con Kaito AKS, Azure Containers (supporto ufficiale)](./md/01.Introduction/03/Kaito_Inference.md) +- [Quantizzazione Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) - [Quantizzare Phi-3.5 / 4 usando llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - [Quantizzare Phi-3.5 / 4 usando estensioni Generative AI per onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Quantizzare Phi-3.5 / 4 usando Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Quantizzare Phi-3.5 / 4 usando Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Quantizzare Phi-3.5 / 4 usando Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Quantizzare Phi-3.5 / 4 usando Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Valutazione Phi - - [IA Responsabile](./md/01.Introduction/05/ResponsibleAI.md) + - [Responsabilità AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry per la Valutazione](./md/01.Introduction/05/AIFoundry.md) - [Usare Promptflow per la Valutazione](./md/01.Introduction/05/Promptflow.md) - + - RAG con Azure AI Search - - [Come usare Phi-4-mini e Phi-4-multimodal(RAG) con Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Come usare Phi-4-mini e Phi-4-multimodal (RAG) con Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG Ibrido Locale Zero-Cloud con SQLite FTS5 e phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Esempi di sviluppo applicazioni Phi - - Applicazioni di Testo & Chat + - Applicazioni testo e chat - Esempi Phi-4 - - [📓] [Chat con il modello ONNX Phi-4-mini](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Chat con modello ONNX Phi-4 locale .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [App Console Chat .NET con Phi-4 ONNX usando Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Esempi Phi-3 / 3.5 - - [Chatbot locale nel browser usando Phi3, ONNX Runtime Web e WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - Phi-3-mini interattivo e OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [📓] [Chat con modello Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Chat con modello Phi-4 locale ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [App console Chat .NET con Phi-4 ONNX usando Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Esempi + - [Chatbot Locale nel browser usando Phi3, ONNX Runtime Web e WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Chat OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Multi Modello - Phi-3-mini interattivo e OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Creare un wrapper e usare Phi-3 con MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Ottimizzazione del modello - Come ottimizzare il modello Phi-3-min per ONNX Runtime Web con Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Ottimizzazione Modello - Come ottimizzare il modello Phi-3-min per ONNX Runtime Web con Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [App WinUI3 con Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Esempio di app WinUI3 Multi Model AI Powered Notes](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Messa a punto e integrazione di modelli Phi-3 personalizzati con Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Messa a punto e integrazione di modelli Phi-3 personalizzati con Prompt flow in Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Valutare il modello Fine-tuned Phi-3 / Phi-3.5 in Microsoft Foundry concentrandosi sui Principi di AI Responsabile di Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Esempio di previsione linguistica Phi-3.5-mini-instruct (Cinese/Inglese)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Uso della GPU di Windows per creare una soluzione Prompt flow con Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Uso di Microsoft Phi-3.5 tflite per creare un'app Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Esempio Q&A .NET che usa modello ONNX Phi-3 locale usando Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + -[Esempio App Note AI Multimodello WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Fine-tuning e integrazione modelli personalizzati Phi-3 con Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Fine-tuning e integrazione modelli personalizzati Phi-3 con Prompt flow in Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Valutare il modello Fine-tuned Phi-3 / Phi-3.5 in Microsoft Foundry concentrandosi sui principi di AI responsabile di Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Esempio di predizione linguistica Phi-3.5-mini-instruct (Cinese/Inglese)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Chatbot RAG Phi-3.5-Instruct WebGPU](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Usare GPU Windows per creare soluzioni Prompt flow con Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Usare Phi-3.5 tflite di Microsoft per creare app Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Esempio Q&A .NET usando modello ONNX Phi-3 locale con Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - [App console chat .NET con Semantic Kernel e Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Esempi di Azure AI Inference SDK basati su codice + - Esempi SDK Azure AI Inference basati su codice - Esempi Phi-4 - [📓] [Genera codice di progetto usando Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Esempi Phi-3 / 3.5 - - [Costruisci il tuo GitHub Copilot Chat per Visual Studio Code con Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Crea il tuo agente Chat Copilot per Visual Studio Code con Phi-3.5 usando modelli GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Crea il tuo Visual Studio Code GitHub Copilot Chat con Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Crea il tuo agente Chat Copilot per Visual Studio Code con Phi-3.5 tramite modelli GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Esempi di ragionamento avanzato - Esempi Phi-4 - - [📓] [Esempi Phi-4-mini-ragionamento o Phi-4-ragionamento](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Messa a punto di Phi-4-mini-ragionamento con Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Messa a punto di Phi-4-mini-ragionamento con Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-ragionamento con modelli GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-ragionamento con modelli Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [📓] [Esempi di ragionamento con Phi-4-mini o Phi-4](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Fine-tuning di Phi-4-mini-reasoning con Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Fine-tuning di Phi-4-mini-reasoning con Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-reasoning con modelli GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-reasoning con modelli Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demo - - [Demo Phi-4-mini ospitate su Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Demo Phi-4-multimodal ospitate su Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [Demo Phi-4-mini ospitati su Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Demo Phi-4-multimodal ospitati su Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - Esempi Vision - Esempi Phi-4 - [📓] [Usa Phi-4-multimodal per leggere immagini e generare codice](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Esempi Phi-3 / 3.5 - - [📓][Phi-3-vision Immagine testo a testo](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-Testo immagine a testo](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 Recycling](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Assistente visivo linguistico - con Phi3-Vision e OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [DEMO: Riciclaggio Phi-3](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - Assistente linguaggio visivo - con Phi3-Vision e OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Esempio Phi-3.5 Vision multi-frame o multi-image](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision modello ONNX locale usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Modello ONNX locale Phi-3 Vision basato su menu usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Esempio multi-frame o multi-immagine Phi-3.5 Vision](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision modello locale ONNX usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Modello locale ONNX Phi-3 Vision basato su menu usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - Esempi Reasoning-Vision - - Phi-4-Reasoning-Vision-15B - - [📓] [Usare Phi-4-Reasoning-Vision-15B per rilevare il attraversamento pedonale improprio](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Usare Phi-4-Reasoning-Vision-15B per matematica](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Usare Phi-4-Reasoning-Vision-15B per rilevare UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - Esempi Ragionamento-Visione + - Phi-4-Ragionamento-Visione-15B + - [📓] [Usare Phi-4-Ragionamento-Visione-15B per rilevare attraversamenti pedonali irregolari](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Usare Phi-4-Ragionamento-Visione-15B per la matematica](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Usare Phi-4-Ragionamento-Visione-15B per rilevare UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - Esempi Math - - Esempi Phi-4-Mini-Flash-Reasoning-Instruct [Demo matematica con Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Esempi Matematica + - Esempi Phi-4-Mini-Flash-Reasoning-Instruct [Demo Matematica con Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - Esempi Audio - Esempi Phi-4 - - [📓] [Estrazione di trascrizioni audio usando Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Esempio audio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Esempio di traduzione vocale Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [Applicazione console .NET usando Phi-4-multimodal Audio per analizzare un file audio e generare trascrizione](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [📓] [Estrazione trascrizioni audio usando Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Esempio Audio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Esempio traduzione vocale Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [App console .NET usando Phi-4-multimodal Audio per analizzare file audio e generare trascrizione](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - Esempi MOE - Esempi Phi-3 / 3.5 - - [📓] [Modelli Mixture of Experts Phi-3.5 (MoEs) esempio Social Media](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Costruire una pipeline Retrieval-Augmented Generation (RAG) con NVIDIA NIM Phi-3 MOE, Azure AI Search e LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Esempio Social Media modelli Phi-3.5 Mixture of Experts (MoEs)](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Costruire pipeline Retrieval-Augmented Generation (RAG) con NVIDIA NIM Phi-3 MOE, Azure AI Search e LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - Esempi Function Calling - Esempi Phi-4 🆕 - - [📓] [Utilizzo di Function Calling con Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Utilizzo di Function Calling per creare multi-agenti con Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Utilizzo di Function Calling con Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Utilizzo di Function Calling con ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Esempi Multimodal Mixing + - [📓] [Uso di Function Calling con Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Uso di Function Calling per creare multi-agenti con Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Uso di Function Calling con Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Uso di Function Calling con ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Esempi Mixed Multimodal - Esempi Phi-4 🆕 - - [📓] [Utilizzo di Phi-4-multimodal come giornalista tecnologico](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [Applicazione console .NET che utilizza Phi-4-multimodal per analizzare immagini](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [📓] [Uso di Phi-4-multimodal come giornalista tecnologico](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [App console .NET usando Phi-4-multimodal per analizzare immagini](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Esempi di Fine-tuning Phi +- Fine-tuning Phi Esempi - [Scenari di fine-tuning](./md/03.FineTuning/FineTuning_Scenarios.md) - [Fine-tuning vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Messa a punto: lascia che Phi-3 diventi un esperto industriale](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Fine-tuning: lascia che Phi-3 diventi un esperto di settore](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - [Fine-tuning Phi-3 con AI Toolkit per VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [Fine-tuning Phi-3 con Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - [Fine-tuning Phi-3 con Lora](./md/03.FineTuning/FineTuning_Lora.md) @@ -203,49 +205,50 @@ Segui questi passaggi per iniziare a usare queste risorse: - [Fine-tuning con Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Laboratorio pratico di fine-tuning con Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - [Fine-tuning Phi-3-vision con Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Fine-tuning Phi-3 con Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Fine-tuning Phi-3-vision (supporto ufficiale)](./md/03.FineTuning/FineTuning_Vision.md) - - [Fine-Tuning Phi-3 con Kaito AKS , Contenitori Azure (Supporto ufficiale)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Fine-Tuning Phi-3 e 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + + - [Messa a punto di Phi-3 con Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Messa a punto di Phi-3-vision (supporto ufficiale)](./md/03.FineTuning/FineTuning_Vision.md) + - [Messa a punto di Phi-3 con Kaito AKS, Azure Containers (supporto ufficiale)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Messa a punto di Phi-3 e 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Laboratorio pratico - [Esplorare modelli all'avanguardia: LLM, SLM, sviluppo locale e altro](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Sbloccare il potenziale NLP: Fine-Tuning con Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Sbloccare il potenziale NLP: Messa a punto con Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Articoli di ricerca accademica e pubblicazioni +- Articoli e pubblicazioni di ricerca accademica - [Textbooks Are All You Need II: rapporto tecnico phi-1.5](https://arxiv.org/abs/2309.05463) - [Rapporto tecnico Phi-3: un modello linguistico altamente capace localmente sul tuo telefono](https://arxiv.org/abs/2404.14219) - [Rapporto tecnico Phi-4](https://arxiv.org/abs/2412.08905) - - [Rapporto tecnico Phi-4-Mini: modelli linguistici multimodali compatti ma potenti tramite Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Ottimizzazione di piccoli modelli linguistici per chiamate di funzione in veicolo](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Fine-Tuning PHI-3 per domande a scelta multipla: metodologia, risultati e sfide](https://arxiv.org/abs/2501.01588) + - [Rapporto tecnico Phi-4-Mini: Modelli linguistici multimodali compatti ma potenti tramite Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Ottimizzazione di modelli linguistici piccoli per il richiamo di funzioni in-vehicle](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Messa a punto di PHI-3 per risposte a domande a scelta multipla: metodo, risultati e sfide](https://arxiv.org/abs/2501.01588) - [Rapporto tecnico Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Rapporto tecnico Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Uso dei modelli Phi +## Utilizzo dei modelli Phi ### Phi su Microsoft Foundry -Puoi imparare come usare Microsoft Phi e come costruire soluzioni end-to-end nei tuoi diversi dispositivi hardware. Per provare Phi personalmente, inizia a sperimentare con i modelli e personalizza Phi per i tuoi scenari usando il [Catalogo modelli Azure AI di Microsoft Foundry](https://aka.ms/phi3-azure-ai). Puoi saperne di più su Come iniziare con [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md). +Puoi imparare come usare Microsoft Phi e come costruire soluzioni E2E sui tuoi diversi dispositivi hardware. Per provare Phi personalmente, inizia a giocare con i modelli e a personalizzare Phi per i tuoi scenari usando il [Catalogo modelli AI Azure Microsoft Foundry](https://aka.ms/phi3-azure-ai). Puoi saperne di più su Come iniziare con [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) **Playground** -Ogni modello ha un playground dedicato per testare il modello su [Azure AI Playground](https://aka.ms/try-phi3). +Ogni modello ha un playground dedicato per testare il modello [Azure AI Playground](https://aka.ms/try-phi3). ### Phi su modelli GitHub -Puoi imparare come usare Microsoft Phi e come costruire soluzioni end-to-end nei tuoi diversi dispositivi hardware. Per provare Phi personalmente, inizia a giocare con il modello e personalizza Phi per i tuoi scenari usando il [Catalogo modelli GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Puoi saperne di più su Come iniziare con [Catalogo modelli GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md). +Puoi imparare come usare Microsoft Phi e come costruire soluzioni E2E sui tuoi diversi dispositivi hardware. Per provare Phi personalmente, inizia a giocare con il modello e a personalizzare Phi per i tuoi scenari usando il [Catalogo modelli GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Puoi saperne di più su Come iniziare con [Catalogo modelli GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) **Playground** Ogni modello ha un [playground dedicato per testare il modello](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi su Hugging Face -Puoi anche trovare il modello su [Hugging Face](https://huggingface.co/microsoft). +Puoi anche trovare il modello su [Hugging Face](https://huggingface.co/microsoft) **Playground** -[Playground Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct). + [Playground Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Altri corsi + ## 🎒 Altri corsi Il nostro team produce altri corsi! Dai un'occhiata a: @@ -263,52 +266,54 @@ Il nostro team produce altri corsi! Dai un'occhiata a: [![Agenti AI per principianti](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Serie AI generativa + +### Serie AI Generativa [![AI generativa per principianti](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![AI generativa (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) [![AI generativa (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) [![AI generativa (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Apprendimento base + +### Apprendimento di base [![ML per principianti](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Data Science per principianti](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI per principianti](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity per principianti](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web Dev per principianti](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![Sicurezza informatica per principianti](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Sviluppo web per principianti](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![IoT per principianti](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![Sviluppo XR per principianti](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Serie Copilot -[![Copilot per programmazione in coppia AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot per programmazione affiancata AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot per C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Avventura Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![Avventure con Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## AI responsabile -Microsoft si impegna ad aiutare i nostri clienti a usare i nostri prodotti AI in modo responsabile, condividendo le nostre esperienze e costruendo partnership basate sulla fiducia tramite strumenti come trasparenza e valutazioni d'impatto. Molte di queste risorse sono disponibili su [https://aka.ms/RAI](https://aka.ms/RAI). -L'approccio di Microsoft all'AI responsabile si basa sui nostri principi AI di equità, affidabilità e sicurezza, privacy e sicurezza, inclusività, trasparenza e responsabilità. +Microsoft si impegna ad aiutare i nostri clienti a usare i nostri prodotti AI in modo responsabile, condividendo le nostre esperienze e costruendo partnership basate sulla fiducia tramite strumenti come le Note di Trasparenza e le Valutazioni di Impatto. Molte di queste risorse sono disponibili su [https://aka.ms/RAI](https://aka.ms/RAI). +L'approccio di Microsoft all'AI responsabile si basa sui nostri principi di equità, affidabilità e sicurezza, privacy e protezione, inclusività, trasparenza e responsabilità. + +Modelli su larga scala di linguaggio naturale, immagini e parlato - come quelli usati in questo esempio - possono potenzialmente comportarsi in modi ingiusti, inaffidabili o offensivi, causando danni. Si prega di consultare la [Nota di trasparenza del servizio Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) per essere informati sui rischi e le limitazioni. + -I modelli di linguaggio naturale, immagini e voce su larga scala - come quelli usati in questo esempio - possono comportarsi in modi ingiusti, inaffidabili o offensivi, causando potenziali danni. Consulta la [nota sulla trasparenza del servizio Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) per essere informato sui rischi e le limitazioni. -L'approccio consigliato per mitigare questi rischi è includere un sistema di sicurezza nella tua architettura in grado di rilevare e prevenire comportamenti dannosi. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) fornisce un livello indipendente di protezione, in grado di rilevare contenuti dannosi generati dagli utenti e dall'IA nelle applicazioni e nei servizi. Azure AI Content Safety include API per testi e immagini che consentono di rilevare materiale dannoso. All'interno di Microsoft Foundry, il servizio Content Safety ti permette di visualizzare, esplorare e provare codice di esempio per rilevare contenuti dannosi attraverso diverse modalità. La seguente [documentazione quickstart](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ti guida nel fare richieste al servizio. +L'approccio raccomandato per mitigare questi rischi è includere un sistema di sicurezza nella tua architettura che possa rilevare e prevenire comportamenti dannosi. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) fornisce un livello di protezione indipendente, in grado di rilevare contenuti dannosi generati dagli utenti e dall'IA nelle applicazioni e nei servizi. Azure AI Content Safety include API di testo e immagine che permettono di rilevare materiale dannoso. All'interno di Microsoft Foundry, il servizio Content Safety consente di visualizzare, esplorare e provare esempi di codice per rilevare contenuti dannosi attraverso diverse modalità. La seguente [documentazione quickstart](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ti guida nella creazione di richieste al servizio. -Un altro aspetto da tenere in considerazione è la performance complessiva dell'applicazione. Con applicazioni multi-modali e multi-modello, consideriamo la performance come il fatto che il sistema si comporti come tu e i tuoi utenti vi aspettate, incluso il non generare output dannosi. È importante valutare la performance della tua applicazione complessiva utilizzando i [valutatori di Performance e Qualità e Rischio e Sicurezza](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Hai anche la possibilità di creare e valutare con [valutatori personalizzati](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Un altro aspetto da considerare è la performance complessiva dell'applicazione. Con applicazioni multi-modali e multi-modello, per performance intendiamo che il sistema si comporti come tu e i tuoi utenti vi aspettate, incluso il non generare output dannosi. È importante valutare la performance della tua applicazione complessiva utilizzando i [valutatori di Performance e Qualità e di Rischio e Sicurezza](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Hai inoltre la possibilità di creare e valutare con [valutatori personalizzati](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Puoi valutare la tua applicazione AI nel tuo ambiente di sviluppo utilizzando l'[Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Fornito un dataset di test o un obiettivo, le generazioni della tua applicazione AI generativa vengono misurate quantitativamente con valutatori integrati o valutatori personalizzati a tua scelta. Per iniziare con l'azure ai evaluation sdk per valutare il tuo sistema, puoi seguire la [guida quickstart](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Una volta eseguito un ciclo di valutazione, puoi [visualizzare i risultati in Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Puoi valutare la tua applicazione IA nel tuo ambiente di sviluppo usando l'[Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Dato un set di dati di test o un obiettivo, le generazioni della tua applicazione AI generativa vengono misurate quantitativamente con valutatori incorporati o valutatori personalizzati a tua scelta. Per iniziare con l'azure ai evaluation sdk per valutare il tuo sistema, puoi seguire la [guida quickstart](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Una volta eseguita una valutazione, puoi [visualizzare i risultati in Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## Marchi registrati +## Marchi commerciali -Questo progetto può contenere marchi o loghi di progetti, prodotti o servizi. L'uso autorizzato dei marchi o loghi Microsoft è soggetto a e deve seguire le [Linee guida sui marchi e sul brand di Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -L'uso di marchi o loghi Microsoft in versioni modificate di questo progetto non deve causare confusione o implicare sponsorizzazione da parte di Microsoft. Qualsiasi uso di marchi o loghi di terzi è soggetto alle politiche di tali terzi. +Questo progetto può contenere marchi o loghi di progetti, prodotti o servizi. L'uso autorizzato dei marchi o loghi Microsoft è soggetto e deve seguire le [Linee guida sui marchi e i brand Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +L'uso di marchi o loghi Microsoft in versioni modificate di questo progetto non deve causare confusione né implicare sponsorizzazione da parte di Microsoft. Qualsiasi uso di marchi o loghi di terze parti è soggetto alle politiche di tali terze parti. -## Ottenere aiuto +## Richiedere assistenza -Se hai difficoltà o domande sulla creazione di app AI, unisciti a: +Se incontri difficoltà o hai domande sulla creazione di app IA, unisciti a: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) @@ -320,5 +325,5 @@ Se hai feedback sul prodotto o errori durante la costruzione visita: **Disclaimer**: -Questo documento è stato tradotto utilizzando il servizio di traduzione AI [Co-op Translator](https://github.com/Azure/co-op-translator). Sebbene ci impegniamo per l'accuratezza, si prega di considerare che le traduzioni automatiche possono contenere errori o imprecisioni. Il documento originale nella sua lingua nativa deve essere considerato la fonte autorevole. Per informazioni critiche, si raccomanda una traduzione professionale umana. Non siamo responsabili per eventuali malintesi o interpretazioni errate derivanti dall'uso di questa traduzione. +Questo documento è stato tradotto utilizzando il servizio di traduzione AI [Co-op Translator](https://github.com/Azure/co-op-translator). Sebbene ci impegniamo per garantire la precisione, si prega di notare che le traduzioni automatizzate possono contenere errori o imprecisioni. Il documento originale nella sua lingua nativa deve essere considerato la fonte autorevole. Per informazioni critiche, si raccomanda una traduzione professionale effettuata da un essere umano. Non siamo responsabili per eventuali malintesi o interpretazioni errate derivanti dall’uso di questa traduzione. \ No newline at end of file diff --git a/translations/it/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/it/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..ed760d30d --- /dev/null +++ b/translations/it/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 RAG Ibrido con Microsoft phi-4-mini & SQLite FTS5 (SLM Zero-Cloud)\n", + "\n", + "> **Autore:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Focus:** Small Language Models (SLM), SQLite FTS5 BM25, Dense Embeddings, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivazione: Il Dilemma del Richiamo delle Parole Chiave negli SLM Locali\n", + "Le architetture RAG standard che si basano esclusivamente su dense vector embeddings spesso non riescono a recuperare token numerici esatti (es. `2,340,000 TL`, codici contratto, numeri di conto). \n", + "Al contrario, la ricerca lessicale sparsa (BM25) perde sinonimi semanticamente equivalenti e domande parafrasate.\n", + "\n", + "Questo cookbook mostra come implementare un motore di ricerca ibrido **ad alta velocità e in-memory** combinando:\n", + "1. **Vettori Densi** (Cosine Similarity)\n", + "2. **Ricerca Lessicale Sparsa** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Generazione di Citazioni Supportate (`[1]`, `[2]`)** con Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Schema Dual SQLite (Vettori Densi + Tabella Virtuale FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Benchmark di Ingestione ed Esecuzione dei Campioni\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Formulazione del Prompt Fondato per Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Disclaimer**:\nQuesto documento è stato tradotto utilizzando il servizio di traduzione AI [Co-op Translator](https://github.com/Azure/co-op-translator). Sebbene ci impegniamo per garantire la precisione, si prega di notare che le traduzioni automatizzate possono contenere errori o imprecisioni. Il documento originale nella sua lingua nativa deve essere considerato la fonte autorevole. Per informazioni critiche, si raccomanda una traduzione professionale effettuata da un essere umano. Non siamo responsabili per eventuali malintesi o interpretazioni errate derivanti dall’uso di questa traduzione.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/ja/.co-op-translator.json b/translations/ja/.co-op-translator.json index 3b01982c9..bed55026a 100644 --- a/translations/ja/.co-op-translator.json +++ b/translations/ja/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "ja" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:13:16+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:24:05+00:00", "source_file": "README.md", "language_code": "ja" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "ja" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:29:40+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "ja" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:05+00:00", diff --git a/translations/ja/README.md b/translations/ja/README.md index 115f756dc..0606f6a8a 100644 --- a/translations/ja/README.md +++ b/translations/ja/README.md @@ -1,6 +1,6 @@ -# Phi クックブック：Microsoft の Phi モデルを使ったハンズオン例 +# Phi クックブック: Microsoftの Phi モデルを使ったハンズオン例 -[![GitHub Codespaces でサンプルを開いて使用する](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![GitHub Codespaces でサンプルを開いて使う](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Dev Containers で開く](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub コントリビューター](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) @@ -8,35 +8,35 @@ [![GitHub プルリクエスト](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![PR歓迎](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ウォッチャー](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub Watchers](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub フォーク](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub スター](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi は Microsoft によって開発されたオープンソースの AI モデルのシリーズです。 +Phi は Microsoft によって開発された一連のオープンソース AI モデルです。 -Phi は現在、非常に優れた多言語・推論・テキスト/チャット生成・コーディング・画像・音声およびその他のシナリオでのベンチマークを誇る、最も強力かつコスト効率の高い小型言語モデル（SLM）です。 +Phi は現在、マルチ言語、推論、テキスト／チャット生成、コーディング、画像、音声などのシナリオで非常に優れたベンチマークを持つ、最も強力でコスト効率の高い小型言語モデル（SLM）です。 -Phi はクラウドやエッジデバイスにデプロイ可能で、限られた計算リソースでも簡単に生成型 AI アプリケーションを構築できます。 +Phi はクラウドまたはエッジデバイスに展開でき、限られた計算リソースでも生成AIアプリケーションを簡単に構築可能です。 -以下のステップに従ってこれらのリソースの使用を開始してください： -1. リポジトリをフォークする: クリック [![GitHub フォーク](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +これらのリソースを使い始めるには次の手順を実行してください： +1. リポジトリをフォークする: クリック [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. リポジトリをクローンする: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Microsoft AI Discord コミュニティに参加し、専門家や他の開発者と交流する**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Microsoft AI Discord コミュニティに参加し、専門家や開発者仲間と交流する**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/ja/cover.eb18d1b9605d754b.webp) -### 🌐 多言語対応 +### 🌐 マルチ言語対応 -#### GitHub Actions によるサポート（自動化＆常に最新） +#### GitHub Action によるサポート（自動化＆常に最新） -[アラビア語](../ar/README.md) | [ベンガル語](../bn/README.md) | [ブルガリア語](../bg/README.md) | [ビルマ語（ミャンマー）](../my/README.md) | [中国語（簡体字）](../zh-CN/README.md) | [中国語（繁体字、香港）](../zh-HK/README.md) | [中国語（繁体字、マカオ）](../zh-MO/README.md) | [中国語（繁体字、台湾）](../zh-TW/README.md) | [クロアチア語](../hr/README.md) | [チェコ語](../cs/README.md) | [デンマーク語](../da/README.md) | [オランダ語](../nl/README.md) | [エストニア語](../et/README.md) | [フィンランド語](../fi/README.md) | [フランス語](../fr/README.md) | [ドイツ語](../de/README.md) | [ギリシャ語](../el/README.md) | [ヘブライ語](../he/README.md) | [ヒンディー語](../hi/README.md) | [ハンガリー語](../hu/README.md) | [インドネシア語](../id/README.md) | [イタリア語](../it/README.md) | [日本語](./README.md) | [カンナダ語](../kn/README.md) | [クメール語](../km/README.md) | [韓国語](../ko/README.md) | [リトアニア語](../lt/README.md) | [マレー語](../ms/README.md) | [マラヤーラム語](../ml/README.md) | [マラーティー語](../mr/README.md) | [ネパール語](../ne/README.md) | [ナイジェリア・ピジン語](../pcm/README.md) | [ノルウェー語](../no/README.md) | [ペルシア語（ファルシ）](../fa/README.md) | [ポーランド語](../pl/README.md) | [ポルトガル語（ブラジル）](../pt-BR/README.md) | [ポルトガル語（ポルトガル）](../pt-PT/README.md) | [パンジャブ語（グルムキー）](../pa/README.md) | [ルーマニア語](../ro/README.md) | [ロシア語](../ru/README.md) | [セルビア語（キリル）](../sr/README.md) | [スロバキア語](../sk/README.md) | [スロベニア語](../sl/README.md) | [スペイン語](../es/README.md) | [スワヒリ語](../sw/README.md) | [スウェーデン語](../sv/README.md) | [タガログ語（フィリピン）](../tl/README.md) | [タミル語](../ta/README.md) | [テルグ語](../te/README.md) | [タイ語](../th/README.md) | [トルコ語](../tr/README.md) | [ウクライナ語](../uk/README.md) | [ウルドゥー語](../ur/README.md) | [ベトナム語](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](./README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **ローカルクローンを希望しますか？** +> **ローカルでクローンしたいですか？** > -> このリポジトリには50以上の言語の翻訳が含まれており、ダウンロードサイズが大幅に増加します。翻訳を含めずにクローンする場合は、スパースチェックアウトを使用してください： +> このリポジトリには50以上の言語翻訳が含まれているため、ダウンロードサイズが大幅に増加します。翻訳なしでクローンするにはスパースチェックアウトを使用してください： > > **Bash / macOS / Linux:** > ```bash @@ -45,33 +45,33 @@ Phi はクラウドやエッジデバイスにデプロイ可能で、限られ > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` > -> **CMD（Windows）:** +> **CMD (Windows):** > ```cmd > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> これにより、コース完了に必要なすべてがより高速にダウンロード可能になります。 +> これにより翻訳なしでより速くダウンロードしてコースを完了するために必要なすべてが得られます。 ## 目次 - はじめに - [Phi ファミリーへようこそ](./md/01.Introduction/01/01.PhiFamily.md) - - [環境のセットアップ](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [環境設定](./md/01.Introduction/01/01.EnvironmentSetup.md) - [主要技術の理解](./md/01.Introduction/01/01.Understandingtech.md) - [Phi モデルの AI セーフティ](./md/01.Introduction/01/01.AISafety.md) - [Phi ハードウェアサポート](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi モデルとプラットフォーム別の利用状況](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Guidance-ai と Phi の使用](./md/01.Introduction/01/01.Guidance.md) - - [GitHub Marketplace Models](https://github.com/marketplace/models) - - [Azure AI Model Catalog](https://ai.azure.com) + - [Phi モデルとプラットフォーム間の可用性](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Guidance-ai と Phi の利用法](./md/01.Introduction/01/01.Guidance.md) + - [GitHub マーケットプレイスのモデル](https://github.com/marketplace/models) + - [Azure AI モデルカタログ](https://ai.azure.com) -- 様々な環境での Phi 推論 +- 様々な環境で Phi 推論 - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [GitHub Models](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry Model Catalog](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [GitHub モデル](./md/01.Introduction/02/02.GitHubModel.md) + - [Microsoft Foundry モデルカタログ](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) @@ -82,60 +82,61 @@ Phi はクラウドやエッジデバイスにデプロイ可能で、限られ - [Android での Phi 推論](./md/01.Introduction/03/Android_Inference.md) - [Jetson での Phi 推論](./md/01.Introduction/03/Jetson_Inference.md) - [AI PC での Phi 推論](./md/01.Introduction/03/AIPC_Inference.md) - - [Apple MLX フレームワークによる Phi 推論](./md/01.Introduction/03/MLX_Inference.md) + - [Apple MLX フレームワークでの Phi 推論](./md/01.Introduction/03/MLX_Inference.md) - [ローカルサーバーでの Phi 推論](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI Toolkit を使用したリモートサーバーでの Phi 推論](./md/01.Introduction/03/Remote_Interence.md) + - [AI Toolkit を使ったリモートサーバでの Phi 推論](./md/01.Introduction/03/Remote_Interence.md) - [Rust での Phi 推論](./md/01.Introduction/03/Rust_Inference.md) - - [ローカルでの Phi ビジョン推論](./md/01.Introduction/03/Vision_Inference.md) - - [Kaito AKS、Azure Containers による Phi 推論（公式サポート）](./md/01.Introduction/03/Kaito_Inference.md) - + - [ローカルでの Phi--Vision 推論](./md/01.Introduction/03/Vision_Inference.md) + - [Kaito AKS、Azure コンテナ（公式サポート）での Phi 推論](./md/01.Introduction/03/Kaito_Inference.md) - [Phi ファミリーの量子化](./md/01.Introduction/04/QuantifyingPhi.md) - [llama.cpp を使った Phi-3.5 / 4 の量子化](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime 用生成 AI 拡張を使った Phi-3.5 / 4 の量子化](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [onnxruntime 用生成AI拡張を使った Phi-3.5 / 4 の量子化](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [Intel OpenVINO を使った Phi-3.5 / 4 の量子化](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - [Apple MLX フレームワークを使った Phi-3.5 / 4 の量子化](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- Phi の評価 - - [Responsible AI](./md/01.Introduction/05/ResponsibleAI.md) - - [評価のための Microsoft Foundry](./md/01.Introduction/05/AIFoundry.md) +- Phi の評価 + - [責任ある AI](./md/01.Introduction/05/ResponsibleAI.md) + - [Microsoft Foundry を使った評価](./md/01.Introduction/05/AIFoundry.md) - [Promptflow を使った評価](./md/01.Introduction/05/Promptflow.md) - + - Azure AI Search を使った RAG - - [Azure AI Search での Phi-4-mini と Phi-4-multimodal(RAG) の使い方](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Phi-4-mini と Phi-4-multimodal（RAG）を Azure AI Search で使う方法](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [SQLite FTS5 と phi-4-mini を使ったゼロクラウドローカルハイブリッド RAG](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Phi アプリケーション開発サンプル - - テキスト＆チャットアプリケーション + - テキスト & チャットアプリケーション - Phi-4 サンプル - [📓] [Phi-4-mini ONNX モデルとのチャット](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 ローカル ONNX モデルとのチャット .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Semantic Kernel を使用した Phi-4 ONNX の .NET コンソールアプリチャット](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - [Phi-4 ローカル ONNX モデル .NET でチャット](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Semantic Kernel を使用した Phi-4 ONNX の .NET コンソールアプリでチャット](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 サンプル - - [ブラウザで Phi3、ONNX Runtime Web と WebGPU を使ったローカルチャットボット](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [マルチモデル - 対話型 Phi-3-mini と OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - ラッパーの構築と Phi-3 を MLFlow で使用する](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [モデル最適化 - Olive を使って Phi-3-mini モデルを ONNX Runtime Web 用に最適化する方法](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Phi3、ONNX Runtime Web、および WebGPU を使用したブラウザ内ローカルチャットボット](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [OpenVino チャット](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [マルチモデル - インタラクティブな Phi-3-mini と OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - ラッパーの構築および Phi-3 の MLFlow での使用](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [モデル最適化 - Olive を使用した ONNX Runtime Web 向け Phi-3-min モデルの最適化方法](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [Phi-3 mini-4k-instruct-onnx を使った WinUI3 アプリ](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 マルチモデル AI 搭載ノートアプリのサンプル](https://github.com/microsoft/ai-powered-notes-winui3-sample) + -[WinUI3 マルチモデル AI 搭載ノートアプリサンプル](https://github.com/microsoft/ai-powered-notes-winui3-sample) - [カスタム Phi-3 モデルのファインチューニングと Prompt flow との統合](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Microsoft Foundry での Prompt flow を使ったカスタム Phi-3 モデルのファインチューニングと統合](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Microsoft Foundry における Prompt flow とカスタム Phi-3 モデルのファインチューニングおよび統合](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - [Microsoft の責任ある AI 原則に焦点を当てた Microsoft Foundry でのファインチューニング済み Phi-3 / Phi-3.5 モデルの評価](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct 言語予測サンプル（中国語／英語）](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [📓] [Phi-3.5-mini-instruct の言語予測サンプル（中国語/英語）](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG チャットボット](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU を使って Phi-3.5-Instruct ONNX の Prompt flow ソリューションを作成する](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Microsoft Phi-3.5 tflite を使って Android アプリを作成する](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Microsoft.ML.OnnxRuntime を使ったローカル ONNX Phi-3 モデルによる Q&A .NET の例](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Semantic Kernel と Phi-3 を使ったコンソールチャット .NET アプリ](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI Inference SDK のコードベースサンプル - - Phi-4 サンプル - - [📓] [Phi-4-multimodal を用いたプロジェクトコード生成](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [Windows GPU を使用して Phi-3.5-Instruct ONNX で Prompt flow ソリューションを作成する方法](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Microsoft Phi-3.5 tflite を使用して Android アプリを作成する方法](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Microsoft.ML.OnnxRuntime を使用したローカル ONNX Phi-3 モデルによる Q&A .NET サンプル](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Semantic Kernel と Phi-3 を用いたコンソールチャット .NET アプリ](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI Inference SDK コードベースのサンプル + - Phi-4 サンプル + - [📓] [Phi-4-multimodal を使用したプロジェクトコード生成](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 サンプル - - [Microsoft Phi-3 ファミリーを使って Visual Studio Code GitHub Copilot Chat を構築する](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub モデルで Phi-3.5 による独自の Visual Studio Code Chat Copilot エージェントを作成する](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Microsoft Phi-3 ファミリーを用いて Visual Studio Code GitHub Copilot Chat を構築する](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub モデルで Phi-3.5 を使った Visual Studio Code チャットコパイロットエージェントの作成](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - 高度な推論サンプル - - Phi-4 サンプル + - Phi-4 サンプル - [📓] [Phi-4-mini-reasoning または Phi-4-reasoning サンプル](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - [📓] [Microsoft Olive を使った Phi-4-mini-reasoning のファインチューニング](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Apple MLX を使った Phi-4-mini-reasoning のファインチューニング](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) @@ -143,112 +144,113 @@ Phi はクラウドやエッジデバイスにデプロイ可能で、限られ - [📓] [Microsoft Foundry モデルを使った Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - デモ - [Hugging Face Spaces でホストされている Phi-4-mini デモ](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Hugginge Face Spaces でホストされている Phi-4-multimodal デモ](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [Hugging Face Spaces でホストされている Phi-4-multimodal デモ](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - ビジョンサンプル - - Phi-4 サンプル - - [📓] [Phi-4-multimodal を用いて画像を読み取りコードを生成する](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-4 サンプル + - [📓] [Phi-4-multimodal を使用して画像を読み取りコードを生成する](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 サンプル - - [📓][Phi-3-vision-画像テキスト変換](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision 画像テキストからテキストへの変換](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP 埋め込み](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 リサイクリング](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - ビジュアル言語アシスタント - Phi3-Vision と OpenVINO を使って](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [デモ: Phi-3 リサイクル](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - Phi3-Vision と OpenVINO を使ったビジュアル言語アシスタント](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - [📓][Phi-3.5 Vision マルチフレームまたはマルチ画像サンプル](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Microsoft.ML.OnnxRuntime .NET を使った Phi-3 Vision ローカル ONNX モデル](../../md/04.HOL/dotnet/src/LabsPhi303) - - [メニュー形式の Phi-3 Vision ローカル ONNX モデル Microsoft.ML.OnnxRuntime .NET を使用](../../md/04.HOL/dotnet/src/LabsPhi304) + - [Microsoft.ML.OnnxRuntime .NET を使用した Phi-3 Vision ローカル ONNX モデル](../../md/04.HOL/dotnet/src/LabsPhi303) + - [メニュー方式 Phi-3 Vision ローカル ONNX モデル Microsoft.ML.OnnxRuntime .NET 使用](../../md/04.HOL/dotnet/src/LabsPhi304) - - Reasoning-Vision サンプル - - Phi-4-Reasoning-Vision-15B - - [📓] [Phi-4-Reasoning-Vision-15B を使って歩行禁止区域違反を検出する](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Phi-4-Reasoning-Vision-15B を使って数学問題を解く](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-Reasoning-Vision-15B を使って UI を検出する](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - 推論・ビジョンサンプル + - Phi-4-Reasoning-Vision-15B + - [📓] [Phi-4-Reasoning-Vision-15B を使った歩行者無視検出](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B を使った数学](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B を使った UI 検出](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - 数学サンプル - - Phi-4-Mini-Flash-Reasoning-Instruct サンプル [Phi-4-Mini-Flash-Reasoning-Instruct による数学デモ](./md/02.Application/09.Math/MathDemo.ipynb) + - 数式サンプル + - Phi-4-Mini-Flash-Reasoning-Instruct サンプル [Phi-4-Mini-Flash-Reasoning-Instruct を使った数学デモ](./md/02.Application/09.Math/MathDemo.ipynb) - オーディオサンプル - - Phi-4 サンプル - - [📓] [Phi-4-multimodal を使った音声書き起こしの抽出](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal 音声サンプル](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - Phi-4 サンプル + - [📓] [Phi-4-multimodal を使ったオーディオ転写抽出](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal オーディオサンプル](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Phi-4-multimodal 音声翻訳サンプル](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [Phi-4-multimodal を使った .NET コンソールアプリケーションで音声ファイルを分析し文字起こしを生成する](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [.NET コンソールアプリで Phi-4-multimodal オーディオを使って音声ファイル分析と転写生成](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE サンプル - Phi-3 / 3.5 サンプル - - [📓] [Phi-3.5 Mixture of Experts モデル (MoEs) ソーシャルメディアサンプル](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE、Azure AI Search、LlamaIndex を使った Retrieval-Augmented Generation (RAG) パイプライン構築](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 Mixture of Experts (MoEs) ソーシャルメディアサンプル](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE、Azure AI Search、および LlamaIndex を使用した Retrieval-Augmented Generation (RAG) パイプライン構築](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - 関数呼び出しサンプル - Phi-4 サンプル 🆕 - [📓] [Phi-4-mini での関数呼び出しの使用](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Phi-4-mini で関数呼び出しを使いマルチエージェントを作成する](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Ollama で関数呼び出しを使用する](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [ONNX で関数呼び出しを使用する](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - [📓] [Phi-4-mini で関数呼び出しを使いマルチエージェントを作成](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Ollama で関数呼び出しを使う](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [ONNX で関数呼び出しを使う](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - マルチモーダルミキシングサンプル - Phi-4 サンプル 🆕 - - [📓] [Phi-4-multimodal をテクノロジージャーナリストとして使用](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET コンソールアプリケーションで Phi-4-multimodal を使用して画像を分析](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [📓] [技術ジャーナリストとして Phi-4-multimodal を使う](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET コンソールアプリで Phi-4-multimodal を使って画像分析](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Phi ファインチューニングサンプル +- Phi ファインチューニング - [ファインチューニングシナリオ](./md/03.FineTuning/FineTuning_Scenarios.md) - [ファインチューニングと RAG の比較](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [ファインチューニング：Phi-3 を業界専門家にする](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [VS Code 用 AI ツールキットでの Phi-3 ファインチューニング](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Azure Machine Learning サービスによる Phi-3 ファインチューニング](./md/03.FineTuning/Introduce_AzureML.md) - - [Lora を使った Phi-3 ファインチューニング](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora を使った Phi-3 ファインチューニング](./md/03.FineTuning/FineTuning_Qlora.md) - - [Microsoft Foundry での Phi-3 ファインチューニング](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Azure ML CLI/SDK での Phi-3 ファインチューニング](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Microsoft Olive を使ったファインチューニング](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive ハンズオンラボでのファインチューニング](./md/03.FineTuning/olive-lab/readme.md) + - [Phi-3 を業界の専門家にするファインチューニング](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [VS Code の AI Toolkit で Phi-3 ファインチューニング](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure Machine Learning Service で Phi-3 ファインチューニング](./md/03.FineTuning/Introduce_AzureML.md) + - [Lora で Phi-3 ファインチューニング](./md/03.FineTuning/FineTuning_Lora.md) + - [QLora で Phi-3 ファインチューニング](./md/03.FineTuning/FineTuning_Qlora.md) + - [Microsoft Foundry で Phi-3 ファインチューニング](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Azure ML CLI/SDK で Phi-3 ファインチューニング](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Microsoft Olive でファインチューニング](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Microsoft Olive ハンズオンラボでファインチューニング](./md/03.FineTuning/olive-lab/readme.md) - [Weights and Bias を使った Phi-3-vision のファインチューニング](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Apple MLX フレームワークを使った Phi-3 ファインチューニング](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-vision ファインチューニング（公式サポート）](./md/03.FineTuning/FineTuning_Vision.md) - - [Kaito AKSによるPhi-3のファインチューニング、Azureコンテナ（公式サポート）](./md/03.FineTuning/FineTuning_Kaito.md) - - [Phi-3および3.5 Visionのファインチューニング](https://github.com/2U1/Phi3-Vision-Finetune) + + - [Apple MLX フレームワークによる Phi-3 のファインチューニング](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision のファインチューニング（公式サポート）](./md/03.FineTuning/FineTuning_Vision.md) + - [Kaito AKS、Azure コンテナによる Phi-3 のファインチューニング（公式サポート）](./md/03.FineTuning/FineTuning_Kaito.md) + - [Phi-3 および 3.5 Vision のファインチューニング](https://github.com/2U1/Phi3-Vision-Finetune) - ハンズオンラボ - - [最先端モデルの探求：LLM、SLM、ローカル開発など](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLPの可能性を引き出す：Microsoft Oliveによるファインチューニング](https://github.com/azure/Ignite_FineTuning_workshop) + - [最先端モデルの探索：LLM、SLM、ローカル開発など](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLPの可能性を解き放つ：Microsoft Oliveでのファインチューニング](https://github.com/azure/Ignite_FineTuning_workshop) - 学術研究論文と出版物 - - [Textbooks Are All You Need II：phi-1.5技術報告](https://arxiv.org/abs/2309.05463) - - [Phi-3技術報告：高性能な言語モデルをローカルの携帯電話で](https://arxiv.org/abs/2404.14219) - - [Phi-4技術報告](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini技術報告：Mixture-of-LoRAsによるコンパクトでパワフルなマルチモーダル言語モデル](https://arxiv.org/abs/2503.01743) - - [車載用機能呼び出しに最適化された小型言語モデル](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) PHI-3の多肢選択問題解答向けファインチューニング：方法論、結果、課題](https://arxiv.org/abs/2501.01588) - - [Phi-4推論技術報告](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini推論技術報告](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Textbooks Are All You Need II: phi-1.5 技術レポート](https://arxiv.org/abs/2309.05463) + - [Phi-3 技術レポート: 高性能な言語モデルをローカルのスマホ上で](https://arxiv.org/abs/2404.14219) + - [Phi-4 技術レポート](https://arxiv.org/abs/2412.08905) + - [Phi-4-Mini 技術レポート: Mixture-of-LoRAs によるコンパクトで強力なマルチモーダル言語モデル](https://arxiv.org/abs/2503.01743) + - [車載機能呼び出しに最適化された小型言語モデル](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) 複数選択式質問応答のための PHI-3 ファインチューニング：方法論、結果、課題](https://arxiv.org/abs/2501.01588) + - [Phi-4-reasoning 技術レポート](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-reasoning 技術レポート](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Phiモデルの使用 +## Phi モデルの利用方法 -### Microsoft Foundry上のPhi +### Microsoft Foundry 上の Phi -Microsoft Phiの使い方や、さまざまなハードウェアデバイスでのE2Eソリューションの構築方法を学べます。Phiを体験するには、モデルを操作し、シナリオに合わせてPhiをカスタマイズすることから始めてください。詳しくは[Microsoft Foundry Azure AIモデルカタログ](https://aka.ms/phi3-azure-ai) と [Microsoft Foundryの開始方法](/md/02.QuickStart/AzureAIFoundry_QuickStart.md)をご覧ください。 +Microsoft Phi の使い方やさまざまなハードウェアデバイス上での E2E ソリューション構築方法を学べます。Phi を実際に体験するには、モデルを操作し、ご自身のシナリオに合わせて Phi をカスタマイズすることから始めましょう。詳しくは [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) で学べます。[Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) の始め方もご覧ください。 -プレイグラウンド -各モデルには専用のプレイグラウンドがあり、モデルのテストが可能です。[Azure AI Playground](https://aka.ms/try-phi3) +プレイグラウンド +各モデルには専用のプレイグラウンドが用意されており、モデルを試すことができます。[Azure AI Playground](https://aka.ms/try-phi3)。 -### GitHubモデル上のPhi +### GitHub モデル上の Phi -Microsoft Phiの使い方や、さまざまなハードウェアデバイスでのE2Eソリューション構築方法を学べます。Phiを体験するには、モデルを操作しシナリオに合わせてカスタマイズしてください。詳しくは[GitHubモデルカタログ](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo)と[GitHubモデルカタログの開始方法](/md/02.QuickStart/GitHubModel_QuickStart.md)をご覧ください。 +Microsoft Phi の使い方やさまざまなハードウェアデバイス上での E2E ソリューション構築方法を学べます。Phi を実際に体験するには、モデルを操作し、ご自身のシナリオに合わせて Phi をカスタマイズすることから始めましょう。[GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) から学ぶことができ、[GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) の始め方もご覧ください。 -プレイグラウンド -各モデルには専用の[プレイグラウンドがあります](/md/02.QuickStart/GitHubModel_QuickStart.md)。 +プレイグラウンド +各モデルには専用の[モデルを試すためのプレイグラウンド](/md/02.QuickStart/GitHubModel_QuickStart.md)があります。 -### Hugging Face上のPhi +### Hugging Face 上の Phi -モデルは[Hugging Face](https://huggingface.co/microsoft)でも提供されています。 +モデルは [Hugging Face](https://huggingface.co/microsoft) でも見つけることができます。 -プレイグラウンド -[Hugging Chatプレイグラウンド](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +プレイグラウンド + [Hugging Chat プレイグラウンド](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 その他のコース + ## 🎒 その他のコース -私たちのチームは他にも様々なコースを作成しています！ぜひチェックしてください： +私たちのチームは他にもコースを制作しています！ぜひご覧ください： ### LangChain @@ -277,48 +279,51 @@ Microsoft Phiの使い方や、さまざまなハードウェアデバイスで [![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![サイバーセキュリティ初心者向け](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web 開発初心者向け](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT 初心者向け](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR 開発初心者向け](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Copilotシリーズ -[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +### Copilot シリーズ +[![AI ペアプログラミング向け Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![C#/.NET 向け Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot アドベンチャー](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## 責任あるAI +## 責任ある AI + +Microsoft は、お客様が私たちの AI 製品を責任を持って使用できるよう支援し、学びを共有し、Transparency Notes や Impact Assessments のようなツールを通じて信頼に基づくパートナーシップを構築することにコミットしています。これらのリソースの多くは [https://aka.ms/RAI](https://aka.ms/RAI) で入手可能です。 +Microsoft の責任ある AI アプローチは、公平性、信頼性と安全性、プライバシーとセキュリティ、包括性、透明性、および説明責任という AI 原則に基づいています。 + +このサンプルで使われているような大規模な自然言語処理、画像処理、音声モデルは、不公平、不信頼、攻撃的な動作をする可能性があり、それによって害を生じさせる危険性があります。リスクと制限については [Azure OpenAI サービスの Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) をご参照ください。 -Microsoftは、お客様が当社のAI製品を責任を持って使用できるよう支援し、学びを共有し、透明性ノートや影響評価などのツールを通じて信頼に基づくパートナーシップを築くことにコミットしています。これらのリソースの多くは[https://aka.ms/RAI](https://aka.ms/RAI)でご覧いただけます。 -Microsoftの責任あるAIへのアプローチは、公平性、信頼性と安全性、プライバシーとセキュリティ、包括性、透明性、説明責任というAIの原則に基づいています。 -このサンプルで使用されている大規模な自然言語、画像、音声モデルは、不公平、不信頼、または不快な挙動をする可能性があり、それによって被害をもたらすことがあります。リスクと制限については、[Azure OpenAIサービスの透明性ノート](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text)をご参照ください。 -これらのリスクを軽減する推奨される方法は、有害な行動を検出し防止できる安全システムをアーキテクチャに組み込むことです。[Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) は独立した保護レイヤーを提供し、アプリケーションやサービス内の有害なユーザー生成コンテンツおよびAI生成コンテンツを検出できます。Azure AI Content Safetyには有害な素材を検出できるテキストおよび画像のAPIが含まれています。Microsoft Foundry内では、Content Safetyサービスを使って異なるモダリティにわたる有害コンテンツの検出用のサンプルコードを閲覧、探索、試すことができます。以下の[クイックスタートドキュメント](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest)では、サービスへのリクエストの方法を案内しています。 +これらのリスクを軽減するための推奨される方法は、有害な行動を検知し防止できる安全システムをアーキテクチャに組み込むことです。[Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview)は独立した保護層を提供し、アプリケーションやサービス内での有害なユーザー生成コンテンツおよびAI生成コンテンツを検出できます。Azure AI Content Safetyには、有害な素材を検出するためのテキストおよび画像APIが含まれています。Microsoft Foundry内では、Content Safetyサービスを利用してさまざまなモダリティにわたる有害コンテンツの検出に関するサンプルコードを閲覧、探索、試用できます。以下の[クイックスタートドキュメント](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest)はサービスへのリクエスト方法を案内しています。 -考慮すべきもう一つの側面は全体的なアプリケーションのパフォーマンスです。マルチモーダルかつマルチモデルのアプリケーションでは、パフォーマンスとは、ユーザーやあなたの期待通りにシステムが動作し、有害な出力を生成しないことを意味します。[Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) を使用して全体のアプリケーション性能を評価することが重要です。また、[カスタム評価ツール](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators)を作成して評価することもできます。 +もう一つ考慮すべき点は全体的なアプリケーション性能です。マルチモーダルかつマルチモデルのアプリケーションでは、性能とはシステムがあなたとユーザーの期待通りに動作し、有害な出力を生成しないことを意味します。[Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)を使用して全体のアプリケーション性能を評価することが重要です。また、[カスタム評価者](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators)を作成して評価することも可能です。 -[Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) を使うと、開発環境でAIアプリを評価できます。テストデータセットまたはターゲットを与えると、生成型AIアプリの生成物が組み込み評価器や任意のカスタム評価器で定量的に測定されます。azure ai evaluation sdkを使ってシステム評価を始めるには、[クイックスタートガイド](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)に従ってください。評価を実行すると、[Microsoft Foundryで結果を可視化](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) できます。 +[Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html)を使用すると、開発環境でAIアプリケーションを評価できます。テストデータセットかターゲットが与えられると、生成AIアプリケーションの生成物は、組み込み評価者または選択したカスタム評価者で定量的に測定されます。システムの評価を始めるためにazure ai evaluation sdkの始め方は、[クイックスタートガイド](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)に従ってください。評価実行を行ったら、[Microsoft Foundryで結果を可視化](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)できます。 -## トレードマーク +## 商標 -このプロジェクトにはプロジェクト名、製品名、またはサービスのトレードマークやロゴが含まれている場合があります。Microsoftのトレードマークやロゴの許可された使用は、[Microsoftのトレードマークとブランドのガイドライン](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) に従う必要があります。 このプロジェクトの改変版でのMicrosoftのトレードマークまたはロゴの使用は混乱を招いたりMicrosoftの後援を示唆したりしてはなりません。第三者のトレードマークやロゴの使用は、それら第三者の方針に従います。 +このプロジェクトには、プロジェクト、製品、またはサービスの商標やロゴが含まれている場合があります。Microsoftの商標またはロゴの使用は、[Microsoftの商標およびブランドガイドライン](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)に従い、許可された範囲で行わなければなりません。 +このプロジェクトの修正バージョンでMicrosoftの商標やロゴを使用する際は、混同やMicrosoftのスポンサーシップの示唆を生じさせてはなりません。第三者の商標やロゴの使用については、それら第三者の方針に従います。 -## ヘルプを得る +## ヘルプの取得 AIアプリの構築で行き詰まったり質問がある場合は、以下に参加してください： [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -製品のフィードバックや構築時のエラーの報告は以下へ： +製品フィードバックやエラー報告は以下をご利用ください： [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**免責事項**: -本書類は AI 翻訳サービス [Co-op Translator](https://github.com/Azure/co-op-translator) を使用して翻訳されています。正確性を期していますが、自動翻訳には誤りや不正確な部分が含まれる可能性があることをご理解ください。原文の母国語版が正式な情報源とみなされます。重要な情報については、専門の人間による翻訳を推奨します。本翻訳の使用による誤解や誤訳について当方は一切責任を負いかねます。 +**免責事項**： +本書類は AI 翻訳サービス [Co-op Translator](https://github.com/Azure/co-op-translator) を使用して翻訳されています。正確性を期していますが、自動翻訳には誤りや不正確な部分が含まれる可能性があることをご承知おきください。原文の原語版が正式な情報源とみなされるべきです。重要な情報については、専門の人間による翻訳を推奨します。本翻訳の利用により生じたいかなる誤解や解釈違いについても、当方は責任を負いかねます。 \ No newline at end of file diff --git a/translations/ja/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/ja/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..d94bb1ae4 --- /dev/null +++ b/translations/ja/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Microsoft phi-4-mini と SQLite FTS5 を用いたハイブリッド RAG（ゼロクラウド SLM）\n", + "\n", + "> **著者:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **注目分野:** 小型言語モデル（SLM）、SQLite FTS5 BM25、密ベクトル埋め込み、相互ランク融合（RRF）\n", + "\n", + "---\n", + "\n", + "## 📌 1. 動機：ローカルSLMにおけるキーワードリコールのジレンマ\n", + "密ベクトル埋め込みのみに依存する標準的なRAGアーキテクチャは、正確な数値トークン（例: `2,340,000 TL`、契約コード、口座番号）を取得できないことがよくあります。 \n", + "一方、スパースな語彙検索（BM25）は、意味的な同義語や言い換えた質問を見逃します。\n", + "\n", + "本クックブックでは、以下を組み合わせた高速インメモリハイブリッド検索エンジンの実装方法を示します：\n", + "1. 密ベクトル（コサイン類似度）\n", + "2. スパース語彙検索（SQLite FTS5 BM25）\n", + "3. **相互ランク融合（RRF, $k=60$）**\n", + "4. Microsoft `phi-4-mini` を用いた**根拠付き引用生成 (`[1]`, `[2]`)**\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. 二重SQLiteスキーマ（密ベクトル + 仮想FTS5 BM25テーブル）\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. サンプル取り込みと実行ベンチマーク\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Microsoft phi-4-mini の基礎付けられたプロンプト作成\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**免責事項**：\n本書類は AI 翻訳サービス [Co-op Translator](https://github.com/Azure/co-op-translator) を使用して翻訳されています。正確性を期していますが、自動翻訳には誤りや不正確な部分が含まれる可能性があることをご承知おきください。原文の原語版が正式な情報源とみなされるべきです。重要な情報については、専門の人間による翻訳を推奨します。本翻訳の利用により生じたいかなる誤解や解釈違いについても、当方は責任を負いかねます。\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/km/.co-op-translator.json b/translations/km/.co-op-translator.json index 7c1456e36..221dfd0aa 100644 --- a/translations/km/.co-op-translator.json +++ b/translations/km/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "km" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T08:16:05+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T16:34:06+00:00", "source_file": "README.md", "language_code": "km" }, @@ -167,6 +167,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "km" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T16:34:37+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "km" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2026-04-20T08:22:44+00:00", diff --git a/translations/km/README.md b/translations/km/README.md index ab5251fef..5de32a52d 100644 --- a/translations/km/README.md +++ b/translations/km/README.md @@ -1,254 +1,256 @@ -# ស៊ុមភី (Phi) Cookbook: ឧទាហរណ៍បញ្ចូលដៃជាមួយម៉ូដែល Phi របស់ Microsoft +# សៀវភៅចម្អិន Phi: ឧទាហរណ៍អនុវត្តដោយប្រើម៉ូដែល Phi របស់ Microsoft -[![បើក និង ប្រើការគំរូនៅ GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![បើកនៅ Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![បើកនិងប្រើឧទាហរណ៍នៅក្នុង GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![បើកនៅក្នុង Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![អ្នករួមចំណែក GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![បញ្ហា GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![សំណើរចងក្រង GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![ស្វាគមន៍ PRs](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![សំណើរសុំបិទពាក្យ GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![រក្សាទុក PRs](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![អ្នកតាម GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![កាំបិត GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![តារា GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![អ្នកមើល GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![កាំង GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![ផ្កាយ GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi គឺជាស៊ុមម៉ូដែល AI ផ្លូវចំហ រដ្ធផលដោយ Microsoft។ +Phi គឺជាស៊េរីម៉ូដែល AI ប្រភពបើកដែលបានអភិវឌ្ឍដោយ Microsoft។ -Phi ជាមួយនឹងភាពមានសមត្ថភាពខ្ពស់បំផុត និងអត្ថប្រយោជន៍ថោកបំផុតសម្រាប់ម៉ូដែលភាសាថ្មីតូច (SLM) ជាមួយគំរូវាស់វែងល្អក្នុងភាសាច្រើន, ការរួមបញ្ចូលផ្នែកគំនិត, ការបង្កើតអត្ថបទ/ជជែក, កូដ, រូបភាព, សំឡេង និងសេណារីយ៉ូនផ្សេងទៀត។ +Phi គឺជាម៉ូដែលភាសាតូច (SLM) ដែលមានសមត្ថភាពខ្លាំងបំផុតនិងមានប្រសិទ្ធភាពថ្លៃកាត់បង្គោលបំផុតបច្ចុប្បន្ននេះ ដែលមានសន្ទស្សន៍ល្អនៅក្នុងភាសាច្រើន ការអភិវឌ្ឍន៍ហេតុផល ការបង្កើតអត្ថបទ/ជជែក កូដរចនា រូបភាព សំលេង និងស្ថានការណ៍ផ្សេងទៀត។ -អ្នកអាចដំណើរការចេញ Phi នៅលើពពក ឬ ឧបករណ៍មុខដូច Edge ហើយអ្នកអាចសាងសង់កម្មវិធី AI បង្កើតដោយងាយស្រួលជាមួយថាមពលគណនា​ពិតប្រាកដកំណត់។ +អ្នកអាច triển phi ទៅកូនពពក ឬឧបករណ៍ចុង និង អ្នកអាចសាងសង់កម្មវិធី AI ផលិតផលបានយ៉ាងងាយស្រួល ជាមួយថាមពលកុំព្យូទ័រតិចតួច។ -អនុវត្តន៍ជំហានខាងក្រោមដើម្បីចាប់ផ្តើមប្រើប្រាស់ធនធានទាំងនេះ៖ -1. **លួចបម្រើផ្ទះបាយ (Fork the Repository)**: ចុច [![កាំបិត GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **ចម្លងបម្រើផ្ទះបាយ (Clone the Repository)**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**ចូលរួមជាមួយសហគមន៍ Microsoft AI Discord និងជួបអ្នកជំនាញ និងអ្នកអភិវឌ្ឍន៍ផ្សេងទៀត**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +អ្នកអាចដើរតាមជំហានទាំងនេះដើម្បីចាប់ផ្តើមប្រើធនធានទាំងនេះ៖ +1. **ស្វែងរក Repository**៖ ចុច [![កាំង GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **ចម្លង Repository**៖ `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**ចូលរួមសហគមន៍ Microsoft AI Discord ហើយជួបជាមួយអ្នកជំនាញ និងអ្នកអភិវឌ្ឍរផ្សេងទៀត**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/km/cover.eb18d1b9605d754b.webp) +![ផ្ទុក](../../translated_images/km/cover.eb18d1b9605d754b.webp) -### 🌐 គាំទ្រភាសាច្រើន +### 🌐 គាំទ្រ​ភាសាច្រើន -#### គាំទ្រដោយ GitHub Action (ស្វ័យក្រិត និង ជាប្រចាំ) +#### គាំទ្រដោយ GitHub Action (ជាអូតូម៉ាតិច និងជានិច្ច) -[អារ៉ាប់](../ar/README.md) | [បង់គ្លា](../bn/README.md) | [ប៊ុលហ្គាប៊ី](../bg/README.md) | [ភូមា (មីយ៉ាន់ម៉ា)](../my/README.md) | [ចិន (រូបរាងសាមញ្ញ)](../zh-CN/README.md) | [ចិន (បែបប្រពៃណី, ហុងកុង)](../zh-HK/README.md) | [ចិន (បែបប្រពៃណី, ម៉ាកាវ)](../zh-MO/README.md) | [ចិន (បែបប្រពៃណី, តៃវ៉ាន់)](../zh-TW/README.md) | [ក្រូអាទីយ៉ាន់](../hr/README.md) | [ចែក](../cs/README.md) | [ដាណា](../da/README.md) | [នឌូស](../nl/README.md) | [អេសតូនី](../et/README.md) | [ហ្វាំងណ៊ីស៍](../fi/README.md) | [បារាំង](../fr/README.md) | [អេស្ប៉ាញ](../de/README.md) | [ហ្គ្រីក](../el/README.md) | [ហេប្រ៊ូវ](../he/README.md) | [ហីនឌី](../hi/README.md) | [ហុងគ្រី](../hu/README.md) | [ឥណ្ឌូណេស៊ី](../id/README.md) | [អ៊ីតាលី](../it/README.md) | [ជប៉ុន](../ja/README.md) | [កនាដា](../kn/README.md) | [ខ្មែរ](./README.md) | [កូរ៉េ](../ko/README.md) | [លីទួយ៉ា](../lt/README.md) | [ម៉ាឡាវី](../ms/README.md) | [ម៉ាឡាលាំ](../ml/README.md) | [ម៉ារ៉ាធី](../mr/README.md) | [នេប៉ាឡេ](../ne/README.md) | [ភីជិននីហ្សេរីយ៉ា](../pcm/README.md) | [ន័រវេស៊ី](../no/README.md) | [ភាសាពួស៊ង់ (ហ្វាស៊ី)](../fa/README.md) | [ប៉ូល្ល្](../pl/README.md) | [ព័រទុយហ្គេស (ប្រេស៊ីល)](../pt-BR/README.md) | [ព័រទុយហ្គេស (ព័រទុយហ្គាល់)](../pt-PT/README.md) | [បង់ជាប្រ៊ូណិច (Gurmukhi)](../pa/README.md) | [រូម៉ានី](../ro/README.md) | [រុស្ស៊ី](../ru/README.md) | [ស៊ែប៊ី (Cyrillic)](../sr/README.md) | [ស្លូវ៉ាគ](../sk/README.md) | [ស្លូវ៉េនី](../sl/README.md) | [អេស្ប៉ាញ](../es/README.md) | [ស្វាហ៊ីលី](../sw/README.md) | [ស៊ុយអែដ](../sv/README.md) | [តាឡាហ្គូល (ហ្វីលីពីន)](../tl/README.md) | [ធាមីល](../ta/README.md) | [តេលុគ្រ៊ូ](../te/README.md) | [ថៃ](../th/README.md) | [ទួរកី](../tr/README.md) | [អ៊ុយក្រែន](../uk/README.md) | [អ៊ុយរឌូ](../ur/README.md) | [វៀតណាម](../vi/README.md) +[អារ៉ាប់](../ar/README.md) | [បង់្លាដេស](../bn/README.md) | [ប៊ុលហ្គារី](../bg/README.md) | [ភាសា​ចិន (សាមញ្ញ)](../zh-CN/README.md) | [ភាសា​ចិន (ត្រង់ហុងកុង)](../zh-HK/README.md) | [ភាសា​ចិន (ត្រង់ម៉ាកាវ)](../zh-MO/README.md) | [ភាសា​ចិន (ត្រង់តៃវ៉ាន់)](../zh-TW/README.md) | [ក្រូអាស៊ី](../hr/README.md) | [ភាសាខេក](../cs/README.md) | [ដាណូម៉ាក](../da/README.md) | [ហុល្លង់](../nl/README.md) | [អេស្ទូនី](../et/README.md) | [ហ្វាំងឡង់](../fi/README.md) | [បារាំង](../fr/README.md) | [ជ័ម៉ុន](../de/README.md) | [ក្រិក](../el/README.md) | [អ៊ីស្រាអែល](../he/README.md) | [ហ៊ីណ្ឌូ](../hi/README.md) | [ហុងហ្គារី](../hu/README.md) | [ឥណ្ឌូនេស៊ី](../id/README.md) | [អ៊ីតាលី](../it/README.md) | [ជប៉ុន](../ja/README.md) | [កណ្ដា](../kn/README.md) | [ខ្មែរ](./README.md) | [កូរ៉េ](../ko/README.md) | [លីទុយអានី](../lt/README.md) | [ម៉ាឡេស៊ី](../ms/README.md) | [ម៉ាឡាលាំ](../ml/README.md) | [ម៉ារាតី](../mr/README.md) | [អ្នក​និពន្ធ](../ne/README.md) | [ភាសា​ណាយេជីរី](../pcm/README.md) | [ន័រវೇ](../no/README.md) | [ព័រចាម (ហ្វាស៊ី)](../fa/README.md) | [ប៉ូឡូញ](../pl/README.md) | [ព័រទុយហ្គាល់ (ប្រេស៊ីល)](../pt-BR/README.md) | [ព័រទុយហ្គាល់ (ព័រទុយហ្គាល់)](../pt-PT/README.md) | [ភាសា​ភុនជាប៊ី (គឺមុគហ៊ី)](../pa/README.md) | [រូម៉ានី](../ro/README.md) | [រុស្ស៊ី](../ru/README.md) | [សេរប៊ី (ស៊ីរីលិក)](../sr/README.md) | [ស្លូវ៉ាក់](../sk/README.md) | [ស្លូវេនី](../sl/README.md) | [ស្ប៉ាញ](../es/README.md) | [ស្វាហ៊ីលី](../sw/README.md) | [ស៊ុយអែ](../sv/README.md) | [តាហ្គាឡុក (ភីលីពីន)](../tl/README.md) | [តាមីល](../ta/README.md) | [តេឡូម្គូ](../te/README.md) | [ថៃ](../th/README.md) | [តួរគី](../tr/README.md) | [អ៊ុយក្រែន](../uk/README.md) | [អ៊ួរឌូ](../ur/README.md) | [វៀតណាម](../vi/README.md) -> **ចូលចិត្តចម្លងក្នុងថ្នាក់ក្រៅបន្ទប់?** -> -> បម្រើផ្ទះនេះរួមបញ្ចូលការបកប្រែជាភាសាម៉ាត់ 50+ ដែលបង្កើនទំហំដោនឡូដយ៉ាងខ្លាំង។ ដើម្បីចម្លងដោយគ្មានការប្រែក្លាយភាសា ប្រើ sparse checkout: -> -> **Bash / macOS / Linux:** +> **ចង់ចម្លងទៅក្នុងកុំព្យូទ័រខ្លួនឯង?** +> +> កម្មវិធីនេះរួមបញ្ចូលការប្រែសម្រួលភាសា 50+ ដែលធ្វើឲ្យទំហំទាញយកធំឡើងយ៉ាងខ្លាំង។ ដើម្បីចម្លងដោយគ្មានបកប្រែ សូមប្រើ sparse checkout: +> +> **Bash / macOS / Linux:** > ```bash > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` -> -> **CMD (Windows):** +> +> **CMD (Windows):** > ```cmd > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` -> -> ធានាថាអ្នកមានគ្រឿងបន្លាស់ទាំងអស់ដែលត្រូវការដើម្បីបញ្ចប់វគ្គសិក្សា ជាមួយនឹងដោនឡូដលឿនជាងមុន។ +> +> វានាំអ្វីគ្រប់យ៉ាងដែលអ្នកត្រូវការដើម្បីបញ្ចប់វគ្គនេះដោយទាញយកឆាប់រហ័សជាងមុន។ -## ខ្សែចំណងចិត្ត +## តារាងមាតិកា -- មាតិកា - - [ស្វាគមន៍ទៅកាន់គ្រួសារភី](./md/01.Introduction/01/01.PhiFamily.md) +- ណែនាំ + - [ស្វាគមន៍មកកាន់គ្រួសារភី](./md/01.Introduction/01/01.PhiFamily.md) - [ការតំឡើងបរិយាកាសរបស់អ្នក](./md/01.Introduction/01/01.EnvironmentSetup.md) - [ការយល់ដឹងអំពីបច្ចេកវិទ្យាសំខាន់ៗ](./md/01.Introduction/01/01.Understandingtech.md) - [សុវត្ថិភាព AI សម្រាប់ម៉ូដែល Phi](./md/01.Introduction/01/01.AISafety.md) - - [គាំទ្រ់សម្រាប់រឹងភី](./md/01.Introduction/01/01.Hardwaresupport.md) - - [ម៉ូដែល Phi និងការចូលប្រើនៅលើវេទិកាផ្សេងៗ](./md/01.Introduction/01/01.Edgeandcloud.md) + - [គាំទ្រ हार्डវែរ Phi](./md/01.Introduction/01/01.Hardwaresupport.md) + - [ម៉ូដែល Phi និងមានភាពអាចប្រើបានសាមញ្ញនៅលើវេទិកាទាំងអស់](./md/01.Introduction/01/01.Edgeandcloud.md) - [ការប្រើប្រាស់ Guidance-ai និង Phi](./md/01.Introduction/01/01.Guidance.md) - - [ម៉ូដែលទីផ្សារ GitHub](https://github.com/marketplace/models) - - [បញ្ជីម៉ូដែល Azure AI](https://ai.azure.com) - -- ការជំរុញ Phi នៅក្នុងបរិយាកាសផ្សេងគ្នា - - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [ម៉ូដែល GitHub](./md/01.Introduction/02/02.GitHubModel.md) - - [បញ្ជីម៉ូដែល Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) - - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) + - [ម៉ូដែល GitHub Marketplace](https://github.com/marketplace/models) + - [កាតាឡុកម៉ូដែល Azure AI](https://ai.azure.com) + +- ការបញ្ចេញ Phi នៅបរិយាកាសផ្សេងៗ + - [Hugging face](./md/01.Introduction/02/01.HF.md) + - [ម៉ូដែល GitHub](./md/01.Introduction/02/02.GitHubModel.md) + - [កាតាឡុកម៉ូដែល Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Ollama](./md/01.Introduction/02/04.Ollama.md) + - [ឧបករណ៍ AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- ការជំរុញ Phi Family - - [ការជំរុញ Phi នៅ iOS](./md/01.Introduction/03/iOS_Inference.md) - - [ការជំរុញ Phi នៅ Android](./md/01.Introduction/03/Android_Inference.md) - - [ការជំរុញ Phi នៅ Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [ការជំរុញ Phi នៅ AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [ការជំរុញ Phi ជាមួយ Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [ការជំរុញ Phi នៅ Server តាមដាន](./md/01.Introduction/03/Local_Server_Inference.md) - - [ការជំរុញ Phi នៅ Server ចម្ងាយដោយប្រើ AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [ការជំរុញ Phi ជាមួយ Rust](./md/01.Introduction/03/Rust_Inference.md) - - [ការជំរុញ Phi--Vision នៅមូលដ្ឋាន](./md/01.Introduction/03/Vision_Inference.md) - - [ការជំរុញ Phi ជាមួយ Kaito AKS, Azure Containers (គាំទ្រផ្លូវការជាផ្លូវការសម្បូរ)](./md/01.Introduction/03/Kaito_Inference.md) -- [ការវាស់រូបមន្ត Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) - - [បម្លែង Phi-3.5 / 4 ដោយប្រើ llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [បម្លែង Phi-3.5 / 4 ដោយប្រើជំនួយបន្ថែម AI បង្កើតសម្រាប់ onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [បម្លែង Phi-3.5 / 4 ដោយប្រើ Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [បម្លែង Phi-3.5 / 4 ដោយប្រើ Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- ការវាយតម្លៃ Phi - - [ការឆ្លើយតប AI](./md/01.Introduction/05/ResponsibleAI.md) - - [Microsoft Foundry សម្រាប់ការវាយតម្លៃ](./md/01.Introduction/05/AIFoundry.md) - - [ការប្រើប្រាស់ Promptflow សម្រាប់ការវាយតម្លៃ](./md/01.Introduction/05/Promptflow.md) +- ការបញ្ចេញ Phi Family + - [ការបញ្ចេញ Phi នៅ iOS](./md/01.Introduction/03/iOS_Inference.md) + - [ការបញ្ចេញ Phi នៅ Android](./md/01.Introduction/03/Android_Inference.md) + - [ការបញ្ចេញ Phi នៅ Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [ការបញ្ចេញ Phi នៅ AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [ការបញ្ចេញ Phi ជាមួយ Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [ការបញ្ចេញ Phi នៅ Local Server](./md/01.Introduction/03/Local_Server_Inference.md) + - [ការបញ្ចេញ Phi នៅ Remote Server ដោយប្រើ AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [ការបញ្ចេញ Phi ជាមួយ Rust](./md/01.Introduction/03/Rust_Inference.md) + - [ការបញ្ចេញ Phi--Visionនៅក្នុង Local](./md/01.Introduction/03/Vision_Inference.md) + - [ការបញ្ចេញ Phi ជាមួយ Kaito AKS, Azure Containers(គាំទ្រផ្លូវការ)](./md/01.Introduction/03/Kaito_Inference.md) +- [ចំណាត់ថ្នាក់ Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) + - [បំលែងចំណាត់ថ្នាក់ Phi-3.5 / 4 ដោយប្រើ llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [បំលែងចំណាត់ថ្នាក់ Phi-3.5 / 4 ដោយប្រើជំនួយ AI ផលិតផលសម្រាប់ onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [បំលែងចំណាត់ថ្នាក់ Phi-3.5 / 4 ដោយប្រើ Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [បំលែងចំណាត់ថ្នាក់ Phi-3.5 / 4 ដោយប្រើ Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- ការវាយតម្លៃ Phi + - [សម្រួល AI](./md/01.Introduction/05/ResponsibleAI.md) + - [Microsoft Foundry សម្រាប់ការវាយតម្លៃ](./md/01.Introduction/05/AIFoundry.md) + - [ប្រើ Promptflow សម្រាប់ការវាយតម្លៃ](./md/01.Introduction/05/Promptflow.md) -- RAG ជាមួយ Azure AI Search - - [របៀបប្រើ Phi-4-mini និង Phi-4-multimodal (RAG) ជាមួយ Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) - -- ឧទាហរណ៍ការអភិវឌ្ឍន៍កម្មវិធី Phi - - កម្មវិធីអត្ថបទ និងការជជែក - - ឧទាហរណ៍ Phi-4 - - [📓] [ជជែកជាមួយម៉ូដែល Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [ជជែកជាមួយម៉ូដែល Phi-4 ONNX នៅក្នុងតំបន់ .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [កម្មវិធីពិភាក្សា .NET Console ជាមួយ Phi-4 ONNX ប្រើ Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - ឧទាហរណ៍ Phi-3 / 3.5 - - [Chatbot តំបន់ក្នុងកម្មវិធីរុករក ប្រើ Phi3, ONNX Runtime Web និង WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [ម៉ូឌែលច្រើន - អន្ដរកម្ម Phi-3-mini និង OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - ការបង្កើត wrapper និងការប្រើ Phi-3 ជាមួយ MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [ចំណុចសំខាន់នៃម៉ូឌែល - របៀបបង្កើតម៉ូឌែល Phi-3-min សម្រាប់ ONNX Runtime Web ជាមួយ Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) +- RAG ជាមួយ Azure AI Search + - [របៀបប្រើ Phi-4-mini និង Phi-4-multimodal(RAG) ជាមួយ Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG ជាញឹកញាប់មិនពាក់ព័ន្ធកូនពពកជាមួយ SQLite FTS5 និង phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) + +- ឧទាហរណ៍ការអភិវឌ្ឍកម្មវិធី Phi + - កម្មវិធីអត្ថបទ និងជជែក + - ឧទាហរណ៍ Phi-4 + - [📓] [ជជែកជាមួយម៉ូដែល Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [ជជែកជាមួយម៉ូដែល Phi-4 ONNX នៅក្នុង NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [កម្មវិធី Console Chat .NET ជាមួយ Phi-4 ONNX ប្រើក្របខ័ណ្ឌ Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - ម៉ូដែល Phi-3 / 3.5 + - [បន្ទាត់ជជែកក្នុងកម្មវិធីរុករក ដោយប្រើ Phi3, ONNX Runtime Web និង WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [ជជែក OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [ម៉ូដែលច្រើន - Phi-3-mini អន្តរក्रिय និង OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - ការកសាង wrapper និងប្រើ Phi-3 ជាមួយ MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [ការបង្កើតម៉ូដែល - វិធីបង្កើតម៉ូដែល Phi-3-min សម្រាប់ ONNX Runtime Web ជាមួយ Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [កម្មវិធី WinUI3 ជាមួយ Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Multi Model AI Powered Notes App Sample](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [ការតំឡើងលម្អិត និងការរួមបញ្ចូលម៉ូឌែល Phi-3 ផ្ទាល់ខ្លួនជាមួយ Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [ការតំឡើងលម្អិត និងការរួមបញ្ចូលម៉ូឌែល Phi-3 ផ្ទាល់ខ្លួនជាមួយ Prompt flow ក្នុង Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [ការវាយតម្លៃម៉ូឌែល Phi-3 / Phi-3.5 តំឡើងថ្មីនៅ Microsoft Foundry មិនលើកវិធីរបស់ Microsoft របស់ AI](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + -[ឯកសារគំរូកម្មវិធី WinUI3 Multi Model AI Powered Notes](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [បង្រួមបង្រួម និង រួមបញ្ចូលម៉ូដែល Phi-3 ផ្ទាល់ខ្លួនជាមួយ Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [បង្រួមបង្រួម និង រួមបញ្ចូលម៉ូដែល Phi-3 ផ្ទាល់ខ្លួនជាមួយ Prompt flow នៅក្នុង Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [វាយតម្លៃម៉ូដែល Phi-3 / Phi-3.5 រួមបញ្ចូល នៅ Microsoft Foundry ដែលផ្ដោតលើគោលការណ៍ AI ទំនួលខុសត្រូវរបស់ Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [គំរូការព្យាករណ៍ភាសា Phi-3.5-mini-instruct (ចិន/អង់គ្លេស)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [ការប្រើកាត GPU របស់ Windows ដើម្បីបង្កើតដំណោះស្រាយ Prompt flow ជាមួយ Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [ការប្រើ Microsoft Phi-3.5 tflite ដើម្បីបង្កើតកម្មវិធី Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [ឧទាហរណ៍ Q&A .NET ដោយប្រើម៉ូឌែល ONNX Phi-3 ក្នុងដំណើរការពីស្រុក Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [កម្មវិធីឆាត console .NET ជាមួយ Semantic Kernel និង Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - - គំរូកូដ Azure AI Inference SDK - - គំរូ Phi-4 - - [📓] [បង្កើតកូដគម្រោងប្រើ Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - គំរូ Phi-3 / 3.5 - - [បង្កើត Visual Studio Code GitHub Copilot Chat ផ្ទាល់ខ្លួនជាមួយ Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [បង្កើត Visual Studio Code Chat Copilot Agent ផ្ទាល់ខ្លួនជាមួយ Phi-3.5 ដោយ GitHub Models](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - គំរូការយល់ដឹងខ្ពស់ - - គំរូ Phi-4 - - [📓] [គំរូ Phi-4-mini-reasoning ឬ Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [ការតំឡើងលម្អិត Phi-4-mini-reasoning ជាមួយ Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [ការតំឡើងលម្អិត Phi-4-mini-reasoning ជាមួយ Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [ប្រើ GPU វីនដូដើម្បីបង្កើតដំណោះស្រាយ Prompt flow ជាមួយ Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [ប្រើ Microsoft Phi-3.5 tflite ដើម្បីបង្កើតកម្មវិធី Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Q&A ឧទាហរណ៍ .NET ប្រើម៉ូដែល Phi-3 ONNX ផ្ទាល់ខ្លួន ដោយប្រើ Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [កម្មវិធីជជែក Console .NET ជាមួយ Semantic Kernel និង Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + + - ឧទាហរណ៍កូដ Azure AI Inference SDK + - ម៉ូដែល Phi-4 + - [📓] [បង្កើតកូដគម្រោងដោយប្រើ Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - ម៉ូដែល Phi-3 / 3.5 + - [បង្កើត Visual Studio Code GitHub Copilot Chat របស់អ្នកផ្ទាល់ជាមួយ Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [បង្កើត Visual Studio Code Chat Copilot Agent របស់អ្នកផ្ទាល់ជាមួយ Phi-3.5 ដោយ GitHub Models](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - ឧទាហរណ៍អភិវឌ្ឍចំណេះដឹងកម្រិតខ្ពស់ + - ម៉ូដែល Phi-4 + - [📓] [Phi-4-mini-reasoning ឬ ម៉ូដែល Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [បង្រួមបង្រួម Phi-4-mini-reasoning ជាមួយ Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [បង្រួមបង្រួម Phi-4-mini-reasoning ជាមួយ Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-reasoning ជាមួយ GitHub Models](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-reasoning ជាមួយ Microsoft Foundry Models](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - ការត្រួតពិនិត្យ - - [Phi-4-mini ដេមោ ដែលផ្ទុកនៅលើ Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal ដេមោ ដែលផ្ទុកនៅលើ Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - គំរូចក្ខុវិស័យ - - គំរូ Phi-4 - - [📓] [ប្រើ Phi-4-multimodal ដើម្បីអានរូបភាព និងបង្កើតកូដ](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - គំរូ Phi-3 / 3.5 + - សាកល្បង + - [ការ​បង្ហាញម៉ូដែល Phi-4-mini ផ្ទុក​នៅលើ Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [ការ​បង្ហាញម៉ូដែល Phi-4-multimodal ផ្ទុក​នៅលើ Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - ឧទាហរណ៍ក្នុងជំនាញមើល + - ម៉ូដែល Phi-4 + - [📓] [ប្រើ Phi-4-multimodal ដើម្បីអានរូបភាព និង បង្កើតកូដ](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - ម៉ូដែល Phi-3 / 3.5 - [📓][Phi-3-vision-រូបភាពអត្ថបទទៅអត្ថបទ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 ការប៉ុនប៉ងជីវពហុ](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - ជំនួយភាសាទស្សនៈ - ជាមួយ Phi3-Vision និង OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [DEMO: Phi-3 ការប្រើឡើងវិញ](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - ជំនួយភាសាវីស្យួល - ជាមួយ Phi3-Vision និង OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision ឧទាហរណ៍រូបភាពច្រើន ឬច្រើនសន្លឹក](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision ម៉ូឌែល ONNX ក្នុងស្រុក ប្រើ Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [មេនុយ ប្រើ Phi-3 Vision ម៉ូឌែល ONNX ក្នុងស្រុក ប្រើ Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Phi-3.5 Vision multi-frame ឬ multi-image គំរូ](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [ម៉ូដែល Phi-3 Vision ONNX ផ្ទាល់ខ្លួនប្រើ Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [ម៉ឺនុយសម្រាប់ម៉ូដែល Phi-3 Vision ONNX ផ្ទាល់ខ្លួនប្រើ Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - គំរូ Reasoning-Vision - - Phi-4-Reasoning-Vision-15B - - [📓] [ប្រើ Phi-4-Reasoning-Vision-15B ដើម្បីស្វែងរកការឆ្លងផ្លូវគន្លងជើង](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - ឧទាហរណ៍ Reasoning-Vision + - ម៉ូដែល Phi-4-Reasoning-Vision-15B + - [📓] [ប្រើ Phi-4-Reasoning-Vision-15B ដើម្បីស្វែងរកការឆ្លងផ្លូវមិនត្រឹមត្រូវ jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [ប្រើ Phi-4-Reasoning-Vision-15B ដើម្បីគណិតវិទ្យា](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - [📓] [ប្រើ Phi-4-Reasoning-Vision-15B ដើម្បីស្វែងរក UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - គំរូគណិតវិទ្យា - - គំរូ Phi-4-Mini-Flash-Reasoning-Instruct [ដេមោគណិតវិទ្យាជាមួយ Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - ឧទាហរណ៍គណិតវិទ្យា + - ម៉ូដែល Phi-4-Mini-Flash-Reasoning-Instruct [គំរូគណិតវិទ្យាជាមួយ Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - គំរូសូរ - - គំរូ Phi-4 - - [📓] [ការបម្រែបម្រួលសំឡេងតាម Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [គំរូសម្លេង Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - ឧទាហរណ៍សំឡេង + - ម៉ូដែល Phi-4 + - [📓] [ដកស្រង់អត្ថបទសំឡេងដោយប្រើ Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [គំរូសំឡេង Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [គំរូបកប្រែសំឡេង Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [កម្មវិធី console .NET ប្រើ Phi-4-multimodal សំឡេង ដើម្បីវិភាគឯកសារ និងបង្កើតច្បាប់](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - គំរូ MOE - - គំរូ Phi-3 / 3.5 - - [📓] [Phi-3.5 ម្សិចឡូចអេកសផឺរ(MoEs) គំរូប្រព័ន្ធបណ្ដាញសង្គម](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [ការបង្កើតបណ្ដាញ RAG ជាមួយ NVIDIA NIM Phi-3 MOE, Azure AI Search និង LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - គំរូការហៅមុខងារ - - គំរូ Phi-4 🆕 - - [📓] [ប្រើការហៅមុខងារជាមួយ Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [ប្រើការហៅមុខងារដើម្បីបង្កើត multi-agents ជាមួយ Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [ប្រើការហៅមុខងារជាមួយ Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [ប្រើការហៅមុខងារជាមួយ ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - - គំរូលាយបញ្ចូល Multimodal - - គំរូ Phi-4 🆕 - - [📓] [ប្រើ Phi-4-multimodal ជាជournalist បច្ចេកវិទ្យា](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [កម្មវិធី console .NET ប្រើ Phi-4-multimodal ដើម្បីវិភាគឯកសារ​សំឡេង និងបង្កើតអត្ថបទ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - ឧទាហរណ៍ MOE + - ម៉ូដែល Phi-3 / 3.5 + - [📓] [ម៉ូដែល Phi-3.5 Mixture of Experts Models (MoEs) គំរូប្រព័ន្ធផ្សព្វផ្សាយសង្គម](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [កសាងប្រព័ន្ធបង្កើតដែលបន្ថែមការយកព័ត៌មានក្រៅ (RAG) ជាមួយ NVIDIA NIM Phi-3 MOE, Azure AI Search និង LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - + - ឧទាហរណ៍អំពី ការហៅមុខងារ Function Calling + - ម៉ូដែល Phi-4 ថ្មី🆕 + - [📓] [ប្រើ Function Calling ជាមួយ Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [ប្រើ Function Calling ដើម្បីបង្កើត multi-agents ជាមួយ Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [ប្រើ Function Calling ជាមួយ Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [ប្រើ Function Calling ជាមួយ ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - ឧទាហរណ៍ចម្រះម៉ូឌែល Multimodal Mixing + - ម៉ូដែល Phi-4 ថ្មី🆕 + - [📓] [ប្រើ Phi-4-multimodal ជាអ្នកសារព័ត៌មានបច្ចេកវិទ្យា](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [កម្មវិធី console .NET ប្រើ Phi-4-multimodal ដើម្បីវិភាគរូបភាព](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- ការតំឡើងលម្អិត Phi - - [ស្ថានភាពការតំឡើងលម្អិត](./md/03.FineTuning/FineTuning_Scenarios.md) - - [ការតំឡើងលម្អិត ប្រៀបធៀប RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [ការតំឡើងលម្អិត ឲ្យ Phi-3 ក្លាយជាអ្នកជំនាញឧស្សាហកម្ម](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [ការតំឡើងលម្អិត Phi-3 ជាមួយ AI Toolkit សម្រាប់ VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [ការតំឡើងលម្អិត Phi-3 ជាមួយសេវាកម្ម Azure Machine Learning](./md/03.FineTuning/Introduce_AzureML.md) - - [ការតំឡើងលម្អិត Phi-3 ជាមួយ Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [ការតំឡើងលម្អិត Phi-3 ជាមួយ QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [ការតំឡើងលម្អិត Phi-3 ជាមួយ Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [ការតំឡើងលម្អិត Phi-3 ជាមួយ Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [ការតំឡើងលម្អិត ជាមួយ Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [ការតំឡើងលម្អិត ជាមួយ Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - - [ការតំឡើងលម្អិត Phi-3-vision ជាមួយ Weights និង Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [ការតំឡើងលម្អិត Phi-3 ជាមួយ Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [ការតំឡើងលម្អិត Phi-3-vision (គាំទ្រផ្លូវការជាផ្លូវការ)](./md/03.FineTuning/FineTuning_Vision.md) - - [ការកែលម្អ Phi-3 ជាមួយ Kaito AKS , Azure Containers(ការគាំទ្រផ្លូវការ)](./md/03.FineTuning/FineTuning_Kaito.md) - - [ការកែលម្អ Phi-3 និង 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - -- កិច្ចសិក្សាផ្ទាល់ - - [ការស្វែងរកគំរូកម្រិតខ្ពស់: LLMs, SLMs, ការអភិវឌ្ឍនៅមូលដ្ឋាន និងផ្សេងៗទៀត](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [ការបើកចំហរ​​​សក្តានុពល NLP៖ ការកែលម្អជាមួយ Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - -- សារព័ត៌មាន និងការបោះពុម្ពផ្សាយស្រាវជ្រាវសិក្សា - - [សៀវភៅសិក្សាគឺគ្រប់គ្រាន់ដែលអ្នកត្រូវការលេខរៀងទី II: របាយការណ៍បច្ចេកទេស phi-1.5](https://arxiv.org/abs/2309.05463) - - [របាយការណ៍បច្ចេកទេស Phi-3៖ គំរូភាសាមានសមត្ថភាពខ្ពស់នៅក្នុងទូរស័ព្ទរបស់អ្នក](https://arxiv.org/abs/2404.14219) +- ការបង្រួមបង្រួមម៉ូឌែល Phi + - [សេចក្ដីណែនាំអំពីការបង្រួមបង្រួម](./md/03.FineTuning/FineTuning_Scenarios.md) + - [ការបង្រួមបង្រួមទល់នឹង RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [អនុញ្ញាតឱ្យ Phi-3 ក្លាយជាអ្នកជំនាញឧស្សាហកម្ម](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [បង្រួមបង្រួម Phi-3 ជាមួយ AI Toolkit សម្រាប់ VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [បង្រួមបង្រួម Phi-3 ជាមួយ សេវាកម្ម Azure Machine Learning](./md/03.FineTuning/Introduce_AzureML.md) + - [បង្រួមបង្រួម Phi-3 ជាមួយ Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [បង្រួមបង្រួម Phi-3 ជាមួយ QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [បង្រួមបង្រួម Phi-3 ជាមួយ Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [បង្រួមបង្រួម Phi-3 ជាមួយ Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [បង្រួមបង្រួមជាមួយ Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [បង្រួមបង្រួមជាមួយ Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) + - [បង្រួមបង្រួម Phi-3-vision ជាមួយ Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [កំណត់ត្រាឡើងវិញ Phi-3 ជាមួយនឹងទូរស័ព្ទ Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [កំណត់ត្រាឡើងវិញ Phi-3-vision (ការគាំទ្រផ្លូវការជាផ្លូវការ)](./md/03.FineTuning/FineTuning_Vision.md) + - [កំណត់ត្រាឡើងវិញ Phi-3 ជាមួយនឹង Kaito AKS, Azure Containers (ការគាំទ្រផ្លូវការជាផ្លូវការ)](./md/03.FineTuning/FineTuning_Kaito.md) + - [កំណត់ត្រាឡើងវិញ Phi-3 និង 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + +- ការប្រឡងដោយដៃ + - [ពង្រីកម៉ូដែលកម្រិតខ្ពស់: LLMs, SLMs, local development និងផ្សេងទៀត](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [សមត្ថភាព NLP: កំណត់ត្រាលម្អិតជាមួយ Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + +- ព័ត៍មានឯកទេស និងការបោះពុម្ពផ្សាយ + - [មូលដ្ឋានសៀវភៅគឺគ្រប់គ្រងដែលអ្នកត្រូវការ II: របាយការណ៍បច្ចេកទេស phi-1.5](https://arxiv.org/abs/2309.05463) + - [របាយការណ៍បច្ចេកទេស Phi-3: ម៉ូដែលភាសាកាន់តែសមត្ថភាពនៅក្នុងទូរស័ព្ទរបស់អ្នក](https://arxiv.org/abs/2404.14219) - [របាយការណ៍បច្ចេកទេស Phi-4](https://arxiv.org/abs/2412.08905) - - [របាយការណ៍បច្ចេកទេស Phi-4-Mini៖ គំរូភាសាទ្រង់ទ្រាយតូច ប៉ុន្តែមានអំណាចខ្ពស់តាមរយៈការបញ្ចូលគ្នារបស់ LoRAs](https://arxiv.org/abs/2503.01743) - - [ការបង្ហាញលទ្ធផលល្អបំផុតសម្រាប់គំរូភាសាតូចសម្រាប់នាទីហៅមុខងារកន្លែងក្នុងឡាន](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) ការកែលម្អ PHI-3 សម្រាប់ការឆ្លើយសំណួរជម្រើសច្រើន៖ វិធីសាស្រ្តលទ្ធផល និងបញ្ហា](https://arxiv.org/abs/2501.01588) + - [របាយការណ៍បច្ចេកទេស Phi-4-Mini: ម៉ូដែលភាសាចម្រុះទាបប៉ុន្តែមានពន្លឺតាមរយៈ Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [បង្កើតសមត្ថភាពម៉ូដែលភាសារាប់តូចសម្រាប់ការហៅអនុគ្រោះនៅក្នុងយានយន្ត](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) កំណត់ត្រាឡើងវិញ PHI-3 សម្រាប់ការឆ្លើយសំណួរប្រចាំជាជម្រើសច្រើន: វិធីសាស្ត្រ, លទ្ធផល និងបញ្ហា](https://arxiv.org/abs/2501.01588) - [របាយការណ៍បច្ចេកទេស Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [របាយការណ៍បច្ចេកទេស Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## ការប្រើប្រាស់គំរូ Phi +## ការប្រើប្រាស់ម៉ូដែល Phi ### Phi នៅលើ Microsoft Foundry -អ្នកអាចស្វែងយល់ពីរបៀបប្រើ Microsoft Phi និងរបៀបសាងសង់ដំណោះស្រាយ E2E នៅលើឧបករណ៍របស់អ្នក តាមរយៈការលេងល្បិចជាមួយគំរូ និងកែប្រែ Phi ទៅតាមស្ថានភាពរបស់អ្នកដោយប្រើ [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) អ្នកអាចស្វែងយល់បន្ថែមពី Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +អ្នកអាចរៀនពីការប្រើប្រាស់ Microsoft Phi និងរបៀបសាងសង់ដំណោះស្រាយ E2E នៅលើឧបករណ៍របស់អ្នក។ ដើម្បីទទួលបទពិសោធន៍ Phi សម្រាប់ខ្លួនអ្នកផ្ទាល់ ចាប់ផ្តើមពិសោធន៍ជាមួយម៉ូដែល និងប្តូរតាមតម្រូវការខ្លួនអ្នកដោយប្រើ [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) អ្នកអាចរៀនបន្ថែមនៅ Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**ប្លង់លំហ** -គំរូរាល់គំរូមានប្លង់លំហដាក់ធ្វើតេស្តគំរូ [Azure AI Playground](https://aka.ms/try-phi3)។ +**កន្លែងលេង** +ម៉ូដែលនីមួយៗមានកន្លែងលេងឯកទេសសម្រាប់សាកល្បងម៉ូដែល [Azure AI Playground](https://aka.ms/try-phi3)។ -### Phi នៅលើ GitHub Models +### Phi នៅលើម៉ូដែល GitHub -អ្នកអាចស្វែងយល់ពីរបៀបប្រើ Microsoft Phi និងរបៀបសាងសង់ដំណោះស្រាយ E2E នៅលើឧបករណ៍របស់អ្នក តាមរយៈការលេងល្បិចជាមួយគំរូ និងកែប្រែ Phi ទៅតាមស្ថានភាពរបស់អ្នកដោយប្រើ [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) អ្នកអាចស្វែងយល់បន្ថែមពី Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +អ្នកអាចរៀនពីការប្រើប្រាស់ Microsoft Phi និងរបៀបសាងសង់ដំណោះស្រាយ E2E នៅលើឧបករណ៍ខុសៗគ្នារបស់អ្នក។ ដើម្បីទទួលបទពិសោធន៍ Phi សម្រាប់ខ្លួនអ្នកផ្ទាល់ ចាប់ផ្តើមសាកល្បងម៉ូដែល និងប្តូរតាមតម្រូវការខ្លួនអ្នកដោយប្រើ [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) អ្នកអាចរៀនបន្ថែមនៅ Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**ប្លង់លំហ** -គំរូរាល់គំរូមាន [ប្លង់លំហដាក់ធ្វើតេស្តគំរូ](/md/02.QuickStart/GitHubModel_QuickStart.md)។ +**កន្លែងលេង** +ម៉ូដែលនីមួយៗមាន [កន្លែងលេងសម្រាប់សាកល្បងម៉ូដែល](/md/02.QuickStart/GitHubModel_QuickStart.md)។ ### Phi នៅលើ Hugging Face -អ្នកក៏អាចស្វែងរកគំរូនៅលើ [Hugging Face](https://huggingface.co/microsoft) +អ្នកក៏អាចស្វែងរកម៉ូដែលនៅលើ [Hugging Face](https://huggingface.co/microsoft) -**ប្លង់លំហ** - [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**កន្លែងលេង** + [កន្លែងលេង Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) - ## 🎒 វគ្គសិក្សាផ្សេងទៀត + ## 🎒 មុខវិជ្ជាផ្សេងទៀត -ក្រុមរបស់យើងផលិតវគ្គសិក្សាផ្សេងទៀត! សូមពិនិត្យមើល៖ +ក្រុមរបស់យើងបង្កើតមុខវិជ្ជាផ្សេងទៀត! សូមពិនិត្យមើល: ### LangChain @@ -257,68 +259,71 @@ Phi ជាមួយនឹងភាពមានសមត្ថភាពខ្ព [![LangChain សម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agents +### Azure / Edge / MCP / ឧបករណ៍ភ្នាក់ងារ [![AZD សម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI សម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP សម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents សម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![ឧបករណ៍ភ្នាក់ងារ AI សម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### ស៊េរី AI បង្កើតថ្មី -[![AI បង្កើតថ្មីសម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI បង្កើតថ្មី (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![AI បង្កើតថ្មី (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![AI បង្កើតថ្មី (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### ស៊េរី AI បង្កើត +[![AI បង្កើតសម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI បង្កើត (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![AI បង្កើត (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![AI បង្កើត (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- -### ការសិក្សាសំខាន់ៗ +### ការរៀនមូលដ្ឋាន [![ML សម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![វិទ្យាសាស្ត្រទិន្នន័យសម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![វិទ្យាសាស្ត្រ​ទិន្នន័យ​សម្រាប់​អ្នក​ចាប់​ផ្តើម](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI សម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![សន្តិសុខបណ្ដាញសម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![ការអភិវឌ្ឍវែបសម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![សន្តិសុខបច្ចេកវិទ្យាសម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![អភិវឌ្ឍន៍បណ្ដាញសម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![IoT សម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![ការអភិវឌ្ឍ XR សម្រាប់អ្នកចាប់ផ្តើម](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- ### ស៊េរី Copilot -[![Copilot សម្រាប់កម្មវិធីបញ្ចូលជាដៃគូ AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot សម្រាប់ការសរសេរកម្មវិធីជាគូ AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot សម្រាប់ C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## AI ដ៏ទំនួលខុសត្រូវ +## ការប្រើប្រាស់ AI ដោយក្តីទទួលខុសត្រូវ -Microsoft មានការប្តេជ្ញាចិត្តជួយអតិថិជនប្រើផលិតផល AI របស់យើងយ៉ាងទំនួលខុសត្រូវ ចែករំលែកចំណេះដឹង និងបង្កើតដៃគូផ្អែកលើជំនឿទុកចិត្ត តាមរយៈឧបករណ៍ដូចជា Transparency Notes និង Impact Assessments ។ មានធនធានជាច្រើនដែលអាចរកបាននៅ [https://aka.ms/RAI](https://aka.ms/RAI)។ -វិធីចូលរួម AI ដ៏ទំនួលខុសត្រូវរបស់ Microsoft គឺផ្ដោតលើគ្រឹះធម្មតា AI របស់យើង រួមមាន ភាពយុត្តិធម៌ ភាពជឿនលឿន និងសុវត្ថិភាព គុណភាពឯកជន និងសុវត្ថិភាព ភាពរួមបញ្ចូល ភាពត្រឹមត្រូវ និងការទទួលខុសត្រូវ។ +Microsoft មានបំណងជួយអតិថិជនក្នុងការប្រើប្រាស់ផលិតផល AI របស់យើងដោយទទួលខុសត្រូវ ការចែករំលែកបទពិសោធន៍ និងការបង្កើតដៃគូជាស្ថាប័នផ្អែកលើការជឿជាក់តាមឧបករណ៍ដូចជាកំណត់ត្រាឯកភាព និងការវាយតម្លៃផលប៉ះពាល់។ ឧបករណ៍ជាច្រើនទាំងនេះអាចរកបាននៅ [https://aka.ms/RAI](https://aka.ms/RAI)។ +ជំហានរបស់ Microsoft ទៅរក AI ដោយក្តីទទួលខុសត្រូវគឺមានមូលដ្ឋានលើគោលការណ៍ AI របស់យើងដែលរួមមានតុល្យភាព ភាពទុកចិត្ត និងសុវត្ថិភាព ការសម្ងាត់ និងសុវត្ថិភាព ភាពរួមបញ្ចូល ​ភាពត្រួតពិនិត្យ និងការទទួលខុសត្រូវ។ -គំរូភាសាធំៗ នឹងរូបភាព និងសំឡេង - ដូចគំរូដែលបានប្រើក្នុងឧទាហរណ៍នេះ - អាចមានបរិបទដែលមិនយុត្តិធម៌ មិនជឿនលឿន ឬមានការត្អូញត្អែរ ដែលអាចបង្ករជាពុល។ សូមពិនិត្យ [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ដើម្បីទទួលបានព័ត៌មានអំពីហានិភ័យ និងកំណត់ចំណាំ។ -វិធីសាស្រ្តដែលបានណែនាំសម្រាប់កាត់បន្ថយហានិភ័យទាំងនេះគឺការរួមបញ្ចូលប្រព័ន្ធសុវត្ថិភាពក្នុងស្ថាបត្យកម្មរបស់អ្នកដែលអាចរកឃើញ និងរារាំងប្រតិបត្តិការបង្កគ្រោះថ្នាក់។ [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ផ្តល់ជាស្រទាប់ការពារឯករាជ្យ ដែលអាចរកឃើញមាតិកាដែលបង្កគ្រោះថ្នាក់ដែលបានបង្កើតដោយអ្នកប្រើ និង AI ក្នុងកម្មវិធីនិងសេវាកម្ម។ Azure AI Content Safety រួមបញ្ចូល API អត្ថបទ និងរូបភាពដែលអនុញ្ញាតឱ្យអ្នករកឃើញសារព័ត៌មានដែលមានគ្រោះថ្នាក់។ ខណៈនៅក្នុង Microsoft Foundry សេវាកម្ម Content Safety អនុញ្ញាតឱ្យអ្នកមើល ស្វែងយល់ និងសាកល្បងកូដឧទាហរណ៍សម្រាប់ការរកឃើញមាតិកាដែលបង្កគ្រោះថ្នាក់នៅក្នុងវិធីសាស្រ្តផ្សេងៗគ្នា។ ​ឯកសារបណ្តុះបណ្តាលរហ័សនេះ [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) នាំអ្នកឆ្ពោះទៅរកការធ្វើសំណើទៅសេវាកម្ម។ +ម៉ូដែលធំៗនៃភាសា, រូបភាព និងសំឡេង - ដូចដែលបានប្រើនៅក្នុងឧទាហរណ៍នេះ - អាចមានប្រតិកម្មដែលមិនត្រឹមត្រូវ មិនទុកចិត្ត ឬគម្របដោយអាក្រក់ ដែលអាចបណ្តាលឲ្យមានហានិភ័យ។ សូមពិនិត្យឯកសារ [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ដើម្បីទទួលបានព័ត៌មានអំពីហានិភ័យ និងកំណត់ខ្ចាប់ខ្ចាយ។ -មួយជាងនេះទៀតដែលត្រូវគិតគូរគឺប្រសិទ្ធភាពរបស់កម្មវិធីទាំងមូល។ ជាមួយកម្មវិធីម៉ុលទីម៉ូឌែល និងម៉ុលទីម៉ូឌាល់ យើងចាត់ទុកប្រសិទ្ធភាពថាជាការដែលប្រព័ន្ធបង្ហាញដូចដែលអ្នកនិងអ្នកប្រើប្រាស់របស់អ្នករំពឹងទុក រួមទាំងមិនបង្កើតលទ្ធផលដែលបង្កគ្រោះថ្នាក់។ វាសំខាន់ក្នុងការវាយតម្លៃប្រសិទ្ធភាពរបស់កម្មវិធីទាំងមូលរបស់អ្នកដោយប្រើ [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)។ អ្នកក៏មានសមត្ថភាពបង្កើតនិងវាយតម្លៃជាមួយ [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) ផងដែរ។ -អ្នកអាចវាយតម្លៃកម្មវិធី AI របស់អ្នកក្នុងបរិស្ថានអភិវឌ្ឍន៍របស់អ្នកដោយប្រើ [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html)។ ជាប្រភេទឧបករណ៍សម្រាប់បញ្ជាក់ ទិន្នន័យតេស្ត ឬគោលដៅ កំណត់ជំនាន់ AI ផលិតមកនៃកម្មវិធីរបស់អ្នកត្រូវបានវាស់ដោយចំនុចវាយតម្លៃស្រាប់ឬចំណាត់ថ្នាក់ផ្ទាល់ខ្លួនដែលអ្នកជ្រើសរើស។ ដើម្បីចាប់ផ្តើមជាមួយ azure ai evaluation sdk សម្រាប់វាយតម្លៃប្រព័ន្ធរបស់អ្នក អ្នកអាចអនុវត្តតាម [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)។ នៅពេលដែលអ្នកបើករត់ការវាយតម្លៃមួយ អ្នកអាច [មើលលទ្ធផលនៅក្នុង Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) បាន។ +វិធីសាស្រ្តណែនាំសម្រាប់កាត់បន្ថយហានិភ័យទាំងនេះគឺរួមបញ្ចូលប្រព័ន្ធសុវត្ថិភាពនៅក្នុងស្ថាបត្យកម្មរបស់អ្នកដែលអាចកំណត់ និងការពារជាងទម្លាក់អាកប្បកិរិយាអាក្រក់។ [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ផ្តល់ជារបារម្ភការពារឯករាជ្យ មកដែលអាចកំណត់មាតិកាដែលមានការខូចខាតពីក្រុមប្រើប្រាស់ និងបង្កើតដោយ AI នៅក្នុងកម្មវិធី និងសេវាកម្ម។ Azure AI Content Safety រួមបញ្ចូល API អត្ថបទ និងរូបភាពដែលអនុញ្ញាតិឱ្យអ្នកកំណត់វត្ថុដែលមានការខូចខាត។ នៅក្នុង Microsoft Foundry, សេវាផ្នែក Content Safety អនុញ្ញាតិឱ្យអ្នកមើល ស្វែងយល់ និងសាកល្បងគូដគំរូសម្រាប់ការកំណត់មាតិកាខូចខាតតាមរយៈរបៀបផ្សេងៗគ្នា។ [ឯកសារចាប់ផ្តើមរហ័ស](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ខាងក្រោមណែនាំអ្នកថាត្រូវបញ្ជូនសំណើទៅសេវាកម្មយ៉ាងដូចម្តេច។ -## ត្រេដម៉ាក +ផ្នែកមួយទៀតដែលត្រូវគិតរួមចំណែកគឺប្រសិទ្ធភាពទូទៅនៃកម្មវិធី។ ជាមួយកម្មវិធីដែលមានមុខងារពហុរូបភាព និងម៉ូដែលច្រើន យើងគិតថាប្រសិទ្ធភាពមានន័យថាប្រព័ន្ធដំណើរការដូចដែលអ្នក និងអ្នកប្រើប្រាស់របស់អ្នករំពឹងទុក រួមទាំងមិនបញ្ចេញលទ្ធផលខូចខាត។ វាសំខាន់ក្នុងការវាយតំលៃប្រសិទ្ធភាពកម្មវិធីទាំងមូលរបស់អ្នកដោយប្រើ [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)។ អ្នកក៏អាចបង្កើត និងវាយតំលៃជាមួយ [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) ផ្ទាល់ខ្លួន។ -គម្រោងនេះអាចមានត្រេដម៉ាក ឬរូបសញ្ញាសម្រាប់គម្រោង ផលិតផល ឬសេវាកម្ម។ ការប្រើប្រាស់ត្រេដម៉ាក ឬរូបសញ្ញារាង Microsoft ដោយមានការអនុញ្ញាត ត្រូវជាប់នឹងនិយមន័យនិងត្រូវអនុវត្តតាម [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)។ ការប្រើប្រាស់ត្រេដម៉ាក រឺរូបសញ្ញារបស់ Microsoft ក្នុងកំណែដែលបានកែប្រែក្នុងគម្រោងនេះមិនគួរធ្វើឱ្យមានការភាន់ច្រឡំនិងមិនបញ្ជាក់ទៅលើការឧបត្ថម្ភរបស់ Microsoft ទេ។ ការប្រើប្រាស់ត្រេដម៉ាក ឬរូបសញ្ញាផ្សេងទៀត ជាផ្នែករបស់ភាគីទីបី នឹងជាប់នឹងនយោបាយរបស់ភាគីទីបីនោះ។ +អ្នកអាចវាយតំលៃកម្មវិធី AI របស់អ្នកនៅក្នុងបរិយាកាសអភិវឌ្ឍដោយប្រើ [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html)។ អ្នកអាចវាស់ទំហំនៃការបង្កើតរបស់កម្មវិធី AI របស់អ្នកដោយប្រើតំណាងបង្ហាញ ប្រៀបធៀបលទ្ធផលជាផ្ទាល់នឹងសំណុំទិន្នន័យសាកល្បង ឬគោលដៅ ជាមួយអ្នកវាយតំលៃស្ដង់ដារ ឬអ្នកវាយតំលៃផ្ទាល់ខ្លួនដែលអ្នកជ្រើសរើស។ ដើម្បីចាប់ផ្តើមជាមួយ azure ai evaluation sdk ដើម្បីវាយតំលៃប្រព័ន្ធរបស់អ្នក អ្នកអាចអនុវត្តតាម [មគ្គុទេសក៍ចាប់ផ្តើមរហ័ស](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)។ ពេលដែលអ្នកបានបើកការវាយតំលៃមួយ អ្នកអាច [មើលលទ្ធផលនៅក្នុង Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)។ -## ការជួយផ្នែកសំណួរ +## សញ្ញាទំនិញ -បើអ្នកជួបកង្វល់ឬមានសំណួរអំពីការសង់កម្មវិធី AI សូមចូលរួម៖ +គម្រោងនេះអាចមានសញ្ញាទំនិញ ឬឡូហ្គូសម្រាប់គម្រោង ផលិតផល ឬសេវាកម្ម។ ការប្រើប្រាស់សញ្ញាទំនិញ ឬឡូហ្គូ Microsoft ដែលមានសិទ្ធិអនុម័តត្រូវគោរពទៅតាម [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)។ +ការប្រើប្រាស់សញ្ញាទំនិញ ឬឡូហ្គូ Microsoft ក្នុងគម្រោងដែលបានកែប្រែនោះមិនគួរបង្កការជ្រុលនិយម ឬបង្ហាញថាមានការឧបត្ថម្ភពីMicrosoftឡើយ។ ការប្រើប្រាស់សញ្ញាទំនិញ ឬឡូហ្គូរបស់បុគ្គលទីបីអាស្រ័យលើគោលការណ៍របស់បុគ្គលទីបីនោះ។ + +## សុំជំនួយ + +ប្រសិនបើអ្នកមានបញ្ហាជួបតែមិនដោះស្រាយ ឬមានសំនួរទាក់ទងនឹងការទាញយកកម្មវិធី AI សូមចូលរួម៖ [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -បើអ្នកមានមតិយោបល់ផលិតផល ឬកំហុសខណៈកំពុងសាងសង់ សូមចូលទៅ៖ +ប្រសិនបើអ្នកមានមតិយោបល់អំពីផលិតផល ឬកំហុសពេលសាងសង់ សូមចូលទៅកាន់៖ [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**ប្រកាស​ស្តី​ពី​ការទទួល​ខុសត្រូវ**ៈ -ឯកសារ​នេះ​ត្រូវ​បាន​បម្លែង​ភាសា​ដោយ​ប្រព័ន្ធ​បកប្រែ AI [Co-op Translator](https://github.com/Azure/co-op-translator)។ ទោះ​យើង​ព្យាយាម​រក្សា​ភាព​ត្រឹមត្រូវ​ក៏​ដោយ សូម​យល់​ព្រម​ថា​ការ​បកប្រែ​ដោយ​ស្វ័យ​ប្រវត្តិ​អាច​មាន​កំហុស ឬ​ភាព​មិន​ត្រឹមត្រូវ។ ឯកសារ​ដើម​នៅ​ក្នុង​ភាសា​ដើម​គួរត្រូវ​បាន​ពិចារណា​ជា​ធនធាន​អធិប្បាយ។ សម្រាប់​ព័ត៌មាន​សំខាន់ គឺ​ត្រូវ​បាន​ផ្ដល់អនុសាសន៍​ឲ្យ​ធ្វើ​ការ​បកប្រែ​ដោយ​បុគ្គលិក​ជំនាញ​មនុស្ស។ យើង​មិន​មាន​ភាព​ទទួល​ខុសត្រូវ​ចំពោះ​ការ​យល់ច្រឡំ ឬការ​បកប្រែ​ខុស​ពី​ការ​ប្រើប្រាស់​ការ​បកប្រែ​នេះឡើយ។ +**ការបដិសេធ**: +ឯកសារនេះត្រូវបានបម្លែងភាសា ដោយប្រើសេវាបម្លែងភាសា AI [Co-op Translator](https://github.com/Azure/co-op-translator)។ ទោះយើងខ្ញុំមានក្តីប្រាថ្នាឱ្យបានច្បាស់លាស់ តែសូមយល់ដឹងថាការបម្លែងដោយស្វ័យប្រវត្តិក៏អាចមានកំហុសឬភាពមិនត្រឹមត្រូវ។ ឯកសារដើមជាភាសាទីតាំងគួរត្រូវបានគេប្រើជាប្រភពច្បាស់លាស់។ សម្រាប់ព័ត៌មានសំខាន់ៗ សូមណែនាំឱ្យប្រើប្រាស់ការប្រែដោយមនុស្សជំនាញ។ យើងខ្ញុំមិនទទួលខុសត្រូវចំពោះការយល់ច្រឡំ ឬការបកស្រាយខុសបន្ទាប់ពីការប្រើប្រាស់ការបម្លែងនេះនោះទេ។ \ No newline at end of file diff --git a/translations/km/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/km/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..f00c1d5f6 --- /dev/null +++ b/translations/km/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 បញ្ចូល RAG ការរួមបញ្ចូល Microsoft phi-4-mini & SQLite FTS5 (SLM គ្មានពពក)\n", + "\n", + "> **អ្នករៀបរៀង៖** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **ផ្តោតសំខាន់៖** ម៉ូដែលភាសាមានទំហំតូច (SLMs), SQLite FTS5 BM25, សន្និដ្ឋានដិត (Dense Embeddings), ការរួមបញ្ចូលចំណាត់ថ្នាក់វិញ (Reciprocal Rank Fusion - RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. 动機：ការលំបាកក្នុងការចងចាំពាក្យគន្លឹះនៅក្នុង SLM ទីលាន\n", + "រចនាសម្ព័ន្ធ RAG ស្តង់ដារដែលពឹងផ្អែកតែលើសន្និដ្ឋានដិតវ៉ិចទ័រមានជាញឹកញាប់បរាជ័យក្នុងការទាញយកលេខកំណត់ត្រឹមត្រូវ (ឧ. `2,340,000 TL`, ឯកសារហត្ថលេខា, លេខគណនី)។ \n", + "ផ្ទុយទៅវិញ ការស្វែងរកពាក្យសរសេរមានប្រយោជន៍តិច (BM25) មិនអាចចាប់យកពាក្យមានន័យដូចគ្នា និងសំណួរមានការប្តូរពាក្យបានទេ។\n", + "\n", + "បង្កប់នេះបង្ហាញពីរបៀបអនុវត្តម៉ូទ័រការទាញយក hybrid ប្រសិទ្ធភាពល្បឿនខ្ពស់ ដោយបញ្ចូល៖\n", + "1. **វ៉ិចទ័រដិត** (Cosine Similarity)\n", + "2. **ស្វែងរកពាក្យសរសេរ BM25** (SQLite FTS5)\n", + "3. **ការរួមបញ្ចូលចំណាត់ថ្នាក់វិញ (RRF, $k=60$)**\n", + "4. **ការបង្កើតសម្រង់យោងដែលមានមូលដ្ឋាន (`[1]`, `[2]`)** ជាមួយ Microsoft `phi-4-mini`។\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. ស្កីមា SQLite គូ (វ៉ិចទ័រមាប់ + តារាង FTS5 BM25 វេជ្ជសាស្រ្ត)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. សំណុំគំរូសម្រាប់បញ្ចូល និងវាយតម្លៃការប្រតិបត្តិការនាំចេញ\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. ការបង្កើតសំណួរដែលមានមូលដ្ឋានសម្រាប់ Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**ការបដិសេធ**:\nឯកសារនេះត្រូវបានបម្លែងភាសា ដោយប្រើសេវាបម្លែងភាសា AI [Co-op Translator](https://github.com/Azure/co-op-translator)។ ទោះយើងខ្ញុំមានក្តីប្រាថ្នាឱ្យបានច្បាស់លាស់ តែសូមយល់ដឹងថាការបម្លែងដោយស្វ័យប្រវត្តិក៏អាចមានកំហុសឬភាពមិនត្រឹមត្រូវ។ ឯកសារដើមជាភាសាទីតាំងគួរត្រូវបានគេប្រើជាប្រភពច្បាស់លាស់។ សម្រាប់ព័ត៌មានសំខាន់ៗ សូមណែនាំឱ្យប្រើប្រាស់ការប្រែដោយមនុស្សជំនាញ។ យើងខ្ញុំមិនទទួលខុសត្រូវចំពោះការយល់ច្រឡំ ឬការបកស្រាយខុសបន្ទាប់ពីការប្រើប្រាស់ការបម្លែងនេះនោះទេ។\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/kn/.co-op-translator.json b/translations/kn/.co-op-translator.json index 273bd69e4..2e75d2830 100644 --- a/translations/kn/.co-op-translator.json +++ b/translations/kn/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "kn" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T08:09:55+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:03:55+00:00", "source_file": "README.md", "language_code": "kn" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "kn" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:05:01+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "kn" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-12-21T15:51:57+00:00", diff --git a/translations/kn/README.md b/translations/kn/README.md index 4afb5bcfc..5d05ad5f1 100644 --- a/translations/kn/README.md +++ b/translations/kn/README.md @@ -1,324 +1,329 @@ -# ಫೈ ಕುಕ್‌ಬುಕ್: ಮೈಕ್ರೋಸಾಫ್ಟ್‌ನ ಫೈ ಮಾದರಿಗಳೊಂದಿಗೆ ಕೈಯಿಂದ ಅನುಭವಿಸುವ ಉದಾಹರಣೆಗಳು +# ಫೈ ಕುಕ್‌ಬುಕ್: ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೈ ಮಾದರಿಗಳೊಂದಿಗೆ ಹ್ಯಾಂಡ್ಸ್-ಆನ್ ಉದಾಹರಣೆಗಳು -[![GitHub ಕೋಡ್‌ಸ್ಪೇಸಸ್‌ನಲ್ಲಿಯೇ ನಮೂನೆಗಳನ್ನು ತೆರೆಯುವ ಮತ್ತು ಬಳಸುವದು](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![ಡಿವ್ ಕಂಟೇನರ್‌ಗಳಲ್ಲಿ ತೆರೆಯಿರಿ](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![GitHub ಕೋಡ್ಸ್ಪೇಸಗಳಲ್ಲಿ ಉದಾಹರಣೆಗಳನ್ನು ತೆರೆಯಿ ಮತ್ತು ಬಳಸಿಕೊಳ್ಳಿ](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![ಡೆವ್ ಕಂಟೈನರ್‌ಗಳಲ್ಲಿ ತೆರೆಯಿರಿ](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub ಕೊಡುಗೆಗಾರರು](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub ಕೊಡುಗೆದಾರರು](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub ಸಮಸ್ಯೆಗಳು](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ಪುಲ್-ವಿನಂತಿಗಳು](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![ಪುಲ್ ವಿನಂತಿಗಳನ್ನು ಸ್ವಾಗತಿಸಲಾಗುತ್ತದೆ](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub ಪಲ್ಲಟ ವಿನಂತಿಗಳು](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![ಪಿಆರ್ಸ್‌ಗೆ ಸ್ವಾಗತ](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![GitHub ವಾಚರ್ಸ್](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ಫೋರ್ಕ್‌ಗಳು](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ಸ್ಟಾರ್ಸ್](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub ಫಾಂಕ್ಸ್](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub ತಾರೆಗಳು](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -ಫೈ ಮೈಕ್ರೋಸಾಫ್ಟ್ ಅಭಿವೃದ್ಧಿಪಡಿಸಿದ ಒಪ್ಪನ್ ಸೋರ್ಸ್ AI ಮಾದರಿಗಳ ಸರಣಿಯಾಗಿದೆ. +ಫೈ ಮೈಕ್ರೋಸಾಫ್ಟ್ ಅಭಿವೃದ್ಧಿಪಡಿಸಿರುವ ಸರಣಿಯ ಓಪನ್ ಸೋರ್ಸ್ AI ಮಾದರಿಗಳಾಗಿವೆ. -ಫೈ ಪ್ರಸ್ತುತ ಅತ್ಯಂತ ಶಕ್ತಿಶಾಲಿ ಮತ್ತು ಖರ್ಚಿನ ಅನುಕೂಲಕಾರಿ ಸಣ್ಣ ಭಾಷಾ ಮಾದರಿ (SLM), ಬಹುಭಾಷಾ, ತರ್ಕ, ಪಠ್ಯ/ಚಾಟ್ ಉತ್ಪಾದನೆ, ಕೋಡಿಂಗ್, ಚಿತ್ರಗಳು, ಶಬ್ದ ಮತ್ತು ಇತರ ಸಂವೇದನಶೀಲ ಪರಿಕಲ್ಪನೆಗಳಲ್ಲಿ ಅತ್ಯುತ್ತಮ ಪ್ರಮಾಣಿಕತೆ ಗಳಿಸುವದು. +ಫೈ ಪ್ರಸ್ತುತ ಅತ್ಯಂತ ಶಕ್ತಿಶಾಲಿ ಮತ್ತು ವೆಚ್ಚ ಪರಿಣಾಮಕಾರಿಯಾದ ಸಣ್ಣ ಭಾಷಾ ಮಾದರಿ (SLM) ಆಗಿದ್ದು, ಬಹುಭಾಷಾ, ತರ್ಕ, ಪಠ್ಯ/ಚಾಟ್ ಜನರೇಶನ್, ಕೋಡಿಂಗ್, ಚಿತ್ರಗಳು, ಧ್ವನಿ ಮತ್ತು ಇತರೆ ಪರಿಸ್ಥಿತಿಗಳಲ್ಲಿ ತುಂಬಾ ಉತ್ತಮ ಗುರ್ತಾಂಕಗಳನ್ನು ಹೊಂದಿದೆ. -ನೀವು ಫೈನನ್ನು ಕ್ಲೌಡ್ ಅಥವಾ ಎಡ್ಜ್ ಸಾಧನಗಳಿಗೆ ನಿಯೋಜಿಸಬಹುದು ಮತ್ತು ಸಂಬಂಧಿತ ಗಣಕ ಶಕ್ತಿಯ ನಿಯಂತ್ರಣದಲ್ಲಿ ಸೃಜನಶೀಲ AI ಅಪ್ಲಿಕೇಶನ್‌ಗಳನ್ನು ಸುಲಭವಾಗಿ ನಿರ್ಮಿಸಬಹುದು. +ನೀವು ಫೈ ಅನ್ನು ಕ್ಲೌಡ್ ಅಥವಾ ಎಡ್ಜ್ ಸಾಧನಗಳಿಗೆ ನಿಯೋಜಿಸಬಹುದು, ಮತ್ತು ಕಡಿಮೆ ಗಣನಾ ಶಕ್ತಿಯೊಂದಿಗೆ ಸುಲಭವಾಗಿ ಜನರೇಟಿವ್ AI ಅಪ್ಲಿಕೇಶನ್‌ಗಳನ್ನು ನಿರ್ಮಿಸಬಹುದು. -ಈ ಸಂಪನ್ಮೂಲಗಳನ್ನು ಬಳಸಲು ಪ್ರಾರಂಭಿಸಲು ಈ ಹಂತಗಳನ್ನು ಅನುಸರಿಸಿ: -1. **ರೆಪೋಸಿಟರಿಯನ್ನು ಫೋರ್ಕ್ ಮಾಡಿ**: ಕ್ಲಿಕ್ ಮಾಡಿ [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **ರೆಪೋಸಿಟರಿಯನ್ನು ಕ್ಲೋನ್ ಮಾಡಿ**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**ಮೈಕ್ರೋಸಾಫ್ಟ್ AI Discord ಸಮುದಾಯದಲ್ಲಿ ಸೇರಿ ಹಾಗೂ ತಜ್ಞರು ಮತ್ತು ಸಹ ಡೆವಲಪರ್‌ಗಳನ್ನು ಭೇಟಿ ಮಾಡಿ**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +ಈ ಸಂಪನ್ಮೂಲಗಳನ್ನು ಬಳಸಲು ಈ ಹಂತಗಳನ್ನು ಅನುಸರಿಸಿ: +1. **ರಿಪೋಜಿಟರಿಯನ್ನು ಫಾರ್ಕ್ ಮಾಡಿ**: ಕ್ಲಿಕ್ ಮಾಡಿ [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **ರಿಪೋಜಿಟರಿಯನ್ನು ಕ್ಲೋನ್ ಮಾಡಿ**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Microsoft AI Discord ಸಮುದಾಯಕ್ಕೆ ಸೇರಿ ಮತ್ತು ತಜ್ಞರನ್ನು ಮತ್ತು ಸಹ-ವಿಕಾಸಕರನ್ನು ಭೇಟಿಮಾಡಿ**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/kn/cover.eb18d1b9605d754b.webp) ### 🌐 ಬಹುಭಾಷಾ ಬೆಂಬಲ -#### GitHub ಕ್ರಿಯೆಯಿಂದ ಬೆಂಬಲಿಸಲಾಗಿದೆ (ಸ್ವಯಂಚಾಲಿತ ಮತ್ತು ಯಾವಾಗಲೂ ನವೀಕರಿಸಲಾಗಿದೆ) +#### GitHub ಕ್ರಿಯೆ ಮೂಲಕ ಬೆಂಬಲಿಸಲಾಗುತ್ತದೆ (ಸ್ವಯಂಚಾಲಿತ ಮತ್ತು எப்போதும் ನವೀಕರಿಸಿಬರುತ್ತಾನೆ) -[ಅರೆಬಿಕ್](../ar/README.md) | [ಬೆಂಗಾಲಿ](../bn/README.md) | [ಬುಲ್ಗೇರಿಯನ್](../bg/README.md) | [ಬರ್ಮೀಸ್ (ಮಿಯಾದರ್)](../my/README.md) | [ಚೈನೀಸ್ (ಸರಳೀಕೃತ)](../zh-CN/README.md) | [ಚೈನೀಸ್ (ಪಾರಂಪರಿಕ, ಹೊಂಡ್ ಕಾಂಗ್)](../zh-HK/README.md) | [ಚೈನೀಸ್ (ಪಾರಂಪರಿಕ, ಮಕಾವ್)](../zh-MO/README.md) | [ಚೈನೀಸ್ (ಪಾರಂಪರಿಕ, ತೈವಾನ್)](../zh-TW/README.md) | [ಕ್ರೋಶಿಯನ್](../hr/README.md) | [ಚೆಕ್](../cs/README.md) | [ಡೆನಿಷ್](../da/README.md) | [ಡಚ್](../nl/README.md) | [ಎಸ್ಟೋನಿಯನ್](../et/README.md) | [ಫಿನ್ನಿಶ್](../fi/README.md) | [ಫ್ರೆಂಚ್](../fr/README.md) | [ಜರ್ಮನ್](../de/README.md) | [ग्रीಕ್](../el/README.md) | [ಹಿವ್ರೂ](../he/README.md) | [ಹಿಂದಿ](../hi/README.md) | [ಹಂಗೇರಿಯನ್](../hu/README.md) | [ಇಂಡोನೇಶಿಯನ್](../id/README.md) | [ಇಟಾಲಿಯನ್](../it/README.md) | [ಜಾಪನೀಸ್](../ja/README.md) | [ಕನ್ನಡ](./README.md) | [ಖ್ಮೇರ್](../km/README.md) | [ಕೋರಿಯನ್](../ko/README.md) | [ಲಿಥುವೇನಿಯನ್](../lt/README.md) | [ಮಲಯ್](../ms/README.md) | [ಮಲಯಾಳಂ](../ml/README.md) | [ಮರाठी](../mr/README.md) | [ನೇಪಾಳಿ](../ne/README.md) | [ನೈಜೀರಿಯನ್ ಪിഡ್ಜಿನ್](../pcm/README.md) | [ನಾರ್ವೇಜಿಯನ್](../no/README.md) | [ಪರ್ಶಿಯನ್ (ಫಾರ್ಸಿ)](../fa/README.md) | [ಪೋಲಿಶ್](../pl/README.md) | [ಪೋರ್ಚುಗೀಸ್ (ಬ್ರೆಜಿಲ್)](../pt-BR/README.md) | [ಪೋರ್ಚುಗೀಸ್ (ಪೋರ್ಚುಗೆಲ)](../pt-PT/README.md) | [ಪಂಜಾಬಿ (ಗುರಮುಖಿ)](../pa/README.md) | [ರೋಮೇನಿಯನ್](../ro/README.md) | [ರಶಿಯನ್](../ru/README.md) | [ಸರ್ಬಿಯನ್ (ಸಿರಿಲಿಕ್)](../sr/README.md) | [ಸ್ಲೋವಾಕ್](../sk/README.md) | [ಸ್ಲೋವೇನಿಯನ್](../sl/README.md) | [ಸ್ಪ್ಯಾನಿಷ್](../es/README.md) | [ಸ್ವಾಹಿಲಿ](../sw/README.md) | [ಸ್ವೀಡಿಶ್](../sv/README.md) | [ಟಗಾಲೋಗ್ (ಫಿಲಿಪಿನೋ)](../tl/README.md) | [ತಮಿಳು](../ta/README.md) | [ತೆಲುಗು](../te/README.md) | [ಥಾಯಿ](../th/README.md) | [ಟರ್ಕಿಷ್](../tr/README.md) | [ಯುಕ್ರೈನಿಯನ್](../uk/README.md) | [ಉರ್ದು](../ur/README.md) | [ವಿಯೆಟ್ನಾಮೀಸ್](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](./README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **ಸ್ಥಳೀಯವಾಗಿ ಕ್ಲೋನ್ ಮಾಡಬೇಕೆಂದು ಇಷ್ಟವಿದೆಯೆ?** +> **ಸ್ಥಳೀಯವಾಗಿ ಕ್ಲೋನ್ ಮಾಡಲು ಇಷ್ಟವಿದೆಯೇ?** > -> ಈ ರೆಪೋಸಿಟರಿ 50ಕ್ಕಿಂತ ಹೆಚ್ಚು ಭಾಷಾ ಅನುವಾದಗಳನ್ನು ಒಳಗೊಂಡಿದೆ ಇದು ಡೌನ್ಲೋಡ್ ಗಾತ್ರವನ್ನು ಬಹಳ ಹೆಚ್ಚಿಸುತ್ತದೆ. ಅನುವಾದಗಳಿಲ್ಲದೆ ಕ್ಲೋನ್ ಮಾಡಲು ಸ್ಪಾರ್ಸ್ ಔಟ್‌ಚೆಕ್ ಅನ್ನು ಬಳಸಿರಿ: +> ಈ ರಿಪೋಜಿಟರಿ 50+ ಭಾಷಾ ಅನುವಾದಗಳನ್ನು ಒಳಗೊಂಡಿದೆ, ಇದು ಡೌನ್ಲೋಡ್ ಗಾತ್ರವನ್ನು ಬಹಳಷ್ಟು ಹೆಚ್ಚಿಸುತ್ತದೆ. ಅನುವಾದಗಳಿಲ್ಲದೆ ಕ್ಲೋನ್ ಮಾಡಲು, ಸ್ಪಾರ್ಸ್ ಔಟ್‌ಪರ್ಸ್ ಅನ್ನು ಬಳಸಿ: > -> **ಬಾಶ್ / macOS / ಲಿನಕ್ಸ:** ->> ```bash +> **ಬಾಷ್ / ಮ್ಯಾಕ್‌ಒಎಸ್ / ಲಿನಕ್ಸ್:** +> ```bash > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` > -> **CMD (ವಿಂಡೋಸ್):** ->> ```cmd +> **CMD (ವಿಂಡೋಸ್):** +> ```cmd > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> ಇದು ನೀವು ಅವಶ್ಯಕವಾದ ಎಲ್ಲಾ ವಿಷಯಗಳನ್ನು ಬಹುಶೀಘ್ರ ಡೌನ್ಲೋಡ್ ಮಾಡಿ ಪೂರ್ಣಗೊಳಿಸಲು ಸಹಾಯ ಮಾಡುತ್ತದೆ. +> ಇದರಿಂದ ನೀವು ಕೋರ್ಸ್ನೆ ಪೂರ್ಣಗೊಳಿಸುವುದಕ್ಕಾಗಿ ಅಗತ್ಯವಿರುವ ಎಲ್ಲವನ್ನು much ವೇಗವಾಗಿ ಡೌನ್ಲೋಡ್ ಮಾಡಬಹುದು. -## ವಿಷಯಗಳ ಪಟ್ಟಿಯು +## ವಿಷಯಗಳ ಪಟ್ಟಿ - ಪರಿಚಯ - - [ಫೈ ಕುಟುಂಬಕ್ಕೆ ಸುಸ್ವಾಗತ](./md/01.Introduction/01/01.PhiFamily.md) - - [ನಿಮ್ಮ ವಾತಾವರಣವನ್ನು ಸಜ್ಜುಗೊಳಿಸುವುದು](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [ಮುಖ್ಯ ತಂತ್ರಜ್ಞಾನಗಳನ್ನು ಅರ್ಥಮಾಡಿಕೊಳ್ಳುವಿಕೆ](./md/01.Introduction/01/01.Understandingtech.md) - - [ಫೈ ಮಾದರಿಗಳಿಗಾಗಿ AI ಸುರಕ್ಷತೆ](./md/01.Introduction/01/01.AISafety.md) - - [ફೈ ಹಾರ್ಡ್‌ವೇರ್ ಬೆಂಬಲ](./md/01.Introduction/01/01.Hardwaresupport.md) - - [ફೈ ಮಾದರಿಗಳು ಮತ್ತು ವಿವಿಧ ವೇದಿಕೆಗಳಲ್ಲಿ ಲಭ್ಯತೆ](./md/01.Introduction/01/01.Edgeandcloud.md) + - [ಫೈ ಕುಟುಂಬಕ್ಕೆ ಸ್ವಾಗತ](./md/01.Introduction/01/01.PhiFamily.md) + - [ನಿಮ್ಮ ಪರಿಸರವನ್ನು ಹೊಂದಿಸುವುದು](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [ಮುಖ್ಯ ತಂತ್ರಜ್ಞಾನಗಳ ಅರ್ಥಮಾಡಿಕೋವುದು](./md/01.Introduction/01/01.Understandingtech.md) + - [ಫೈ ಮಾದರಿಗಳಿಗೆ AI ಸುರಕ್ಷತೆ](./md/01.Introduction/01/01.AISafety.md) + - [ಫೈ ಹಾರ್ಡ್‌ವೇರ್ ಬೆಂಬಲ](./md/01.Introduction/01/01.Hardwaresupport.md) + - [ಫೈ ಮಾದರಿಗಳು ಮತ್ತು ವೇದಿಕೆಗಳ ವ್ಯಾಪಕ ಲಭ್ಯತೆ](./md/01.Introduction/01/01.Edgeandcloud.md) - [Guidance-ai ಮತ್ತು ಫೈ ಬಳಕೆ](./md/01.Introduction/01/01.Guidance.md) - - [GitHub ಮಾರುಕಟ್ಟೆ ಮಾದರಿಗಳು](https://github.com/marketplace/models) - - [Azure AI ಮಾದರಿ ಕ್ಯಾಟಲಾಗ್](https://ai.azure.com) + - [GitHub ಮಾರುಕಟ್ಟೆ Models](https://github.com/marketplace/models) + - [ಅಜೂರ್ AI ಮಾದರಿ ಕ್ಯಾಟಲೋಗ್](https://ai.azure.com) -- ವಿವಿಧ ವಾತಾವರಣಗಳಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್ - - [ಹಗ್ಗಿಂಗ್ ಫೇಸ್](./md/01.Introduction/02/01.HF.md) +- ವಿವಿಧ ಪರಿಸರಗಳಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್ + - [ಹಗಿಂಗ್ ಫೇಸ್](./md/01.Introduction/02/01.HF.md) - [GitHub ಮಾದರಿಗಳು](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry ಮಾದರಿ ಕ್ಯಾಟಲಾಗ್](./md/01.Introduction/02/03.AzureAIFoundry.md) - - [ಒಲ್ಲಾಮಾ](./md/01.Introduction/02/04.Ollama.md) - - [AI ಟೂಲ್‌ಕಿಟ್ VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [Microsoft Foundry ಮಾದರಿ ಕ್ಯಾಟಲೋಗ್](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [ಒಳ್ಳಾಮಾ](./md/01.Introduction/02/04.Ollama.md) + - [AI ಟೂಲ್ಕಿಟ್ VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [ಫೌಂಡ್ರಿ ಲೊಕಲ್](./md/01.Introduction/02/07.FoundryLocal.md) + - [Foundry ಲೊಕಲ್](./md/01.Introduction/02/07.FoundryLocal.md) - ಫೈ ಕುಟುಂಬ ಇನ್ಫರೆನ್ಸ್ - [iOS ನಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/iOS_Inference.md) - - [ಆಂಡ್ರಾಯ್ಡ್‌ನಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Android_Inference.md) + - [ಆಂಡ್ರಾಯ್ಡ್ ನಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Android_Inference.md) - [ಜೆಟ್ಸನ್‌ನಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Jetson_Inference.md) - - [AI PC ನಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/AIPC_Inference.md) - - [ಆಪಲ್ MLX ಫ್ರೇಮ್ವರ್ಕ್ ಬಳಸಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/MLX_Inference.md) - - [ಲೋಕಲ್ ಸರ್ವರ್‌ನಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI ಟೂಲ್‌ಕಿಟ್ ಬಳಸಿ ರಿಮೋಟ್ ಸರ್ವರ್‌ನಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Remote_Interence.md) - - [ರಸ್ಟ್ ಬಳಸಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Rust_Inference.md) - - [ಲೋಕಲಲ್ಲಿಯೇ ಫೈ-ವಿಷನ್ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Vision_Inference.md) - - [ಕೈಟೊ AKS, Azure Containers (ಆಧಿಕೃತ ಬೆಂಬಲ) ಬಳಸಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Kaito_Inference.md) -- [ಫೈ ಕುಟುಂಬದ ಪ್ರಮಾಣೀಕರಣ](./md/01.Introduction/04/QuantifyingPhi.md) - - [llama.cpp ಬಳಸಿ ಫೈ-3.5 / 4 ಪ್ರಮಾಣೀಕರಣ](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime ಗೆನೆರೇಟಿವ್ AI ವಿಸ್ತರಣೆಗಳ ಬಳಕೆಯಿಂದ ಫೈ-3.5 / 4 ಪ್ರಮಾಣೀಕರಣ](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [ಇಂಟೆಲ್ ಓಪನ್ ವಿನೋ ಬಳಸಿ ಫೈ-3.5 / 4 ಪ್ರಮಾಣೀಕರಣ](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [ಆಪಲ್ MLX ಫ್ರೇಮ್ವರ್ಕ್ ಬಳಸಿ ಫೈ-3.5 / 4 ಪ್ರಮಾಣೀಕರಣ](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [AI ಪಿಸಿ ನಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/AIPC_Inference.md) + - [ಆಪಲ್ MLX ಫ್ರೇ임ವರ್ಕ್‌ನೊಂದಿಗೆ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/MLX_Inference.md) + - [ಲೊಕಲ್ ಸರ್ವರ್‌ನಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Local_Server_Inference.md) + - [AI ಟೂಲ್ಕಿಟ್ ಬಳಸಿ ದೂರಸ್ಥ ಸರ್ವರ್‌ನಲ್ಲಿ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Remote_Interence.md) + - [ರಸ್ಟ್ ನೊಂದಿಗೆ ಫೈ ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Rust_Inference.md) + - [ಸ್ಥಳೀಯ Vision Phi ಇನ್ಫರೆನ್ಸ್](./md/01.Introduction/03/Vision_Inference.md) + - [ಕೈಟೋ AKS, ಅಜೂರ್ ಕಂಟೈನರ್‌ಗಳೊಂದಿಗೆ ಫೈ ಇನ್ಫರೆನ್ಸ್ (ಅಧिकारिक ಬೆಂಬಲ)](./md/01.Introduction/03/Kaito_Inference.md) +- [ಫೈ ಕುಟುಂಬದ ಮೌಲ್ಯಮಾಪನ](./md/01.Introduction/04/QuantifyingPhi.md) + - [llama.cpp ಬಳಸಿ Phi-3.5 / 4 ಅನ್ನು ಮೌಲ್ಯಮಾಪನ ಮಾಡುವುದು](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntime ಗಾಗಿ ಜನರೇಟಿವ್ AI ವಿಸ್ತರಣೆಗಳೊಂದಿಗೆ Phi-3.5 / 4 ಅನ್ನು ಮೌಲ್ಯಮಾಪನ ಮಾಡುವುದು](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Intel OpenVINO ಬಳಸಿ Phi-3.5 / 4 ಅನ್ನು ಮೌಲ್ಯಮಾಪನ ಮಾಡುವುದು](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [ಆಪಲ್ MLX ಫ್ರೇ임ವರ್ಕ್ ಬಳಸಿ Phi-3.5 / 4 ಅನ್ನು ಮೌಲ್ಯಮಾಪನ ಮಾಡುವುದು](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - ಫೈ ಮೌಲ್ಯಮಾಪನ - - [ಜವಾಬ್ದಾರಿ AI](./md/01.Introduction/05/ResponsibleAI.md) - - [ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೌಂಡ್ರಿ ಮೌಲ್ಯಮಾಪನಕ್ಕಾಗಿ](./md/01.Introduction/05/AIFoundry.md) - - [ಪ್ರಾಂಪ್ಟ್‌ಫ್ಲೋ ಬಳಸಿ ಮೌಲ್ಯಮಾಪನ](./md/01.Introduction/05/Promptflow.md) + - [ಜವಾಬ್ದಾರಿಯುತ AI](./md/01.Introduction/05/ResponsibleAI.md) + - [ಮೌಲ್ಯಮಾಪನಕ್ಕಾಗಿ Microsoft Foundry](./md/01.Introduction/05/AIFoundry.md) + - [ಮೌಲ್ಯಮಾಪನಕ್ಕಾಗಿ Promptflow ಬಳಸು](./md/01.Introduction/05/Promptflow.md) -- Azure AI ಶೋಧನ ಸಹಿತ RAG - - [Phi-4-mini ಮತ್ತು Phi-4-multimodal(RAG) ಅನ್ನು Azure AI ಶೋಧನೊಂದಿಗೆ ಹೇಗೆ ಬಳಸುವುದು](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- ಅಜೂರ್ AI ಶೋಧನೆಯೊಂದಿಗೆ RAG + - [Phi-4-mini ಮತ್ತು Phi-4-multimodal(RAG) ಅನ್ನು ಅಜೂರ್ AI ಶೋಧನೆಯೊಂದಿಗೆ ಹೇಗೆ ಬಳಸುವುದು](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud ಲೊಕಲ್ ಹೈಬ್ರಿಡ್ RAG ಸ್ಪಷ್ಟೀಕರಣ SQLite FTS5 ಮತ್ತು phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- ಫೈ ಅಪ್ಲಿಕೇಶನ್ ಅಭಿವೃದ್ಧಿ ನುಡಿಗೊಳ್ಳುವಿಕೆಗಳು +- ಫೈ ಅಪ್ಲಿಕೇಶನ್ ಅಭಿವೃದ್ಧಿ ಉದಾಹರಣೆಗಳು - ಪಠ್ಯ ಮತ್ತು ಚಾಟ್ ಅಪ್ಲಿಕೇಶನ್‌ಗಳು - - Phi-4 ಉದಾಹರಣೆಗಳು - - [📓] [Phi-4-mini ONNX ಮಾದರಿಯಿಂದ ಚಾಟ್](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 ಲೋಕಲ್ ONNX ಮಾದರಿಯಿಂದ ಚಾಟ್ ಡಾಟ್ ನೆಟ್](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [ಸೆಮ್ಯಾಂಟಿಕ್ ಕರ್ಣಲ್ ಬಳಸಿ Phi-4 ONNX ಚಾಟ್ ನ ಅನುಭವ ಡಾಟ್ ನೆಟ್ ಕൺಸೋಲ್ ಅಪ್ಲಿಕೇಶನ್](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 ನುಡಿಗೊಳ್ಳುವಿಕೆಗಳು - - [Phi3, ONNX ರನ್ಟೈಮ್ ವೆಬ್ ಮತ್ತು ವೆಬ್ GPU ಬಳಸಿ ಬ್ರೌಸರ್‌ನಲ್ಲಿ ತಾಲೀಮು ಮಾಡಲಾದ ಸ್ಥಳೀಯ ಚಾಟ್ ಬಾಟ್](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino ಚಾಟ್](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [ಮಲ್ಟಿ ಮಾದರಿ - ಇಂಟರ್ಯಾಕ್ಟಿವ್ ಫೈ-3-ಮಿನಿ ಮತ್ತು ಓಪನ್AI ವಿಸ್ಪರ್](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - ರ್ಯಾಪರ್ ನಿರ್ಮಾಣ ಮತ್ತು Phi-3 ಅನ್ನು MLFlow ಜೊತೆ ಬಳಸುವುದು](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [ಮಾದರಿ ಆಪ್ಟಿಮೈಜೆಶನ್ - ONNX ರನ್‌ಟೈಮ್ ವೆಬ್‌ಗೆ Phi-3-ಮಿನ್ ಮಾದರಿಯನ್ನು Olive ಸಹಾಯದಿಂದ ಹೇಗೆ ಆಪ್ಟಿಮೈಸ್ ಮಾಡುವುದು](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 ಅಪ್ Phi-3 ಮಿನಿ-4k-ಇನ್‌ಸ್ಟ್ರಕ್ಟ್-onnx ಸಹಿತ](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 ಮಲ್ಟಿ ಮಾದರಿ AI ಚಾಲಿತ ನೋಟ್ಗಳ ಅಪ್ ಮಾದರಿ](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [ಕಸ್ಟಮ್ Phi-3 ಮಾದರಿಗಳನ್ನು ಸೂಕ್ಷ್ಮವಾಗಿ ಮುನಿಸಿಕೊಂಡು Prompt flow ಜೊತೆ ಸಂಯೋಜಿಸುವುದು](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Microsoft Foundry ಯಲ್ಲಿ Prompt flow ಜೊತೆ ಕಸ್ಟಮ್ Phi-3 ಮಾದರಿಗಳನ್ನು ಸೂಕ್ಷ್ಮವಾಗಿ ಮುನಿಸಿಕೊಂಡು ಸಂಯೋಜಿಸುವುದು](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Microsoft Foundry ಯಲ್ಲಿ Microsoft's ಜವಾಬ್ದಾರಿಯ AI ತತ್ವಗಳಿಂದ Phi-3 / Phi-3.5 ಸೂಕ್ಷ್ಮಾಯೋಜಿತ ಮಾದರಿಯನ್ನು ಮೌಲ್ಯಮಾಪನ ಮಾಡುವುದು](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-ಮಿನಿ-ಇನ್‌ಸ್ಟ್ರಕ್ಟ್ ಭಾಷಾ ಅನುಮಾನ ಮಾದರಿ (ಚೈನೀಸ್/ಇಂಗ್ಲಿಷ್)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-ಇನ್‌ಸ್ಟ್ರಕ್ಟ್ WebGPU RAG ಚಾಟ್‌ಬಾಟ್](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU ಬಳಸಿ Phi-3.5-ಇನ್‌ಸ್ಟ್ರಕ್ಟ್ ONNX ಜೊತೆ Prompt flow ಪರಿಹಾರವನ್ನು ನಿರ್ಮಿಸುವುದು](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Microsoft Phi-3.5 tflite ಬಳಸಿ Android ಅಪ್ ನಿರ್ಮಾಣ](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Microsoft.ML.OnnxRuntime ಬಳಸಿ ಸ್ಥಳೀಯ ONNX Phi-3 ಮಾದರಿಯನ್ನು ಬಳಸುವ Q&A .NET ಉದಾಹರಣೆ](../../md/04.HOL/dotnet/src/LabsPhi301) - - [ಸಮಾನಾರ್ಥಕ ಕರ್ಣелекೆಯೊಂದಿಗೆ ಮತ್ತು Phi-3 ಸಹಿತ ಕನ್ಸೋಲ್ ಚಾಟ್ .NET ಅಪ್](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI ಇನ್‌ಫರೆನ್ಸ್ SDK ಕೋಡ್ ಆಧಾರಿತ ಮಾದರಿಗಳು - - Phi-4 ಮಾದರಿಗಳು - - [📓] [Phi-4-ಮಲ್ಟಿಮೋಡಲ್ ಬಳಸಿ ಪ್ರಾಜೆಕ್ಟ್ ಕೋಡ್ ಸೃಷ್ಟಿಸುವುದು](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 ಮಾದರಿಗಳು - - [Microsoft Phi-3 ಕುಟುಂಬದೊಂದಿಗೆ ನಿಮ್ಮದೇ Visual Studio Code GitHub Copilot ಚಾಟ್ ನಿರ್ಮಿಸು](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub ಮಾದರಿಗಳೊಂದಿಗೆ Phi-3.5 ಬಳಸಿ ನಿಮ್ಮದೇ Visual Studio Code ಚಾಟ್ ಕೋಪೈಲಟ್ ಏಜೆಂಟ್ ಸೃಷ್ಟಿ](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - ಅಭಿವೃದ್ಧಿಶೀಲ ತರ್ಕ ಮಾದರಿಗಳು - - Phi-4 ಮಾದರಿಗಳು - - [📓] [Phi-4-ಮಿನಿ-ತರ್ಕ ಅಥವಾ Phi-4-ತರ್ಕ ಮಾದರಿಗಳು](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Microsoft Olive ಬಳಸಿ Phi-4-ಮಿನಿ-ತರ್ಕದ ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Apple MLX ಬಳಸಿ Phi-4-ಮಿನಿ-ತರ್ಕದ ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [GitHub ಮಾದರಿಗಳೊಂದಿಗೆ Phi-4-ಮಿನಿ-ತರ್ಕ](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Microsoft Foundry ಮಾದರಿಗಳೊಂದಿಗೆ Phi-4-ಮಿನಿ-ತರ್ಕ](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - ಪ್ರದರ್ಶನಗಳು - - [Phi-4-ಮಿನಿ ಪ್ರದರ್ಶನಗಳು Hugging Face Spaces ನಲ್ಲಿ ಹೋಸ್ಟ್ ಮಾಡಲಾಗಿದೆ](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-ಮಲ್ಟಿಮೋಡಲ್ ಪ್ರದರ್ಶನಗಳು Hugginge Face Spaces ನಲ್ಲಿ ಹೋಸ್ಟ್ ಮಾಡಲಾಗಿದೆ](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Phi-4 ಉದಾಹರಣೆಗಳು + - [📓] [Phi-4-mini ONNX ಮಾದರಿಯೊಂದಿಗೆ ಚಾಟ್](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Phi-4 ಲೊಕಲ್ ONNX ಮಾದರಿಯೊಂದಿಗೆ .NET ಚಾಟ್](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [ಸಿಮೆಂಟಿಕ್ ಕರ್ಣಲ್ ಬಳಸಿ Phi-4 ONNX ಸಹಿತ .NET ಕൺಸೋಲ್ ಅಪ್ಲಿಕೇಶನ್](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - ಫೈ-3 / 3.5 ಮಾದರಿಗಳು + - [ಫೈ3, ONNX ರಂಟೈಮ್ ವೆಬ್ ಮತ್ತು ವೆಬ್‌ಜಿಪಿಯು ಬಳಸಿ ಬ್ರ라우ಸರ್‌ನಲ್ಲಿ ಸ್ಥಳೀಯ ಚಾಟ್‌ಬಾಟ್](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [ಒಪನ್‌ವಿನೋ ಚಾಟ್](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [ಬಹು ಮಾದರಿ - ಪರಸ್ಪರ ಕ್ರಿಯಾತ್ಮಕ ಫೈ-3-ಮಿನಿ ಮತ್ತು ಓಪನ್‌ಎಐ ವಿಶ್ಪರ್](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [ಎಂಎಲ್ಫ್ಲೋ - ರ್ಯಾಪರ್ ನಿರ್ಮಿಸಿ ಮತ್ತು ಫೈ-3 ಅನ್ನು ಎಂಎಲ್ಫ್ಲೋ ಜೊತೆಗೆ ಬಳಸಿ](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [ಮಾದರಿ ಅನ್ವಯಿಕೆ - ಓನ್ಎಎಕ್ಸ್ಎಸ್ ರಂಟೈಮ್ ವೆಬ್ ಗಾಗಿ ಫೈ-3-ಮಿನ್ ಮಾದರಿಯನ್ನು Olive ಬಳಸಿ ಹೇಗೆ ಒಪ್ಟಿಮೈಸ್ ಮಾಡುವುದು](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [ಫೈ-3 ಮಿನಿ-4ಕೆ-ಇನ್ಸ್ಟ್ರಕ್ಟ್-onnx ಜೊತೆ WinUI3 ಅಪ್ಲಿಕೇಶನ್](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 ಬಹು ಮಾದರಿ AI ಚಾಲಿತ ನೋಟುಗಳ ಅಪ್ಲಿಕೇಶನ್ ಮಾದರಿ](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [ಕಸ್ಟಮ್ ಫೈ-3 ಮಾದರಿಗಳನ್ನು ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ಮಾಡಿ ಮತ್ತು ಬನ್ನಿ ಪ್ರಾಂಪ್ಟ್ ಫ್ಲೋೊಂದಿಗೆ ಸಂಯೋಜಿಸಿ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೌಂಡ್ರಿಯಲ್ಲಿ ಪ್ರಾಂಪ್ಟ್ ಫ್ಲೋ ಮೂಲಕ ಕಸ್ಟಮ್ ಫೈ-3 ಮಾದರಿಗಳನ್ನು ಫೈನ್ ಟ್ಯೂನಿಂಗ್ ಮಾಡಿ ಮತ್ತು ಹೊಂದಿಸಿ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೌಂಡ್ರಿಯಲ್ಲಿ ಫೈನ್-ಟ್ಯೂನ್ಡ್ ಫೈ-3 / ಫೈ-3.5 ಮಾದರಿಯನ್ನು ಮೈಕ್ರೋಸಾಫ್ಟ್‌ನ ಜವಾಬ್ದಾರಿಯುತ AI ತತ್ವಗಳಿಗೆ ಕೇಂದ್ರೀಕೃತವಾಗಿ ಮೌಲ್ಯಮಾಪನ ಮಾಡಿ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [ಫೈ-3.5-ಮಿನಿ-ಇನ್ಸ್ಟ್ರಕ್ಟ್ ಭಾಷಾ ಭವಿಷ್ಯವಾಣಿ ಮಾದರಿ (ಚೈನೀಸ್/ಇಂಗ್ಲಿಷ್)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [ಫೈ-3.5-ಇನ್ಸ್ಟ್ರಕ್ಟ್ WebGPU RAG ಚಾಟ್‌ಬಾಟ್](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [ವಿಂಡೋಸ್ GPU ಬಳಸಿ ಫೈ-3.5 ಇನ್ಸ್ಟ್ರಕ್ಟ್ ONNX ಜೊತೆಗೆ ಪ್ರಾಂಪ್ಟ್ ಫ್ಲೋ ಪರಿಹಾರ ಸೃಷ್ಟಿ](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೈ-3.5 ಟಿಎಫ್‌ಲೈಟ್ ಬಳಸಿ ಆಂಡ್ರಾಯ್ಡ್ ಅಪ್ ಸೃಷ್ಟಿ](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [ಪ್ರಶ್ನೋತ್ತರ .NET ಉದಾಹರಣೆ ಸ್ಥಳೀಯ ONNX ಫೈ-3 ಮಾದರಿಯನ್ನು Microsoft.ML.OnnxRuntime ಬಳಸಿ](../../md/04.HOL/dotnet/src/LabsPhi301) + - [ಕಾನ್ಸೋಲ್ ಚಾಟ್ .NET ಅಪ್ ಸೆಮ್ಯಾಂಟಿಕ್ ಕರ್ಣಲ್ ಮತ್ತು ಫೈ-3 ಜೊತೆಗೆ](../../md/04.HOL/dotnet/src/LabsPhi302) + + - ಅಜೂರ್ AI ಇನ್ಫರೆನ್ಸ್ SDK ಕೋಡ್ ಆಧಾರಿತ ಮಾದರಿಗಳು + - ಫೈ-4 ಮಾದರಿಗಳು + - [📓] [ಫೈ-4-ಮಲ್ಟಿಮೋಡಲ್ ಬಳಸಿ ಪ್ರಾಜೆಕ್ಟ್ ಕೋಡ್ ತಯಾರಿಕೆ](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - ಫೈ-3 / 3.5 ಮಾದರಿಗಳು + - [ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೈ-3 ಕುಟುಂಬದೊಂದಿಗೆ ನೀವು ನಿಮ್ಮ ಸ್ವಂತ ವೀಸಿಯಲ್ ಸ್ಟುಡಿಯೋ ಕೋಡ್ GitHub ಕೊಪಿಲಟ್ ಚಾಟ್ ನಿರ್ಮಿಸಿ](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub ಮಾದರಿಗಳೊಂದಿಗೆ ಫೈ-3.5 ಬಳಸಿ ನಿಮ್ಮ ಸ್ವಂತ ವೀಸಿಯಲ್ ಸ್ಟುಡಿಯೋ ಕೋಡ್ ಚಾಟ್ ಕೊಪಿಲಟ್ ಏಜೆಂಟ್ ಸೃಷ್ಟಿ](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - ಮುಂದಿನ ಮಟ್ಟದ ತಾರ್ಕಿಕತೆ ಮಾದರಿಗಳು + - ಫೈ-4 ಮಾದರಿಗಳು + - [📓] [ಫೈ-4-ಮಿನಿ-ತಾರ್ಕಿಕತೆ ಅಥವಾ ಫೈ-4-ತಾರ್ಕಿಕತೆ ಮಾದರಿಗಳು](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [ಮೈಕ್ರೋಸಾಫ್ಟ್ Olive ಬಳಸಿ ಫೈ-4-ಮಿನಿ-ತಾರ್ಕಿಕತೆ ಫೈನ್-ಟ್ಯೂನಿಂಗ್](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [ಆಪಲ್ MLX ಬಳಸಿ ಫೈ-4-ಮಿನಿ-ತಾರ್ಕಿಕತೆಗೆ ಫೈನ್-ಟ್ಯೂನಿಂಗ್](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [GitHub ಮಾದರಿಗಳೊಂದಿಗೆ ಫೈ-4-ಮಿನಿ-ತಾರ್ಕಿಕೆ](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೌಂಡ್ರಿ ಮಾದರಿಗಳೊಂದಿಗೆ ಫೈ-4-ಮಿನಿ-ತಾರ್ಕಿಕೆ](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - ಡೆಮೋಗಳು + - [ಹಗ್ಗಿಂಗ್ ಫೇಸ್ ಸ್ಪೇಸಸ್‌ನಲ್ಲಿ ಫೈ-4-ಮಿನಿ ಡೆಮೋಗಳು](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [ಹಗ್ಗಿಂಗ್ ಫೇಸ್ ಸ್ಪೇಸಸ್‌ನಲ್ಲಿ ಫೈ-4-ಮಲ್ಟಿಮೋಡಲ್ ಡೆಮೋಗಳು](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - ದೃಷ್ಟಿ ಮಾದರಿಗಳು - - Phi-4 ಮಾದರಿಗಳು - - [📓] [Phi-4-ಮಲ್ಟಿಮೋಡಲ್ ಬಳಸಿ ಚಿತ್ರಗಳನ್ನು ಓದಿ ಕೋಡ್ ರಚನೆ](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 ಮಾದರಿಗಳು - - [📓][Phi-3-ದೃಷ್ಟಿ ಚಿತ್ರ ಪಠ್ಯದಿಂದ ಪಠ್ಯಕ್ಕೆ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-ದೃಷ್ಟಿ ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-ದೃಷ್ಟಿ CLIP ಪ್ರತಿರೂಪಣಾ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [ಪ್ರದರ್ಶನ: Phi-3 ರಿಸೈಕ್ಲಿಂಗ್](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-ದೃಷ್ಟಿ - ದೃಷ್ಟಿ ಭಾಷಾ ಸಹಾಯಕ - Phi3-ದೃಷ್ಟಿ ಮತ್ತು OpenVINO ಸಹಿತ](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 ದೃಷ್ಟಿ Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 ದೃಷ್ಟಿ OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 ದೃಷ್ಟಿ ಬಹು-ಫ್ರೆ임್ ಅಥವಾ ಬಹು-ಚಿತ್ರ ಮಾದರಿ](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Microsoft.ML.OnnxRuntime .NET ಬಳಸಿ Phi-3 ದೃಷ್ಟಿ ಸ್ಥಳೀಯ ONNX ಮಾದರಿ](../../md/04.HOL/dotnet/src/LabsPhi303) - - [ಮೆನು ಆಧಾರಿತ Phi-3 ದೃಷ್ಟಿ ಸ್ಥಳೀಯ ONNX ಮಾದರಿ Microsoft.ML.OnnxRuntime .NET ಬಳಸಿ](../../md/04.HOL/dotnet/src/LabsPhi304) - - - ತರ್ಕ-ದೃಷ್ಟಿ ಮಾದರಿಗಳು - - Phi-4-ತರ್ಕ-ದೃಷ್ಟಿ-15B - - [📓] [Phi-4-ತರ್ಕ-ದೃಷ್ಟಿ-15B ಬಳಸಿ ಜ್ಯಾವಾಕಿಂಗ್ ಪತ್ತೆಮಾಡುವುದು](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Phi-4-ತರ್ಕ-ದೃಷ್ಟಿ-15B ಬಳಸಿ ಗಣಿತ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-ತರ್ಕ-ದೃಷ್ಟಿ-15B ಬಳಸಿ UI ಪತ್ತೆಮಾಡುವುದು](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - ಫೈ-4 ಮಾದರಿಗಳು + - [📓] [ಫೈ-4-ಮಲ್ಟಿಮೋಡಲ್ ಬಳಸಿ ಚಿತ್ರಗಳನ್ನು ಓದಿ ಮತ್ತು ಕೋಡ್ ಬಳಸಿ ಸೃಷ್ಟಿ](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - ಫೈ-3 / 3.5 ಮಾದರಿಗಳು + - [📓][ಫೈ-3-ದೃಷ್ಟಿ-ಚಿತ್ರ ಪಠ್ಯದಿಂದ ಪಠ್ಯಕ್ಕೆ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [ಫೈ-3-ದೃಷ್ಟಿ-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][ಫೈ-3-ದೃಷ್ಟಿ CLIP ಎಂಬೆಡ್ಡಿಂಗ್](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [ಡೆಮೋ: ಫೈ-3 ಮರುಬಳಕೆ](https://github.com/jennifermarsman/PhiRecycling/) + - [ಫೈ-3-ದೃಷ್ಟಿ - ದೃಷ್ಟಿ ಭಾಷಾ ಸಹಾಯಕ - ಫೈ3-ದೃಷ್ಟಿ ಮತ್ತು ಓಪನ್‌ವಿನೊ ಸಹಿತ](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [ಫೈ-3 ದೃಷ್ಟಿ ನಿವಿಡಿಯಾ NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [ಫೈ-3 ದೃಷ್ಟಿ ಓಪನ್‌ವಿನೊ](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][ಫೈ-3.5 ದೃಷ್ಟಿ ಬಹು-ಫ್ರೇಮ್ ಅಥವಾ ಬಹು-ಚಿತ್ರ ಮಾದರಿ](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [ಮೈಕ್ರೋಸಾಫ್ಟ್.ML.OnnxRuntime .NET ಬಳಸಿ ಫೈ-3 ದೃಷ್ಟಿ ಸ್ಥಳೀಯ ONNX ಮಾದರಿ](../../md/04.HOL/dotnet/src/LabsPhi303) + - [ಮೆನು ಆಧಾರಿತ ಫೈ-3 ದೃಷ್ಟಿ ಸ್ಥಳೀಯ ONNX ಮಾದರಿ ಮೈಕ್ರೋಸಾಫ್ಟ್.ML.OnnxRuntime .NET ಬಳಸಿ](../../md/04.HOL/dotnet/src/LabsPhi304) + + - ತಾರ್ಕಿಕ-ದೃಷ್ಟಿ ಮಾದರಿಗಳು + - ಫೈ-4-ತಾರ್ಕಿಕ-ದೃಷ್ಟಿ-15B + - [📓] [ಜಯ್‌ವಾಕಿಂಗ್ ಪತ್ತೆಗಾಗಿ ಫೈ-4-ತಾರ್ಕಿಕ-ದೃಷ್ಟಿ-15B ಬಳಕೆ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [ಗಣಿತಕ್ಕಾಗಿ ಫೈ-4-ತಾರ್ಕಿಕ-ದೃಷ್ಠಿ-15B ಬಳಕೆ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [UI ಪತ್ತೆಗಾಗಿ ಫೈ-4-ತಾರ್ಕಿಕ-ದೃಷ್ಟಿ-15B ಬಳಕೆ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - ಗಣಿತ ಮಾದರಿಗಳು - - Phi-4-ಮಿನಿ-ಫ್ಲಾಶ್-ತರ್ಕ-ಇನ್‌ಸ್ಟ್ರಕ್ಟ್ ಮಾದರಿಗಳು [Phi-4-ಮಿನಿ-ಫ್ಲಾಶ್-ತರ್ಕ-ಇನ್‌ಸ್ಟ್ರಕ್ಟ್ ಗಣಿತ ಪ್ರದರ್ಶನ](./md/02.Application/09.Math/MathDemo.ipynb) + - ಫೈ-4-ಮಿನಿ-ಫ್ಲ್ಯಾಶ್-ತಾರ್ಕಿಕತೆ-ಇನ್ಸ್ಟ್ರಕ್ಟ್ ಮಾದರಿಗಳು [ಫೈ-4-ಮಿನಿ-ಫ್ಲ್ಯಾಶ್-ತಾರ್ಕಿಕತೆ-ಇನ್ಸ್ಟ್ರಕ್ಟ್ ಜೊತೆಗೆ ಗಣಿತ ಡೆಮೋ](./md/02.Application/09.Math/MathDemo.ipynb) - - ಧ್ವನಿ ಮಾದರಿಗಳು - - Phi-4 ಮಾದರಿಗಳು - - [📓] [Phi-4-ಮಲ್ಟಿಮೋಡಲ್ ಬಳಸಿ ಧ್ವನಿ ಪಠ್ಯಾಂಶಗಳನ್ನು ಹೊರತೆಗೆಯುವುದು](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-ಮಲ್ಟಿಮೋಡಲ್ ಧ್ವನಿ ಮಾದರಿ](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-ಮಲ್ಟಿಮೋಡಲ್ ವಾಚನ ಅನುವಾದ ಮಾದರಿ](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET ಕನ್ಸೋಲ್ ಅಪ್ಲಿಕೇಶನ್ Phi-4-ಮಲ್ಟಿಮೋಡಲ್ ಧ್ವನಿಯನ್ನು ವಿಶ್ಲೇಷಿಸಿ ಪಠ್ಯಾಂಶ ರಚಿಸುವುದಕ್ಕೆ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - ಆڊيو ಮಾದರಿಗಳು + - ಫೈ-4 ಮಾದರಿಗಳು + - [📓] [ಫೈ-4-ಮಲ್ಟಿಮೋಡಲ್ ಬಳಸಿ ಧ್ವನಿ ಲಿಪ್ಯಂತರಣ ಮಾಡುವುದು](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [ಫೈ-4-ಮಲ್ಟಿಮೋಡಲ್ ಆಡಿಯೋ ಮಾದರಿ](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [ಫೈ-4-ಮಲ್ಟಿಮೋಡಲ್ ಭಾಷಾಂತರ ಧ್ವನಿ ಮಾದರಿ](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET ಕಾನ್ಸೋಲ್ ಅಪ್ಲಿಕೇಶನ್ ಫೈ-4-ಮಲ್ಟಿಮೋಡಲ್ ಆಡಿಯೋ ಬಳಸಿ ಫೈಲ್ ವಿಶ್ಲೇಷಿಸಿ ಲಿಪ್ಯಂತರಣ ಸೃಷ್ಟಿ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE ಮಾದರಿಗಳು - - Phi-3 / 3.5 ಮಾದರಿಗಳು - - [📓] [Phi-3.5 ತಜ್ಞರ ಮಿಶ್ರಣ ಮಾದರಿಗಳು (MoEs) ಸಾಮಾಜಿಕ ಮಾಧ್ಯಮ ಮಾದರಿ](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search ಮತ್ತು LlamaIndex ಜೊತೆ Retrieval-Augmented Generation (RAG) ಪೈಪ್ಲೈನ್ ನಿರ್ಮಿಸಿ](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - ಫೈ-3 / 3.5 ಮಾದರಿಗಳು + - [📓] [ಫೈ-3.5 ಮಿಕ್ಸ್ಚರ್ ಆಫ್ ಎಕ್ಸ್ಪರ್ಟ್ಸ್ ಮಾದರಿಗಳು (MoEs) ಸೊಶಿಯಲ್ ಮೀಡಿಯಾ ಮಾದರಿ](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [ನಿವಿಡಿಯಾ NIM ಫೈ-3 MOE, ಅಜೂರ್ AI سرچ್ ಮತ್ತು ಲ್ಲಾಮಾ ಇಂಡೆಕ್ಸ್ ಬಳಸಿ ರಿಟ್ರಿವಲ್-ಆಗ್ಮೆಂಟಡ್ ಜನರೇಷನ್ (RAG) ಪೈಪ್ಲೈನ್ ನಿರ್ಮಿಸಿ](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - ಫಂಕ್ಷನ್ ಕರೆ ಮಾದರಿಗಳು - - Phi-4 ಮಾದರಿಗಳು 🆕 - - [📓] [Phi-4-ಮಿನಿ ಜೊತೆ ಫಂಕ್ಷನ್ ಕರೆ ಉಪಯೋಗಿಸುವುದು](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Phi-4-ಮಿನಿ ಬಳಸಿ ಬಹು-ಏಜೆಂಟ್‌ಗಳು ಸೃಷ್ಟಿಸಲು ಫಂಕ್ಷನ್ ಕರೆ ಉಪಯೋಗಿಸುವುದು](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Ollama ಜೊತೆಗೆ ಫಂಕ್ಷನ್ ಕರೆ ಉಪಯೋಗಿಸುವುದು](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [ONNX ಜೊತೆಗೆ ಫಂಕ್ಷನ್ ಕರೆ ಉಪಯೋಗಿಸುವುದು](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - ಮಲ್ಟಿಮೋಡಲ್ ಮಿಶ್ರಣ ಮಾದರಿಗಳು - - Phi-4 ಮಾದರಿಗಳು 🆕 - - [📓] [ತಂತ್ರಜ್ಞಾನ ಪತ್ರಕರ್ತನಾಗಿ Phi-4-ಮಲ್ಟಿಮೋಡಲ್ ಬಳಸು](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET ಕನ್ಸೋಲ್ ಅಪ್ Phi-4-ಮಲ್ಟಿಮೋಡಲ್ ಬಳಸಿ ಚಿತ್ರಗಳನ್ನು ವಿಶ್ಲೇಷಿಸಲು](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Phi ಮಾದರಿಗಳ ಸೂಕ್ಷ್ಮ ಗಾಯನ - - [ಸೂಕ್ಷ್ಮ ಗಾಯನ ಹಿನ್ನೆಲೆ ಪ್ರದೇಶಗಳು](./md/03.FineTuning/FineTuning_Scenarios.md) - - [ಸೂಕ್ಷ್ಮ ಗಾಯನ ವಿರುದ್ಧ RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Phi-3 ಅನ್ನು ಕೈಗಾರಿಕಾ ತಜ್ಞ ಮಾಡಿಕೊಳ್ಳಲು ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [VS Code ಗಾಗಿ AI ಉಪಕರಣಗಳೊಂದಿಗೆ Phi-3 ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Azure Machine Learning ಸೇವೆ ಬಳಸಿ Phi-3 ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/Introduce_AzureML.md) - - [Lora ಬಳಸಿ Phi-3 ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora ಬಳಸಿ Phi-3 ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/FineTuning_Qlora.md) - - [Microsoft Foundry ಬಳಸಿ Phi-3 ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Azure ML CLI/SDK ಬಳಸಿ Phi-3 ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Microsoft Olive ಬಳಸಿ ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive Hands-On Lab ಬಳಸಿ ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/olive-lab/readme.md) - - [Weights and Bias ಬಳಸಿ Phi-3-vision ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Apple MLX ಫ್ರೇಮ್ವರ್ಕ್ ಬಳಸಿ Phi-3 ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-vision (ಅಧಿಕೃತ ಬೆಂಬಲ) ಸೂಕ್ಷ್ಮ ಗಾಯನ](./md/03.FineTuning/FineTuning_Vision.md) - - [ಕೈಟೋ AKS , ಅಜೂರ್ ಕಂಟೈನರ್‍ಗಳೊಂದಿಗೆ ಫೈ-3 ಅನ್ನು ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ಮಾಡುವುದು (ಅಧಿಕೃತ ಬೆಂಬಲ)](./md/03.FineTuning/FineTuning_Kaito.md) - - [ಫಿ-3 ಮತ್ತು 3.5 ವೀಕ್ಷಣೆಯನ್ನು ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ಮಾಡುವುದು](https://github.com/2U1/Phi3-Vision-Finetune) - -- ಕೈಯಲ್ಲಿ ಪ್ರಯೋಗಾಲಯ - - [ಅತ್ಯಾಧುನಿಕ ಮಾದರಿಗಳನ್ನು ಅನ್ವೇಷಣೆ ಮಾಡುವುದು: LLMs, SLMs, ಸ್ಥಳೀಯ ಅಭಿವೃದ್ಧಿ ಮತ್ತು ಇನ್ನಷ್ಟು](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP ಶಕ್ತಿ ಅನ್ಲಾಕ್ ಮಾಡುವುದು: ಮೈಕ್ರೋಸಾಫ್ಟ್ ಒಲಿವ್ ಮೂಲಕ ಫೈನ್-ಟ್ಯೂನಿಂಗ್](https://github.com/azure/Ignite_FineTuning_workshop) - -- ಶೈಕ್ಷಣಿಕ ಸಂಶೋಧನಾ ಕೆಳಪುಸ್ತಕಗಳು ಮತ್ತು ಪ್ರಕಟಣೆಗಳು - - [ಪಾಠಪುಸ್ತಕಗಳು ನಿಮಗೆ ಬೇಕಾದ ಎಲ್ಲವನ್ನೂ II: phi-1.5 ತಾಂತ್ರಿಕ ವರದಿ](https://arxiv.org/abs/2309.05463) - - [Phi-3 ತಾಂತ್ರಿಕ ವರದಿ: ನಿಮ್ಮ ಫೋನಿನಲ್ಲಿ ಸ್ಥಳೀಯವಾಗಿ ಅತ್ಯಂತ ಶಕ್ತಿಶಾಲಿ ಭಾಷಾ ಮಾದರಿ](https://arxiv.org/abs/2404.14219) + - ಕಾರ್ಯಾಕಾಲ ಹಕ್ಕು ಮಾದರಿಗಳು + - ಫೈ-4 ಮಾದರಿಗಳು 🆕 + - [📓] [ಫೈ-4-ಮಿನಿ ಜೊತೆ ಕಾರ್ಯಾಕಾಲ ಹಕ್ಕು ಬಳಸುವುದು](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [ಫೈ-4-ಮಿನಿ ಜೊತೆ ಬಹು-ಏಜೆಂಟ್‌ಗಳನ್ನು ಕಾರ್ಯಾಕಾಲ ಹಕ್ಕಾಗಿ ಸೃಷ್ಟಿಸುವುದು](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [ಒಳ್ಳಾಮಾ ಜೊತೆಗೆ ಕಾರ್ಯಾಕಾಲ ಹಕ್ಕು ಬಳಸಿ](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [ONNX ಜೊತೆಗೆ ಕಾರ್ಯಾಕಾಲ ಹಕ್ಕು ಬಳಸುವುದು](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - ಬಹುಮುಖ ಮಿಕ್ಸಿಂಗ್ ಮಾದರಿಗಳು + - ಫೈ-4 ಮಾದರಿಗಳು 🆕 + - [📓] [ಫೈ-4-ಮಲ್ಟಿಮೋಡಲ್ ಬಳಸಿ ತಂತ್ರಜ್ಞಾನ ಪತ್ರಕರ್ತನಾಗಿ](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET ಕಾನ್ಸೋಲ್ ಅಪ್ಲಿಕೇಶನ್ ಫೈ-4-ಮಲ್ಟಿಮೋಡಲ್ ಬಳಸಿ ಚಿತ್ರಗಳನ್ನು ವಿಶ್ಲೇಷಿಸಿ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ಫೈ ಮಾದರಿಗಳು + - [ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ದೃಶ್ಯಗಳು](./md/03.FineTuning/FineTuning_Scenarios.md) + - [ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ಮತ್ತು RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ಫೈ-3 ಅನ್ನು ಉದ್ಯಮ ತಜ್ಞನಾಗಿ ಮಾಡುವುದು](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [VS ಕೋಡ್ కోసం AI ಟೂಲ್‌ಕಿಟ್ ಬಳಸಿ ಫೈ-3 ಫೈನ್-ಟ್ಯೂನಿಂಗ್](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [ಅಜೂರ್ ಮೆಷಿನ್ ಲರ್ನಿಂಗ್ ಸರ್ವಿಸ್ ಬಳಸಿ ಫೈ-3 ಫೈನ್-ಟ್ಯೂನಿಂಗ್](./md/03.FineTuning/Introduce_AzureML.md) + - [ಲೋರಾ ಬಳಸಿ ಫೈ-3 ಫೈನ್-ಟ್ಯೂನಿಂಗ್](./md/03.FineTuning/FineTuning_Lora.md) + - [ಕ್ವಿಕ್ ಲೋರಾ ಬಳಸಿ ಫೈ-3 ಫೈನ್- ಟ್ಯೂನಿಂಗ್](./md/03.FineTuning/FineTuning_Qlora.md) + - [ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೌಂಡ್ರಿ ಬಳಸಿ ಫೈ-3 ಫೈನ್-ಟ್ಯೂನಿಂಗ್](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [ಅಜೂರ್ ML CLI/SDK ಬಳಸಿ ಫೈ-3 ಫೈನ್-ಟ್ಯೂನಿಂಗ್](./md/03.FineTuning/FineTuning_MLSDK.md) + - [ಮೈಕ್ರೋಸಾಫ್ಟ್ Olive ಬಳಸಿ ಫೈನ್-ಟ್ಯೂನಿಂಗ್](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [ಮೈಕ್ರೋಸಾಫ್ಟ್ Olive ಹ್ಯಾಂಡ್ಸ್-ಆನ್ ಲ್ಯಾಬ್ ಜೊತೆ ಫೈನ್-ಟ್ಯೂನಿಂಗ್](./md/03.FineTuning/olive-lab/readme.md) + - [ವೈಟ್ಸ್ಅಂಡ್ ಬಯಾಸ್ ಬಳಸಿ ಫೈ-3-ದೃಷ್ಟಿ ಫೈನ್-ಟ್ಯೂನಿಂಗ್](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Apple MLX ಫ್ರೇಮ್‌ವರ್ಕ್‌ನೊಂದಿಗೆ ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ಫೈ-3](./md/03.FineTuning/FineTuning_MLX.md) + - [ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ಫೈ-3-ವಿಷನ್ (ಅಧಿಕೃತ ಬೆಂಬಲ)](./md/03.FineTuning/FineTuning_Vision.md) + - [ಕೈಟೋ AKS, ಅಜ್ಯೂರ್ ಕಂಟೇನರ್‌ಗಳೊಂದಿಗೆ ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ಫೈ-3 (ಅಧಿಕೃತ ಬೆಂಬಲ)](./md/03.FineTuning/FineTuning_Kaito.md) + - [ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ಫೈ-3 ಮತ್ತು 3.5 ವಿಝನ್](https://github.com/2U1/Phi3-Vision-Finetune) + +- ಹ್ಯಾಂಡ್ಸ್ ಆನ್ ಲ್ಯಾಬ್ + - [ಅತ್ಯಾಧುನಿಕ ಮಾದರಿಗಳನ್ನು ಅನ್ವೇಷಣೆ: LLMಗಳು, SLMಗಳು, ಸ್ಥಳೀಯ ಅಭಿವೃದ್ಧಿ ಮತ್ತು ಇತರೆ](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP ಸಾಮರ್ಥ್ಯ ಅನ್ಲಾಕ್ ಮಾಡುವುದು: Microsoft Olive ಮೂಲಕ ಫೈನ್-ಟ್ಯೂನಿಂಗ್](https://github.com/azure/Ignite_FineTuning_workshop) + +- ಶೈಕ್ಷಣಿಕ ಸಂಶೋಧನಾ ಪೇಪರ್‌ಗಳು ಮತ್ತು ಪ್ರಕಟಣೆಗಳು + - [ತರಗತಿ ಪುಸ್ತಕಗಳು ನಿಮಗೆ ಬೇಕಾದ ಎಲ್ಲವಷ್ಟೆ II: phi-1.5 ತಾಂತ್ರಿಕ ವರದಿ](https://arxiv.org/abs/2309.05463) + - [Phi-3 ತಾಂತ್ರಿಕ ವರದಿ: ನಿಮ್ಮ ಫೋನಿನಲ್ಲಿ ಸ್ಥಳೀಯವಾಗಿ ಅತ್ಯಂತ ಸಾಮರ್ಥ್ಯವಂತ ಭಾಷಾ ಮಾದರಿ](https://arxiv.org/abs/2404.14219) - [Phi-4 ತಾಂತ್ರಿಕ ವರದಿ](https://arxiv.org/abs/2412.08905) - - [Phi-4-ಮಿನಿ ತಾಂತ್ರಿಕ ವರದಿ: ಮಿಶ್ರಣ-ಆಫ್-ಲೋರಾಗಳು ಮೂಲಕ ಸಂಕೋಚನೀಯ ಆದರೆ ಬಲಿಷ್ಠ ಬಹುಮಾಧ್ಯ ಭಾಷಾ ಮಾದರಿಗಳು](https://arxiv.org/abs/2503.01743) - - [ವಾಹನ ಕಾರ್ಯ-ಕಾಲಿಂಗ್ ಗಾಗಿ ಸಣ್ಣ ಭಾಷಾ ಮಾದರಿಗಳನ್ನು tốiಮೀಕರಿಸುವುದು](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) ಬಹು-ಆಯ್ಕೆ ಪ್ರಶ್ನೋತ್ತರಕ್ಕಾಗಿ PHI-3 ಅನ್ನು ಫೈನ್-ಟ್ಯೂನಿಂಗ್ ಮಾಡುವುದು: ವಿಧಾನಶಾಸ್ತ್ರ, ಫಲಿತಾಂಶಗಳು ಮತ್ತು ಸವಾಲುಗಳು](https://arxiv.org/abs/2501.01588) - - [Phi-4 ನಿರ್ಣಾಯಕ ತಾಂತ್ರಿಕ ವರದಿ](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-ಮಿನಿ-ನಿರ್ಣಾಯಕ ತಾಂತ್ರಿಕ ವರದಿ](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Phi-4-ಮಿನಿ ತಾಂತ್ರಿಕ ವರದಿ: ಮಿಶ್ರಣ-ಆಫ್-ಲೋರಾಸ್ ಮೂಲಕ ಸೂಕ್ಷ್ಮ ಆದರೆ ಶಕ್ತಿಶಾಲಿ ಬಹುಮಾಧ್ಯಮ ಭಾಷಾ ಮಾದರಿಗಳು](https://arxiv.org/abs/2503.01743) + - [ವಾಹನದಲ್ಲಿ ಕಾರ್ಯ-ಕರೆ ಕಾಳಜಿಗಾಗಿ ಸಣ್ಣ ಭಾಷಾ ಮಾದರಿಗಳನ್ನು ತಗುಲಿಸುವುದು](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) ಬಹು-ಆಯ್ಕೆ ಪ್ರಶ್ನೆ ಉತ್ತರಿಸುವಿಕೆಯಿಗಾಗಿ PHI-3 ಫೈನ್-ಟ್ಯೂನಿಂಗ್: ವಿಧಾನಶಾಸ್ತ್ರ, ಫಲಿತಾಂಶಗಳು ಮತ್ತು ಸವಾಲುಗಳು](https://arxiv.org/abs/2501.01588) + - [Phi-4-ತರ್ಕಾತ್ಮಕ ತಾಂತ್ರಿಕ ವರದಿ](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-ಮಿನಿ-ತರ್ಕಾತ್ಮಕ ತಾಂತ್ರಿಕ ವರದಿ](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## ಫೈ ಮಾದರಿಗಳನ್ನು ಬಳಸುವುದು -### ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೌಂಡ್ರಿಯಲ್ಲಿನ ಫೈ +### ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೌಂಡ್ರಿಯಲ್ಲಿ ಫೈ -ನೀವು ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೈ ಅನ್ನು ಹೇಗೆ ಬಳಸುವುದು ಮತ್ತು ನಿಮ್ಮ ವಿಭಿನ್ನ ಹಾರ್ಡ್‌ವೇರ್ ಸಾಧನಗಳಲ್ಲಿ E2E ಪರಿಹಾರಗಳನ್ನು ನಿರ್ಮಿಸುವ ಬಗ್ಗೆ ಕಲಿಯಬಹುದು. ಫೈಯನ್ನು ನಿಜವಾಗಿ ಅನುಭವಿಸಲು, ಈ ಮಾದರಿಗಳೊಂದಿಗೆ ಆಟವಾಡಿ ಮತ್ತು ನಿಮ್ಮ ಸಂದರ್ಭಗಳಿಗೆ ಫೈಯನ್ನು ಕಸ್ಟಮೈಸ್ ಮಾಡುವುದು ಪ್ರಾರಂಭಿಸಿ, [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) ಮೂಲಕ ನೀವು ಹೆಚ್ಚಿನದನ್ನು ತಿಳಿಯಬಹುದು, Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) ಅನ್ನು ನೋಡಿರಿ +ನೀವು ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೈ ಅನ್ನು ಹೇಗೆ ಬಳಸುವುದು ಮತ್ತು ನಿಮ್ಮ ವಿಭಿನ್ನ ಹಾರ್ಡ್‌ವೇರ್ ಸಾಧನಗಳಲ್ಲಿ E2E ಪರಿಹಾರಗಳನ್ನು ಹೇಗೆ ನಿರ್ಮಿಸುವುದು ಎಂಬುದನ್ನು ಕಲಿಯಬಹುದು. ಫೈ ಅನ್ನು ಸ್ವತಃ ಅನುಭವಿಸಲು, ಮಾದರಿಗಳ ಜೊತೆ ಆಟವಾಡುವುದರಿಂದ ಮತ್ತು ನಿಮ್ಮ ಸಂದರ್ಭಗಳಿಗೆ ಫೈ ಅನ್ನು ವೈಯಕ್ತೀಕರಿಸುವುದರಿಂದ ಪ್ರಾರಂಭಿಸಬಹುದು. [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) ಬಳಸಿ ನೀವು [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) ಮೂಲಕ ಹೆಚ್ಚಿನ ಮಾಹಿತಿಯನ್ನು ಪಡೆಯಬಹುದು. -**ಆಟದ ಮೈದಾನ** -ಪ್ರತಿ ಮಾದರಿಗೂ ಮಾದರಿಯನ್ನು ಪರೀಕ್ಷಿಸಲು ಒಂದು ನಿಯೋಜಿತ ಆಡಳಿತ ಮೈದಾನವಿದೆ [Azure AI Playground](https://aka.ms/try-phi3). +**ಪ್ಲೇಗ್ರೌಂಡ್** +ಪ್ರತಿ ಮಾದರಿಯು ಮಾದರಿಯನ್ನು ಪರೀಕ್ಷಿಸಲು ಸಮರ್ಪಿತ ಪ್ಲೇಗ್ರೌಂಡ್ನೊಂದಿಗೆ ಬರುತ್ತದೆ [Azure AI Playground](https://aka.ms/try-phi3). ### GitHub ಮಾದರಿಗಳಲ್ಲಿ ಫೈ -ನೀವು ಮೈಕ್ರೋಸಾಫ್ಟ್ ಫೈ ಅನ್ನು ಹೇಗೆ ಬಳಸುವುದು ಮತ್ತು ನಿಮ್ಮ ವಿಭಿನ್ನ ಹಾರ್ಡ್‌ವೇರ್ ಸಾಧನಗಳಲ್ಲಿ E2E ಪರಿಹಾರಗಳನ್ನು ಹೇಗೆ ನಿರ್ಮಿಸುವುದು ಎಂದು ಕಲಿಯಬಹುದು. ಫೈ ಅನ್ನು ನಿಜವಾಗಿಯೂ ಅನುಭವಿಸಲು, ಈ ಮಾದರಿಯೊಂದಿಗೇ ಆಟವಾಡಿ ಮತ್ತು ನಿಮ್ಮ ಸಂದರ್ಭಗಳಿಗೆ ಫೈ ಅನ್ನು ಕಸ್ಟಮೈಸ್ ಮಾಡಿರಿ [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ಬಳಸಿ, Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) ನಲ್ಲಿ ಹೆಚ್ಚಿನ ಮಾಹಿತಿಯನ್ನು ಪಡೆಯಿರಿ +ನೀವು Microsoft Phi ಅನ್ನು ಹೇಗೆ ಬಳಸುವುದು ಮತ್ತು ನಿಮ್ಮ ವಿಭಿನ್ನ ಹಾರ್ಡ್‌ವೇರ್ ಸಾಧನಗಳಲ್ಲಿ E2E ಪರಿಹಾರಗಳನ್ನು ಹೇಗೆ ನಿರ್ಮಿಸುವುದು ಎಂಬುದನ್ನು ಕಲಿಯಬಹುದು. ಫೈ ಅನ್ನು ಅನುಭವಿಸಲು, ಮಾದರಿಯನ್ನು ಬಳಸಿಕೊಂಡು ಹಾಗೂ ನಿಮ್ಮ ಸಂದರ್ಭಗಳಿಗೆ ಫೈ ಅನ್ನು ಕಸ್ಟಮೈಸ್ ಮಾಡಿಕೊಳ್ಳುವುದರಿಂದ ಪ್ರಾರಂಭಿಸಿ [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ಬಳಸಿ ನೀವು [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) ಮೂಲಕ ಹೆಚ್ಚಿನ ವಿಷಯವನ್ನು ತಿಳಿದುಕೊಳ್ಳಬಹುದು. -**ಆಟದ ಮೈದಾನ** -ಪ್ರತಿ ಮಾದರಿಗೂ [ಮಾದರಿಯನ್ನು ಪರೀಕ್ಷಿಸುವ ಆಟದ ಮೈದಾನ](/md/02.QuickStart/GitHubModel_QuickStart.md) ಇದೆ. +**ಪ್ಲೇಗ್ರೌಂಡ್** +ಪ್ರತಿ ಮಾದರಿಯು ಪರೀಕ್ಷಿಸಲು ಸಮರ್ಪಿತ [ಪ್ಲೇಗ್ರೌಂಡ್](/md/02.QuickStart/GitHubModel_QuickStart.md) ಹೊಂದಿದೆ. ### Hugging Face ನಲ್ಲಿ ಫೈ -ನೀವು [Hugging Face](https://huggingface.co/microsoft) ನಲ್ಲಿ ಮಾದರಿಯನ್ನು ಕೂಡಾ ಕಂಡುಹಿಡಿಯಬಹುದು +ನೀವು ಮಾದರಿಯನ್ನು [Hugging Face](https://huggingface.co/microsoft) ನಲ್ಲಿ ಸಹ ಹುಡುಕಬಹುದು -**ಆಟದ ಮೈದಾನ** - [Hugging Chat ಆಟದ ಮೈದಾನ](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**ಪ್ಲೇಗ್ರೌಂಡ್** + [Hugging Chat ಪ್ಲೇಗ್ರೌಂಡ್](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) - ## 🎒 ಇತರೆ ಕೋರ್ಸುಗಳು + ## 🎒 ಇತರೆ ಕೋಸರಗಳು -ನಮ್ಮ ತಂಡ ಇತರೆ ಕೋರ್ಸುಗಳನ್ನು ಉತ್ಪಾದಿಸುತ್ತದೆ! ಪರಿಶೀಲಿಸಿ: +ನಮ್ಮ ತಂಡ ಇತರೆ ಕೋಸರ್ಗಳನ್ನು ಉತ್ಪಾದಿಸುತ್ತದೆ! ನೋಡಿ: ### LangChain -[![ಆರಂಭಿಕರಿಗಾಗಿ LangChain4j](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![ಆರಂಭಿಕರಿಗಾಗಿ LangChain.js](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![ಆರಂಭಿಕರಿಗಾಗಿ LangChain](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain for Beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / ಏಜೆಂಟ್ಗಳು -[![ಆರಂಭಿಕರಿಗಾಗಿ AZD](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![ಆರಂಭಿಕರಿಗಾಗಿ ಎಡ್ಜ್ AI](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![ಆರಂಭಿಕರಿಗಾಗಿ MCP](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![ಆರಂಭಿಕರಿಗಾಗಿ AI ಏಜೆಂಟ್‌ಗಳು](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### ಅಜ್ಯೂರ್ / ಎಡ್ಜ್ / MCP / ಏಜೆಂಟ್ಸ್ +[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### ಸೃಜನಶೀಲ AI ಸರಣಿ -[![ಆರಂಭಿಕರಿಗಾಗಿ ಸೃಜನಶೀಲ AI](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![ಸೃಜನಶೀಲ AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![ಸೃಜನಶೀಲ AI (ಜಾವಾ)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![ಸೃಜನಶೀಲ AI (ಜಾವಾಸ್ಕೋರಿಪ್ಟ್)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### ಜನರೇಟಿವ್ AI ಸರಣಿ +[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- -### ಮೂಲ ಕಲಿಕೆ -[![ಆರಂಭಿಕರಿಗಾಗಿ ಎಂಎಲ್](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![ಆರಂಭಿಕರಿಗಾಗಿ ಡೇಟಾ ಸೈನ್ಸ್](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![ಆರಂಭಿಕರಿಗಾಗಿ AI](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![ಸಿಬ್ಬಂದಿ ಸುರಕ್ಷತೆಗಾಗಿ ಆರಂಭಿಕರು](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![ವೆಬ್ ಡೆವ್ ಆರಂಭಿಕರಿಗಾಗಿ](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![ಐಒಟಿ ಆರಂಭಿಕರಿಗಾಗಿ](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR ಅಭಿವೃದ್ಧಿ ಆರಂಭಿಕರಿಗಾಗಿ](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +### ಮೂಲ ಅಧ್ಯಯನ +[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![ದತ್ತಾಂಶ ವಿಜ್ಞಾನ ಆರಂಭಿಕರಿಗೆ](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![ಸೈಬರ್ ಸುರಕ್ಷತೆ ಆರಂಭಿಕರಿಗೆ](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![ವೆಬ್ ಡೆವ್ ಆರಂಭಿಕರಿಗೆ](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR ಡೆವಲಪ್‌ಮೆಂಟ್ ಆರಂಭಿಕರಿಗೆ](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### ಸಹಾಯಕ ಸರಣಿ -[![AI ಪೇಯರ್ಡ್ ಪ್ರೋಗ್ರಾಮಿಂಗ್‌ಗಾಗಿ ಸಹಾಯಕ](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![C#/.NET ಗಾಗಿ ಸಹಾಯಕ](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![ಸಹಾಯಕ ಸಾಹಸ](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +### ಕೋಪೈಲಟ್ ಸರಣಿ +[![AI ಜೋಡಿಯ ಕಾರ್ಯಕ್ರಮಕ್ಕಾಗಿ ಕೋಪೈಲಟ್](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![C#/.NETಗಾಗಿ ಕೋಪೈಲಟ್](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![ಕೋಪೈಲಟ್ ಸಾಹಸ](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## ಜವಾಬ್ದಾರಿಯುತ AI +## ಜವಾಬ್ದಾರಿಯುತ AI -ಮೈಕ್ರೋಸಾಫ್ಟ್ ನಮ್ಮ ಗ್ರಾಹಕರು AI ಉತ್ಪನ್ನಗಳನ್ನು ಜವಾಬ್ದಾರಿಯಿಂದ ಬಳಸಲು ಸಹಾಯ ಮಾಡುವ, ನಮ್ಮ ಅಧ್ಯಯನಗಳನ್ನು ಹಂಚಿಕೊಳ್ಳುವ, ಮತ್ತು ಪಾರದರ್ಶಕತೆ ಟಿಪ್ಪಣಿಗಳು ಮತ್ತು ಪರಿಣಾಮ ಮೌಲ್ಯಮಾಪನ ಕುರಿತಂತೆ ನಂಬಿಕೆಯ ಆಧಾರಿತ ಕೌಟುಂಬಿಕತೆಯನ್ನು ನಿರ್ಮಿಸಲು ಬದ್ಧವಾಗಿದೆ. ಇವುಗಳಲ್ಲಿ ಅನೇಕ ಸಂಪನ್ಮೂಲಗಳನ್ನು [https://aka.ms/RAI](https://aka.ms/RAI) ನಲ್ಲಿ ಕಾಣಬಹುದು. -ಮೈಕ್ರೋಸಾಫ್ಟ್ ನ ಜವಾಬ್ದಾರಿಯುತ AI ಗೆ ನಿಲುವು ನ್ಯಾಯತೆ, ವಿಶ್ವಾಸಾರ್ಹತೆ ಮತ್ತು ಸುರಕ್ಷತೆ, ಗೌಪ್ಯತೆ ಮತ್ತು ಭದ್ರತೆ, ಒಳಗೊಂಡು, ಪಾರದರ್ಶಕತೆ ಮತ್ತು ಹೊಣೆಗಾರಿಕೆ ಎಂಬ AI ಸಿದ್ಧಾಂತಗಳ ಮೇಲೆ ಆಧಾರಿತವಾಗಿದೆ. +ಮೈಕ್ರೋಸಾಫ್ಟ್ ನಮ್ಮ ಗ್ರಾಹಕರಿಗೆ ನಮ್ಮ AI ಉತ್ಪನ್ನಗಳನ್ನು ಜವಾಬ್ದಾರಿತನದಿಂದ ಬಳಸಲು, ನಮ್ಮ ಅನುಭವಗಳನ್ನು ಹಂಚಿಕೊಳ್ಳಲು ಮತ್ತು Transparency Notes ಮತ್ತು ಪ್ರಭಾವ ಮೌಲ್ಯಮಾಪನಗಳಂತಹ ಉಪಕರಣಗಳ ಮೂಲಕ ನಂಬಿಕೆ ಆಧಾರಿತ ಸಹಭಾಗಿತ್ವಗಳನ್ನು ನಿರ್ಮಿಸಲು ಬದ್ಧವಾಗಿದೆ. ಈ ಸಂಪನ್ಮೂಲಗಳ ಬಹುಮತವನ್ನು [https://aka.ms/RAI](https://aka.ms/RAI) ನಲ್ಲಿ ಕಂಡುಹಿಡಿಯಬಹುದು. +ಮೈಕ್ರೋಸಾಫ್ಟ್ ಜವಾಬ್ದಾರಿಯುತ AI ಗೆ ಹೊಂದಿರುವ அணಿಮೆದುಗಳು ನ್ಯಾಯತೀಯತೆ, ವಿಶ್ವಾಸಾರ್ಹತೆ ಮತ್ತು ಭದ್ರತೆ, ಗೌಪ್ಯತೆ ಮತ್ತು ಭದ್ರತೆ, ಸಂಗ್ರಹಣೆಯು, ಪಾರದರ್ಶಕತೆ ಮತ್ತು જવાબ್ದಾರಿತನದ ನಮ್ಮ AI ತತ್ವಗಳ ಮೇಲೆ ಆಧಾರಿತವಾಗಿದೆ. -ಈ ಉದಾಹರಣೆಯಲ್ಲಿ ಬಳಸಿದ ದೊಡ್ಡ-ಪರಿಮಾಣ ಪ್ರಕೃತಭಾಷೆ, ಚಿತ್ರ ಮತ್ತು ಧ್ವನಿ ಮಾದರಿಗಳು ಅನ್ಯಾಯಕರ, ವಿಶ್ವಾಸಾರ್ಹವನ್ನಲ್ಲದ, ಅಥವಾ ಅಸಹ್ಯವಾಗಿರಬಹುದಾದ ರೀತಿಯಲ್ಲಿ ವರ್ತಿಸಬಹುದು, ಇದು ಹಾನಿಗಳನ್ನುಂಟುಮಾಡಬಹುದು. ದಯವಿಟ್ಟು ಅಪಾಯಗಳು ಮತ್ತುಗಳ ಉಲ್ಲೇಖಗಳಿಗೆ [ಅಜೂರ್ ಓಪನ್AI ಸೇವೆ ಪಾರದರ್ಶಕತೆ ಟಿಪ್ಪಣಿ](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ಅನ್ನು ನೋಡಿ. -ಈ ಅಪಾಯಗಳನ್ನು ಕಡಿಮೆ ಮಾಡುವ ಶಿಫಾರಸ್ಸಾದ ಮೂಲಕ ನಿಮ್ಮ ವಾಸ್ತುಶಿಲ್ಪದಲ್ಲಿ ಹಾನಿಕರವಾದ ವರ್ತನೆಗಳನ್ನು ಪತ್ತೆಹಚ್ಚುವುದಕ್ಕೆ ಮತ್ತು ತಡೆಯುವುದಕ್ಕೆ ಸೇಜಾನಿ ವ್ಯವಸ್ಥೆಯನ್ನು ಸೇರಿಸುವುದು. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ಸ್ವತಂತ್ರ ರಕ್ಷಣಾ ಪದರವನ್ನು ಒದಗಿಸುತ್ತದೆ, ಇದು ಅಪ್ಲಿಕೇಶನ್ಗಳು ಮತ್ತು ಸೇವೆಗಳಲ್ಲಿನ ಹಾನಿಕರ ಬಳಕೆದಾರ-ಉತ್ಪಾದಿತ ಮತ್ತು AI-ಉತ್ಪಾದಿತ ವಿಷಯವನ್ನು ಪತ್ತೆಹಚುತ್ತದೆ. Azure AI Content Safety ನಲ್ಲಿ ಪಠ್ಯ ಮತ್ತು ಚಿತ್ರ APIಗಳು ಸೇರಿವೆ, ಅವು ಹಾನಿಕರ ವಿಷಯವನ್ನು ಪತ್ತೆಹಚ್ಚಲು ಸಹಾಯ ಮಾಡುತ್ತವೆ. Microsoft Foundryಯಲ್ಲಿ, Content Safety ಸೇವೆಯು ವಿಭಿನ್ನ ಪದ್ಧತಿಗಳಲ್ಲಿನ ಹಾನಿಕರ ವಿಷಯಗಳನ್ನು ಪತ್ತೆಹಚ್ಚಲು ಉದಾಹರಣೆ ಕೋಡ್ ಅನ್ನು ವೀಕ್ಷಿಸಲು, ಅನ್ವೇಷಿಸಲು ಮತ್ತು ಪ್ರಯತ್ನಿಸಲು ಅವಕಾಶ ನೀಡುತ್ತದೆ. ಕೆಳಗಿನ [quickstart ದಸ್ತಾವೇಜು](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ಸೇವೆಗೆ ವಿನಂತಿಗಳನ್ನು ಮಾಡಿ ನಿಮ್ಮನ್ನು ಮಾರ್ಗದರ್ಶಿಸುತ್ತದೆ. +ಉದಾಹರಣೆಯಾಗಿ ಬಳಸಲಾದ ದೊಡ್ಡ ಪ್ರಮಾಣದ ನೈಸರ್ಗಿಕ ಭಾಷೆ, ಚಿತ್ರ ಮತ್ತು ಭಾಷಣ ಮಾದರಿಗಳು - ಅನ್ಯಾಯಕರ, ಅಸ್ಥಿರ ಅಥವಾ ಅವಮಾನಕಾರಿ ರೀತಿಯಲ್ಲಿ ನಡಹೊಡಲು ಸಾಧ್ಯವಿದೆ, ಇದರಿಂದ ಹಾನಿ ಉಂಟಾಗಬಹುದು. ಅಪಾಯಗಳು ಮತ್ತು ಹಿರಿತನಗಳ ಕುರಿತು ತಿಳಿಯಲು ದಯವಿಟ್ಟು [ಅಜ್ಯೂರ್ OpenAI ಸೇವೆಯ Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ಅನ್ನು ಪರಿಶೀಲಿಸಿ. -ಮತ್ತೊಂದು ಪರಿಗಣಿಸುವ ಅಂಶವೆಂದರೆ ಒಟ್ಟಾರೆ ಅಪ್ಲಿಕೇಶನ್ ಕಾರ್ಯಕ್ಷಮತೆ. ಬಹು-ಮಾರ್ಗ ಮತ್ತು ಬಹು-ಮಾದರಿ ಅಪ್ಲಿಕೇಶನ್ಗಳಲ್ಲಿ, ಕಾರ್ಯಕ್ಷಮತೆ ಎಂದರೆ ನೀವು ಮತ್ತು ನಿಮ್ಮ ಬಳಕೆದಾರರು ನಿರೀಕ್ಷಿಸುವಂತೆ ವ್ಯವಸ್ಥೆ ಕಾರ್ಯನಿರ್ವಹಿಸುತ್ತದೆ, ಹಾನಿಕರವಾದ ಫಲಿತಾಂಶಗಳನ್ನು ಸೃಷ್ಟಿಸದೆ. ನಿಮ್ಮ ಒಟ್ಟು ಅಪ್ಲಿಕೇಶನಿನ ಕಾರ್ಯಕ್ಷಮತೆಯನ್ನು ಅಂದಾಜಿಸಲು [ಕಾರ್ಯಕ್ಷಮತೆ ಮತ್ತು ಗುಣಮಟ್ಟ ಮತ್ತು ಅಪಾಯ ಮತ್ತು ಸುರಕ್ಷತೆ ಮೌಲ್ಯಮಾಪಕರು](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) ಉಪಯೋಗಿಸಬಹುದು. ನೀವು [ಕಸ್ಟಮ್ ಮೌಲ್ಯಮಾಪಕರು](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) ಕೂಡ ರಚಿಸಿ ಅಂದಾಜಿಸಬಹುದು. -ನೀವು ನಿಮ್ಮ ಅಭಿವೃದ್ಧಿ ವಾತಾವಣದಲ್ಲಿ ನಿಮ್ಮ AI ಅಪ್ಲಿಕೇಶನನ್ನು [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ಬಳಸಿ ಮೌಲ್ಯಮಾಪನ ಮಾಡಬಹುದು. ಪರೀಕ್ಷಾ ಡೇಟಾ ಸೆಟ್ ಅಥವಾ ಗುರಿಯನ್ನು ನೀಡಿದರೆ, ನಿಮ್ಮ ಜನರೇಟಿವ್ AI ಅಪ್ಲಿಕೇಶನಿನ ಉತ್ಪತ್ತಿಗಳನ್ನು ನಿಮ್ಮ ಆಯ್ಕೆಯಲ್ಲಿರುವ ಒಳನಿರ್ಮಿತ ಮೌಲ್ಯಮಾಪಕರು ಅಥವಾ ಕಸ್ಟಮ್ ಮೌಲ್ಯಮಾಪಕರೊಂದಿಗೆ ಗಣಿತಾತ್ಮಕವಾಗಿ ಅಂದಾಜಿಸಲಾಗುತ್ತದೆ. ನಿಮ್ಮ ವ್ಯವಸ್ಥೆಯನ್ನು ಅಂದಾಜಿಸಲು ಮತ್ತು ಬಳಸಲು azure ai evaluation sdk ಕುರಿತು ಪ್ರಾರಂಭಿಸಲು, ನೀವು [quickstart ಮಾರ್ಗದರ್ಶಿಯನ್ನು](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) ಅನುಸರಿಸಬಹುದು. ಒಂದು ಮೌಲ್ಯಮಾಪನ ಓಟವನ್ನು ನಿಭಾಯಿಸಿದ ಮೇಲೆ, ನೀವು [Microsoft Foundryನಲ್ಲಿ ಫಲಿತಾಂಶಗಳನ್ನು ದೃಶ್ಯೀಕರಿಸಬಹುದು](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +ಈ ಅಪಾಯಗಳನ್ನು ತಗ್ಗಿಸಲು ಶಿಫಾರಸು ಮಾಡಲಾದ ದಾರಿ ಇದು ನಿಮ್ಮ ವಾಸ್ತುಶಿಲ್ಪದಲ್ಲಿ ಸುರಕ್ಷತಾ ವ್ಯವಸ್ಥೆಯನ್ನು ಸೇರಿಸಬೇಕು, ಇದು ಹಾನಿಕಾರಕ ವರ್ತನೆಯನ್ನು ಪತ್ತೆಮಾಡಿ ತಡೆಯಬಹುದು. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ಸ್ವತಂತ್ರ ರಕ್ಷಣಾ ಪದರವನ್ನು ಒದಗಿಸುತ್ತದೆ, ಇದು ಅಪ್ಲಿಕೇಷನ್‌ಗಳು ಮತ್ತು ಸೇವೆಗಳಲ್ಲಿ ಹಾನಿಕಾರಕ ಬಳಕೆದಾರ-ಜನಿತ ಮತ್ತು AI-ಜನಿತ ವಿಷಯವನ್ನು ಪತ್ತೆಮಾಡಲು ಸಾಧ್ಯವಾಗುತ್ತದೆ. Azure AI Content Safety ಟೈಪಿಂಗ್ ಮತ್ತು ಚಿತ್ರ APIಗಳನ್ನು ಒಳಗೊಂಡಿದ್ದು, ಹಾನಿಕಾರಕ ವಸ್ತುಗಳನ್ನು ಪತ್ತೆಮಾಡಲು ಅವಕಾಶ ನೀಡುತ್ತದೆ. Microsoft Foundry-ನೊಳಗೆ, Content Safety ಸೇವೆ ನಿಮಗೆ ವಿಭಿನ್ನ ಮಾದರಿಗಳಲ್ಲಿ ಹಾನಿಕಾರಕ ವಿಷಯವನ್ನು ಪತ್ತೆಮಾಡಲು ಭಾರತಿಸಲು, ಪರಿಶೀಲಿಸಲು ಮತ್ತು ಉದಾಹರಣಾ ಕೋಡ್ ಅನ್ನು ಪ್ರಯತ್ನಿಸಲು ಅವಕಾಶ ನೀಡುತ್ತದೆ. ಕೆಳಗಿನ [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ನಿಮಗೆ ಸೇವೆಗೆ ವಿನಂತಿಗಳನ್ನು ಮಾಡಲು ಮಾರ್ಗನಿರ್ದೇಶನ ನೀಡುತ್ತದೆ. -## ಟ್ರೇಡ್ಮಾರ್ಕ್ಗಳು +ಪರಿಗಣಿಸಬೇಕಾದ ಮತ್ತೊಂದು ಅಂಶವೆಂದರೆ ಒಟ್ಟು ಅಪ್ಲಿಕೇಷನ್ ಕಾರ್ಯಕ್ಷಮತೆ. ಬಹು-ಮಾದರಿ ಮತ್ತು ಬಹು-ಮಾಡೆಲ್ ಅಪ್ಲಿಕೇಶನ್ಗಳಲ್ಲಿ, ಕಾರ್ಯಕ್ಷಮತೆ ಅಂದ್ರೆ ನೀವು ಮತ್ತು ನಿಮ್ಮ ಬಳಕೆದಾರರು ನಿರೀಕ್ಷಿಸುವಂತೆ ವ್ಯವಸ್ಥೆ ಕಾರ್ಯನಿರ್ವಹಿಸುತ್ತದೆ, ಹಾನಿಕಾರಕ ಫಲಿತಾಂಶಗಳನ್ನು ಉತ್ಪನ್ನಿಸದಿರುವುದಾಗಿ ಅರ್ಥಮಾಡಿಕೊಳ್ಳಲಾಗುತ್ತದೆ. ನಿಮ್ಮ ಒಟ್ಟು ಅಪ್ಲಿಕೇಶನ್ ನ ಕಾರ್ಯಕ್ಷಮತೆಯನ್ನು [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) ಬಳಸಿ ಮೌಲ್ಯಮಾಪನ ಮಾಡುವುದು ಮುಖ್ಯ. ನೀವು ಹಾಗು[custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) ರಚಿಸಿ ಮೌಲ್ಯಮಾಪನ ಮಾಡಬಲ್ಲಿರಿ. -ಈ ಯೋಜನೆಯಲ್ಲಿ ಯೋಜನೆಗಳು, ಉತ್ಪನ್ನಗಳು ಅಥವಾ ಸೇವೆಗಳು ಸಂಬಂಧಿಸಿದ ಟ್ರೇಡ್ಮಾರ್ಕ್‌ಗಳು ಅಥವಾ ಲೋಗೋಗಳನ್ನು ಹೊಂದಿರಬಹುದು. Microsoft ಟ್ರೇಡ್ಮಾರ್ಕ್‌ಗಳು ಅಥವಾ ಲೋಗೋಗಳ ಅನುಮೋದಿತ ಬಳಕೆ [Microsoftನ ಟ್ರೇಡ್ಮಾರ್ಕ್ & ಬ್ರಾಂಡ್ ಮಾರ್ಗಸೂಚಿಗಳ](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) ಅಧೀನದಲ್ಲಿದ್ದರೆ ಮಾತ್ರ ಸಾಧ್ಯವಾಗುತ್ತದೆ. -ಈ ಯೋಜನೆಯ ತಿದ್ದುಪಡಿ ಮಾಡಿದ ಆವೃತ್ತಿಗಳಲ್ಲಿ Microsoft ಟ್ರೇಡ್ಮಾರ್ಕ್ ಅಥವಾ ಲೋಗೋಗಳ ಬಳಕೆ Microsoft ಪ್ರಾಯೋಜಿತತ್ವವನ್ನು ಗೊಂದಲಕ್ಕೆ ಅಥವಾ ಸೂಚನೆಗೆ ಕಾರಣವಾಗಬಾರದು. ತೃತೀಯಪಕ್ಷದ ಟ್ರೇಡ್ಮಾರ್ಕ್‌ಗಳು ಅಥವಾ ಲೋಗೋಗಳ ಯಾವುದೇ ಬಳಕೆ ಆ ತೃತೀಯಪಕ್ಷದ ನೀತಿಗಳಿಗೆ ಒಳಪಟ್ಟಿರುತ್ತದೆ. +ನೀವು ನಿಮ್ಮ ಅಭಿವೃದ್ಧಿ ಪರಿಸರದಲ್ಲಿ [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ಬಳಸಿ ನಿಮ್ಮ AI ಅಪ್ಲಿಕೇಶನ್ ಅನ್ನು ಮೌಲ್ಯಮಾಪನ ಮಾಡಬಹುದು. ಪರೀಕ್ಷಾ ಡೇಟಾಸೆಟ್ ಅಥವಾ ಗುರಿ ಇದ್ದಾಗ, ನಿಮ್ಮ ಜನರೇಟಿವ್ AI ಅಪ್ಲಿಕೇಶನ್ ಜನರೇಷನ್ಗಳನ್ನು ನಿರ್ಮಿತ ಮೌಲ್ಯಮಾಪಕಗಳು ಅಥವಾ ನಿಮ್ಮ ಆಯ್ಕೆಯ कಸ್ಟಮ್ ಮೌಲ್ಯಮಾಪಕಗಳೊಂದಿಗೆ ಪ್ರಮಾಣಬದ್ಧವಾಗಿ ಅಳೆಯಲಾಗುತ್ತದೆ. ನಿಮ್ಮ ವ್ಯವಸ್ಥೆಯನ್ನು ಮೌಲ್ಯಮಾಪನ ಮಾಡಲು azure ai evaluation sdk ಬಳಸಿ ಪ್ರಾರಂಭಿಸಲು, ನೀವು [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) ಅನುಸರಿಸಬಹುದು. ಒಂದು ಮೌಲ್ಯಮಾಪನ ರನ್ ಅನ್ನು ಕಾರ್ಯಗತಗೊಳಿಸಿದ ನಂತರ, ನೀವು [Microsoft Foundry ನಲ್ಲಿ ಫಲಿತಾಂಶಗಳನ್ನು ದೃಶ್ಯೀಕರಿಸಬಹುದು](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). + +## ಟ್ರೇಡ್ಮಾರ್ಕ್‌ಗಳು + +ಈ ಯೋಜನೆಯು ಯೋಜನೆಗಳು, ಉತ್ಪನ್ನಗಳು ಅಥವಾ ಸೇವೆಗಳ ಟ್ರೇಡ್ಮಾರ್ಕ್‌ಗಳು ಅಥವಾ ಲೋಗೋಗಳನ್ನು ಹೊಂದಿರಬಹುದು. Microsoft ಟ್ರೇಡ್ಮಾರ್ಕ್‌ಗಳು ಅಥವಾ ಲೋಗೋಗಳ ಹಕ್ಕುಪತ್ರಿತ ಬಳಕೆ [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) ಅನುಸರಿಸಬೇಕು ಮತ್ತು ಅದರ ಮೇಲೆ ಅವಲಂಬಿತವಾಗಿದೆ. +ಈ ಯೋಜನೆಯ ಪರಿಷ್ಕೃತ ಆವೃತ್ತಿಗಳಲ್ಲಿ Microsoft ಟ್ರೇಡ್ಮಾರ್ಕ್‌ಗಳು ಅಥವಾ ಲೋಗೋಗಳ ಬಳಕೆ Microsoft ಸ್ಪಾನ್ಸರ್‌ಶಿಪ್ ಅನ್ನು ಭ್ರಮೆ ಮೂಡಿಸುವುದು ಅಥವಾ ಸೂಚಿಸುವುದನ್ನು ತಪ್ಪಿಸಬೇಕು. ಮೂರನೇ ಪಕ್ಷದ ಟ್ರೇಡ್ಮಾರ್ಕ್‌ಗಳು ಅಥವಾ ಲೋಗೋಗಳ ಯಾವುದೇ ಬಳಕೆ ಆ ಮೂರನೇ ಪಕ್ಷದ ನীতಿಗಳಿಗೆ ಒಳಪಟ್ಟಿರುತ್ತದೆ. ## ಸಹಾಯ ಪಡೆಯುವುದು -ನೀವು ಅಡಚಣೆಯೊಳಗಾದರೆ ಅಥವಾ AI ಅಪ್ಲಿಕೇಶನ್ಗಳ ನಿರ್ಮಾಣ ಕುರಿತು ಯಾವುದೇ ಪ್ರಶ್ನೆಗಳಿದ್ದರೆ, ಸೇರಿ: +ನೀವು ಸ್ಥಗಿತಗೊಳ್ಳಿದರೆ ಅಥವಾ AI ಅಪ್ಲಿಕೇಶನ್ಗಳ ನಿರ್ಮಾಣ ಕುರಿತು ಯಾವುದೇ ಪ್ರಶ್ನೆಗಳು ಇದ್ದರೆ, ಈ ಮೂಲಕ ಸೇರಿ: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -ನೀವು ಉತ್ಪನ್ನ ಪ್ರತಿಕ್ರಿಯೆ ಅಥವಾ ದೋಷಗಳನ್ನು ಕಂಡುಹಿಡಿದರೆ ಇಲ್ಲಿ ಭೇಟಿ ನೀಡಿ: +ನೀವು ಉತ್ಪನ್ನ ಪ್ರತಿಕ್ರಿಯೆ ಅಥವಾ ದೋಷಗಳನ್ನು ಹೊಂದಿದ್ದರೆ, ಈ ಮೂಲಕ ಭೇಟಿ ನೀಡಿ: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**ಅಂಕಣ ನಿರಾಕರಣೆ**: -ಈ ದಾಖಲೆನ್ನು AI ಅನುವಾದ ಸೇವೆ [Co-op Translator](https://github.com/Azure/co-op-translator) ಬಳಸಿ ಅನುವಾದಿಸಲಾಗಿದೆ. ನಾವು ನಿಖರತೆಗಾಗಿ ಪ್ರಯತ್ನಿಸುತ್ತೇವೆ, ಆದಾಗ್ಯೂ ಸ್ವಯಂಚಾಲಿತ ಅನುವಾದಗಳಲ್ಲಿ ತಪ್ಪುಗಳು ಅಥವಾ ಅಸತ್ಯತೆಗಳು ಇರಬಹುದೆಂದು ದಯವಿಟ್ಟು ಗಮನಿಸಿ. ಮೂಲ ಭಾಷೆಯಲ್ಲಿ ಇರುವ ಮೂಲ ದಾಖಲೆ ಅಧಿಕಾರ ಸಂಪನ್ನವಾದ ಮೂಲವಾಗಿರಬೇಕು. ಗಂಭೀರ ಮಾಹಿತಿಗಾಗಿ, ವೃತ್ತಿಪರ ಮಾನವ ಅನುವಾದವನ್ನು ಶಿಫಾರಸು ಮಾಡಲಾಗುತ್ತದೆ. ಈ ಅನುವಾದ ಬಳಕೆಯಿಂದ ಉಂಟಾಗುವ ಯಾವುದೇ ಊಹಾಪೋಹಗಳು ಅಥವಾ ತಪ್ಪುವ್ಯಾಖ್ಯಾನಗಳಿಗೆ ನಾವು ಜವಾಬ್ದಾರಿಯಲ್ಲ. +**ಅಸ್ವೀಕಾರ**: +ಈ ದಸ್ತಾವೇಜು AI ಅನುವಾದ ಸೇವೆ [Co-op Translator](https://github.com/Azure/co-op-translator) ಬಳಸಿ ಅನುವಾದಿಸಲಾಗಿದೆ. ನಾವು ನಿಖರತೆಯನ್ನು ಸಾಧಿಸಲು ಪ್ರಯತ್ನಿಸುತ್ತಿದ್ದರೂ, ದಯವಿಟ್ಟು ಗಮನಿಸಿ, ಸ್ವಯಂಚಾಲಿತ ಅನುವಾದಗಳಲ್ಲಿ ದೋಷಗಳು ಅಥವಾ ಅಸಡ್ಡೆಗಳು ಇರಬಹುದು. ಮೂಲ ಭಾಷೆಯಲ್ಲಿರುವ ಮೂಲ ದಸ್ತಾವೇಜು ಪ್ರಾಮಾಣಿಕ ಮೂಲವೆಂದು ಪರಿಗಣಿಸಬೇಕು. ಪ್ರಮುಖ ಮಾಹಿತಿಗಾಗಿ, ವೃತ್ತಿಪರ ಮಾನವ ಅನುವಾದವನ್ನು ಶಿಫಾರಸು ಮಾಡಲಾಗುತ್ತದೆ. ಈ ಅನುವಾದವನ್ನು ಬಳಸುವ ಮೂಲಕ ಉಂಟಾಗುವ ಯಾವುದೇ ತಪ್ಪು ಅರ್ಥಗಳ ಅಥವಾ ತಪ್ಪು ವ್ಯಾಖ್ಯಾನಗಳ ಬಗ್ಗೆ ನಾವು ಹೊಣೆಗಾರರಲ್ಲ. \ No newline at end of file diff --git a/translations/kn/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/kn/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..e307e47b3 --- /dev/null +++ b/translations/kn/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 ಮೈಕ್ರೋಸಾಫ್ಟ್ phi-4-mini ಮತ್ತು SQLite FTS5 (ಶೂನ್ಯ-ಮೇಘ SLM) ಜತೆರೆಯ ಹೈಬ್ರಿಡ್ RAG\n", + "\n", + "> **ಲೇಖಕ:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **ಕೇಂದ್ರ:** ಸಣ್ಣ ಭಾಷಾ ಮಾದರಿಗಳು (SLMs), SQLite FTS5 BM25, ಡೆನ್ಸ್ ಎಂಬೆಡಿಂಗ್ಸ್, ರಿಸಿಪ್ರೋಕಲ್ ರ್ಯಾಂಕ್ ಫ್ಯೂಶನ್ (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. ಪ್ರೇರಣೆ: ಸ್ಥಳೀಯ SLM ಗಳಲ್ಲಿ ಕೀವರ್ಡ್ ರಿಕಾಲ್ ದ್ವಿಧ್ವನಿ\n", + "ಡೆನ್ಸ್ ವೇಕ್ಟರ್ ಎಂಬೆಡಿಂಗ್ಸ್ ಮೇಲೆ ಸಂಪೂರ್ಣ ನಂಬಿಕೆಯಿರುವ ಸ್ಟ್ಯಾಂಡರ್ಡ್ RAG ನಿರ್ಮಿತಿಗಳು ನಿಖರ ಸಂಖ್ಯೆ ಟೋಕನ್‌ಗಳು (ಉದಾ: `2,340,000 TL`, ಒಪ್ಪಂದ ಕೋಡ್ಗಳು, ಖಾತೆ ಸಂಖ್ಯೆಗಳು) ಅನ್ನು ಪಡೆಯಲು ವಿಫಲವಾಗುತ್ತವೆ. \n", + "ಇದಕ್ಕೊಳಗೆ, ಸ್ಪಾರ್ಸ್ ಲೆಕ್ಸಿಕಲ್ ಹುಡುಕು (BM25) ಸಾಂಕೇತಿಕ ಬಹುಮುಖ ಪದಗಳು ಮತ್ತು ಪರ್ಯಾಯ ಪ್ರಶ್ನೆಗಳನ್ನು ತಟ್ಟದಂತೆ ಮಾಡುತ್ತದೆ.\n", + "\n", + "ಈ ಕುಕ್‌ಬುಕ್ ಹೇಗೆ **ಹೊಸ-ವೇಗ, ಸ್ಮೃತಿ ಆಧಾರಿತ ಹೈಬ್ರಿಡ್ ರಿಟ್ರೀವಲ್ ಎಂಜಿನ್** ಅನ್ನು ಜೋಡಿಸುವುದನ್ನು ತೋರಿಸುತ್ತದೆ:\n", + "1. **ಡೆನ್ಸ್ ವೇಕ್ಟರ್‌ಗಳು** (ಕೋಸೈನ್ ಸادೃಶ್ಯತೆ)\n", + "2. **ಸ್ಪಾರ್ಸ್ ಲೆಕ್ಸಿಕಲ್ ಹುಡುಕು** (SQLite FTS5 BM25)\n", + "3. **ರಿಸಿಪ್ರೋಕಲ್ ರ್ಯಾಂಕ್ ಫ್ಯೂಶನ್ (RRF, $k=60$)**\n", + "4. **ಭೂಮಿತ ಉಲ್ಲೇಖ ತಯಾರಿಕೆ (`[1]`, `[2]`)** ಮೈಕ್ರೋಸಾಫ್ಟ್ `phi-4-mini` ಮೂಲಕ.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. ಡುಯಲ್ SQLite ಸ್ವರೂಪ (ಘನ ವעקטರ್‌ಗಳು + ವರ್ಚುವಲ್ FTS5 BM25 ಟೇಬಲ್)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. ಮಾದರಿ ಸೇವನೆ ಮತ್ತು ನಿರ್ವಹಣೆ ಬೆಂಚ್‌ಮಾರ್ಕ್\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. ಮೈಕ್ರೋಸಾಫ್ಟ್ phi-4-mini ಗಾಗಿ ನೆಲಸೇಸಿದ ಪ್ರಾಂಪ್ಟ್ ರೂಪರೇಷೆ\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**ಅಸ್ವೀಕಾರ**:\nಈ ದಸ್ತಾವೇಜು AI ಅನುವಾದ ಸೇವೆ [Co-op Translator](https://github.com/Azure/co-op-translator) ಬಳಸಿ ಅನುವಾದಿಸಲಾಗಿದೆ. ನಾವು ನಿಖರತೆಯನ್ನು ಸಾಧಿಸಲು ಪ್ರಯತ್ನಿಸುತ್ತಿದ್ದರೂ, ದಯವಿಟ್ಟು ಗಮನಿಸಿ, ಸ್ವಯಂಚಾಲಿತ ಅನುವಾದಗಳಲ್ಲಿ ದೋಷಗಳು ಅಥವಾ ಅಸಡ್ಡೆಗಳು ಇರಬಹುದು. ಮೂಲ ಭಾಷೆಯಲ್ಲಿರುವ ಮೂಲ ದಸ್ತಾವೇಜು ಪ್ರಾಮಾಣಿಕ ಮೂಲವೆಂದು ಪರಿಗಣಿಸಬೇಕು. ಪ್ರಮುಖ ಮಾಹಿತಿಗಾಗಿ, ವೃತ್ತಿಪರ ಮಾನವ ಅನುವಾದವನ್ನು ಶಿಫಾರಸು ಮಾಡಲಾಗುತ್ತದೆ. ಈ ಅನುವಾದವನ್ನು ಬಳಸುವ ಮೂಲಕ ಉಂಟಾಗುವ ಯಾವುದೇ ತಪ್ಪು ಅರ್ಥಗಳ ಅಥವಾ ತಪ್ಪು ವ್ಯಾಖ್ಯಾನಗಳ ಬಗ್ಗೆ ನಾವು ಹೊಣೆಗಾರರಲ್ಲ.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/ko/.co-op-translator.json b/translations/ko/.co-op-translator.json index cf2e37e17..52e5a99a4 100644 --- a/translations/ko/.co-op-translator.json +++ b/translations/ko/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "ko" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:14:58+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:26:42+00:00", "source_file": "README.md", "language_code": "ko" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "ko" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:29:55+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "ko" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:09+00:00", diff --git a/translations/ko/README.md b/translations/ko/README.md index db7cf3861..6a61d828e 100644 --- a/translations/ko/README.md +++ b/translations/ko/README.md @@ -1,40 +1,40 @@ -# Phi Cookbook: Microsoft의 Phi 모델을 사용한 실습 예제 +# Phi 요리책: Microsoft의 Phi 모델을 활용한 실습 예제 [![GitHub Codespaces에서 샘플 열기 및 사용하기](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Dev Containers에서 열기](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub 기여자들](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub 이슈들](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 기여자](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 이슈](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub 풀 리퀘스트](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![PR 환영](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub 감시자](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 워처](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub 포크](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub 스타](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 별](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi는 Microsoft가 개발한 일련의 오픈 소스 AI 모델입니다. +Phi는 Microsoft에서 개발한 오픈 소스 AI 모델 시리즈입니다. -Phi는 현재 다언어, 추론, 텍스트/채팅 생성, 코딩, 이미지, 오디오 및 기타 시나리오에서 매우 우수한 벤치마크를 가진 가장 강력하고 비용 효율적인 소형 언어 모델(SLM)입니다. +Phi는 현재 다국어, 추론, 텍스트/채팅 생성, 코딩, 이미지, 오디오 및 기타 시나리오에서 매우 우수한 벤치마크를 갖춘 가장 강력하고 비용 효율적인 소형 언어 모델(SLM)입니다. -Phi는 클라우드나 엣지 장치에 배포할 수 있으며, 제한된 컴퓨팅 파워로도 쉽게 생성형 AI 애플리케이션을 구축할 수 있습니다. +Phi를 클라우드 또는 엣지 디바이스에 배포할 수 있으며, 제한된 컴퓨팅 파워로 생성형 AI 애플리케이션을 쉽게 개발할 수 있습니다. -이 리소스 사용을 시작하려면 다음 단계를 따르세요: +다음 단계를 따라 이 리소스를 사용해 보십시오: 1. **저장소 포크하기**: 클릭 [![GitHub 포크](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **저장소 클론하기**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Microsoft AI Discord 커뮤니티에 가입하여 전문가 및 동료 개발자 만나기**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Microsoft AI Discord 커뮤니티에 참여하여 전문가 및 개발자들과 만남**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/ko/cover.eb18d1b9605d754b.webp) ### 🌐 다국어 지원 -#### GitHub Action을 통해 지원 (자동화 및 항상 최신 유지) +#### GitHub Action을 통한 지원 (자동화 및 항상 최신 상태 유지) -[아랍어](../ar/README.md) | [벵골어](../bn/README.md) | [불가리아어](../bg/README.md) | [버마어 (미얀마)](../my/README.md) | [중국어 (간체)](../zh-CN/README.md) | [중국어 (번체, 홍콩)](../zh-HK/README.md) | [중국어 (번체, 마카오)](../zh-MO/README.md) | [중국어 (번체, 대만)](../zh-TW/README.md) | [크로아티아어](../hr/README.md) | [체코어](../cs/README.md) | [덴마크어](../da/README.md) | [네덜란드어](../nl/README.md) | [에스토니아어](../et/README.md) | [핀란드어](../fi/README.md) | [프랑스어](../fr/README.md) | [독일어](../de/README.md) | [그리스어](../el/README.md) | [히브리어](../he/README.md) | [힌디어](../hi/README.md) | [헝가리어](../hu/README.md) | [인도네시아어](../id/README.md) | [이탈리아어](../it/README.md) | [일본어](../ja/README.md) | [칸나다어](../kn/README.md) | [크메르어](../km/README.md) | [한국어](./README.md) | [리투아니아어](../lt/README.md) | [말레이어](../ms/README.md) | [말라얄람어](../ml/README.md) | [마라티어](../mr/README.md) | [네팔어](../ne/README.md) | [나이지리아 피진](../pcm/README.md) | [노르웨이어](../no/README.md) | [페르시아어 (파르시)](../fa/README.md) | [폴란드어](../pl/README.md) | [포르투갈어 (브라질)](../pt-BR/README.md) | [포르투갈어 (포르투갈)](../pt-PT/README.md) | [펀자브어 (구르무키)](../pa/README.md) | [루마니아어](../ro/README.md) | [러시아어](../ru/README.md) | [세르비아어 (키릴문자)](../sr/README.md) | [슬로바키아어](../sk/README.md) | [슬로베니아어](../sl/README.md) | [스페인어](../es/README.md) | [스와힐리어](../sw/README.md) | [스웨덴어](../sv/README.md) | [따갈로그어 (필리피노)](../tl/README.md) | [타밀어](../ta/README.md) | [텔루구어](../te/README.md) | [태국어](../th/README.md) | [터키어](../tr/README.md) | [우크라이나어](../uk/README.md) | [우르두어](../ur/README.md) | [베트남어](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](./README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **로컬에 클론하는 것을 선호하시나요?** +> **로컬에서 클론하기를 선호하시나요?** > > 이 저장소에는 50개 이상의 언어 번역이 포함되어 있어 다운로드 크기가 크게 증가합니다. 번역 없이 클론하려면 스파스 체크아웃을 사용하세요: > @@ -52,7 +52,7 @@ Phi는 클라우드나 엣지 장치에 배포할 수 있으며, 제한된 컴 > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> 이렇게 하면 코스 완료에 필요한 모든 것을 훨씬 빠른 다운로드로 얻을 수 있습니다. +> 이렇게 하면 훨씬 빠른 다운로드 속도로 코스 완료에 필요한 모든 것을 받을 수 있습니다. ## 목차 @@ -60,15 +60,15 @@ Phi는 클라우드나 엣지 장치에 배포할 수 있으며, 제한된 컴 - 소개 - [Phi 패밀리에 오신 것을 환영합니다](./md/01.Introduction/01/01.PhiFamily.md) - [환경 설정](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [주요 기술 이해하기](./md/01.Introduction/01/01.Understandingtech.md) + - [핵심 기술 이해하기](./md/01.Introduction/01/01.Understandingtech.md) - [Phi 모델을 위한 AI 안전성](./md/01.Introduction/01/01.AISafety.md) - [Phi 하드웨어 지원](./md/01.Introduction/01/01.Hardwaresupport.md) - - [플랫폼별 Phi 모델 및 가용성](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Phi 모델 및 플랫폼별 가용성](./md/01.Introduction/01/01.Edgeandcloud.md) - [Guidance-ai 및 Phi 사용하기](./md/01.Introduction/01/01.Guidance.md) - [GitHub 마켓플레이스 모델](https://github.com/marketplace/models) - [Azure AI 모델 카탈로그](https://ai.azure.com) -- 다양한 환경에서 Phi 추론 +- 다양한 환경에서 Phi 추론하기 - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub 모델](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry 모델 카탈로그](./md/01.Introduction/02/03.AzureAIFoundry.md) @@ -77,177 +77,180 @@ Phi는 클라우드나 엣지 장치에 배포할 수 있으며, 제한된 컴 - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Phi 패밀리 추론 +- Phi 패밀리 추론하기 - [iOS에서 Phi 추론](./md/01.Introduction/03/iOS_Inference.md) - [Android에서 Phi 추론](./md/01.Introduction/03/Android_Inference.md) - [Jetson에서 Phi 추론](./md/01.Introduction/03/Jetson_Inference.md) - [AI PC에서 Phi 추론](./md/01.Introduction/03/AIPC_Inference.md) - - [Apple MLX 프레임워크와 함께하는 Phi 추론](./md/01.Introduction/03/MLX_Inference.md) + - [Apple MLX 프레임워크로 Phi 추론](./md/01.Introduction/03/MLX_Inference.md) - [로컬 서버에서 Phi 추론](./md/01.Introduction/03/Local_Server_Inference.md) - [AI Toolkit을 사용한 원격 서버에서 Phi 추론](./md/01.Introduction/03/Remote_Interence.md) - - [Rust와 함께하는 Phi 추론](./md/01.Introduction/03/Rust_Inference.md) - - [로컬에서 Phi--Vision 추론](./md/01.Introduction/03/Vision_Inference.md) - - [Kaito AKS, Azure Containers(공식 지원)와 함께하는 Phi 추론](./md/01.Introduction/03/Kaito_Inference.md) + - [Rust로 Phi 추론](./md/01.Introduction/03/Rust_Inference.md) + - [로컬에서 Phi 비전 추론](./md/01.Introduction/03/Vision_Inference.md) + - [Kaito AKS, Azure 컨테이너(공식 지원)로 Phi 추론](./md/01.Introduction/03/Kaito_Inference.md) - [Phi 패밀리 양자화](./md/01.Introduction/04/QuantifyingPhi.md) - [llama.cpp를 사용한 Phi-3.5 / 4 양자화](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime용 생성형 AI 확장을 사용한 Phi-3.5 / 4 양자화](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [onnxruntime용 생성형 AI 확장 기능을 사용한 Phi-3.5 / 4 양자화](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [Intel OpenVINO를 사용한 Phi-3.5 / 4 양자화](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - [Apple MLX 프레임워크를 사용한 Phi-3.5 / 4 양자화](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- 평가 Phi +- Phi 평가 - [Responsible AI](./md/01.Introduction/05/ResponsibleAI.md) - [평가를 위한 Microsoft Foundry](./md/01.Introduction/05/AIFoundry.md) - - [Promptflow를 사용한 평가](./md/01.Introduction/05/Promptflow.md) + - [평가를 위한 Promptflow 사용법](./md/01.Introduction/05/Promptflow.md) -- Azure AI Search와 함께하는 RAG - - [Azure AI Search와 함께 Phi-4-mini 및 Phi-4-multimodal(RAG) 사용 방법](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- Azure AI Search와 RAG + - [Azure AI Search와 Phi-4-mini 및 Phi-4-multimodal(RAG) 사용 방법](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [SQLite FTS5 및 phi-4-mini를 사용한 제로-클라우드 로컬 하이브리드 RAG](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi 응용 프로그램 개발 샘플 - - 텍스트 및 채팅 애플리케이션 - - Phi-4 샘플 +- Phi 애플리케이션 개발 샘플 + - 텍스트 & 채팅 애플리케이션 + - Phi-4 샘플 - [📓] [Phi-4-mini ONNX 모델과 채팅하기](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 로컬 ONNX 모델과 .NET으로 채팅하기](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Phi-4 로컬 ONNX 모델 .NET과 채팅하기](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - [Semantic Kernel을 사용한 Phi-4 ONNX 채팅 .NET 콘솔 앱](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 샘플 - [Phi3, ONNX Runtime Web 및 WebGPU를 사용한 브라우저 내 로컬 챗봇](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino 채팅](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [멀티 모델 - 대화형 Phi-3-mini와 OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - 래퍼 빌드 및 Phi-3와 MLFlow 사용](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [OpenVino 챗](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [멀티 모델 - 대화형 Phi-3-mini 및 OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - 래퍼 빌딩 및 MLFlow와 함께 Phi-3 사용](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [모델 최적화 - Olive를 사용하여 ONNX Runtime Web용 Phi-3-min 모델 최적화 방법](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [Phi-3 mini-4k-instruct-onnx와 함께하는 WinUI3 앱](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + - [Phi-3 mini-4k-instruct-onnx를 사용하는 WinUI3 앱](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) -[WinUI3 멀티 모델 AI 기반 노트 앱 샘플](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [사용자 지정 Phi-3 모델 미세 조정 및 Prompt flow 통합](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Microsoft Foundry에서 Prompt flow로 사용자 지정 Phi-3 모델 미세 조정 및 통합](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Microsoft의 책임 있는 AI 원칙에 중점을 두고 Microsoft Foundry에서 미세 조정된 Phi-3 / Phi-3.5 모델 평가](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [Prompt flow와 함께 사용자 지정 Phi-3 모델 미세 조정 및 통합](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Microsoft Foundry에서 Prompt flow와 함께 사용자 지정 Phi-3 모델 미세 조정 및 통합](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Microsoft의 책임 있는 AI 원칙에 중점을 둔 Microsoft Foundry에서 미세 조정한 Phi-3 / Phi-3.5 모델 평가](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Phi-3.5-mini-instruct 언어 예측 샘플 (중국어/영어)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG 챗봇](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU를 사용하여 Phi-3.5-Instruct ONNX로 Prompt flow 솔루션 만들기](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Windows GPU를 사용하여 Phi-3.5-Instruct ONNX와 함께 Prompt flow 솔루션 생성](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Microsoft Phi-3.5 tflite를 사용하여 Android 앱 만들기](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Microsoft.ML.OnnxRuntime를 사용해 로컬 ONNX Phi-3 모델을 이용한 Q&A .NET 예제](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Semantic Kernel 및 Phi-3과 함께하는 콘솔 채팅 .NET 앱](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Microsoft.ML.OnnxRuntime를 사용한 로컬 ONNX Phi-3 모델을 활용한 Q&A .NET 예제](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Semantic Kernel과 Phi-3를 사용한 콘솔 챗 .NET 앱](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI 추론 SDK 코드 기반 샘플 - - Phi-4 샘플 - - [📓] [Phi-4-multimodal로 프로젝트 코드 생성하기](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Azure AI Inference SDK 코드 기반 샘플 + - Phi-4 샘플 + - [📓] [Phi-4-multimodal을 사용하여 프로젝트 코드 생성](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 샘플 - - [Microsoft Phi-3 가족을 사용하여 Visual Studio Code GitHub Copilot Chat 만들기](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub 모델로 Phi-3.5 Visual Studio Code 채팅 코파일럿 에이전트 만들기](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Microsoft Phi-3 패밀리로 자신만의 Visual Studio Code GitHub Copilot 챗 빌드하기](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub 모델로 Phi-3.5와 함께 자신만의 Visual Studio Code 챗 코파일럿 에이전트 만들기](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - 고급 추론 샘플 - - Phi-4 샘플 - - [📓] [Phi-4-mini-추론 또는 Phi-4-추론 샘플](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Microsoft Olive로 Phi-4-mini-추론 미세 조정](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Apple MLX로 Phi-4-mini-추론 미세 조정](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [GitHub 모델로 Phi-4-mini-추론](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Microsoft Foundry 모델로 Phi-4-mini-추론](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Phi-4 샘플 + - [📓] [Phi-4-mini-reasoning 또는 Phi-4-reasoning 샘플](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Microsoft Olive로 Phi-4-mini-reasoning 미세 조정하기](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLX로 Phi-4-mini-reasoning 미세 조정하기](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [GitHub 모델로 Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Microsoft Foundry 모델로 Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - 데모 - - [Hugging Face Spaces에 호스팅된 Phi-4-mini 데모](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Hugging Face Spaces에 호스팅된 Phi-4-multimodal 데모](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [Hugging Face Spaces에서 호스팅하는 Phi-4-mini 데모](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Hugginge Face Spaces에서 호스팅하는 Phi-4-multimodal 데모](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - 비전 샘플 - - Phi-4 샘플 - - [📓] [Phi-4-multimodal을 사용하여 이미지 읽기 및 코드 생성](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-4 샘플 + - [📓] [Phi-4-multimodal을 사용하여 이미지 읽기 및 코드 생성](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 샘플 - - [📓][Phi-3-vision-이미지 텍스트에서 텍스트로](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-이미지 텍스트 변환](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP 임베딩](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [데모: Phi-3 재활용](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - 시각적 언어 도우미 - Phi3-Vision 및 OpenVINO와 함께](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 비전 Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 비전 OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 비전 다중 프레임 또는 다중 이미지 샘플](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Microsoft.ML.OnnxRuntime .NET을 사용한 Phi-3 Vision 로컬 ONNX 모델](../../md/04.HOL/dotnet/src/LabsPhi303) - - [메뉴 기반 Phi-3 Vision 로컬 ONNX 모델 Microsoft.ML.OnnxRuntime .NET 사용](../../md/04.HOL/dotnet/src/LabsPhi304) + - [Phi-3-vision - 시각 언어 도우미 - Phi3-Vision 및 OpenVINO와 함께](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision 다중 프레임 또는 다중 이미지 샘플](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Microsoft.ML.OnnxRuntime .NET을 사용하는 Phi-3 Vision 로컬 ONNX 모델](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Microsoft.ML.OnnxRuntime .NET을 사용하는 메뉴 기반 Phi-3 Vision 로컬 ONNX 모델](../../md/04.HOL/dotnet/src/LabsPhi304) - 추론-비전 샘플 - - Phi-4-추론-비전-15B - - [📓] [Phi-4-추론-비전-15B를 사용한 무단 횡단 감지](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Phi-4-추론-비전-15B로 수학 문제 풀기](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-추론-비전-15B를 사용한 UI 감지](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - Phi-4-Reasoning-Vision-15B + - [📓] [Phi-4-Reasoning-Vision-15B를 사용한 무단횡단 감지](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B를 사용한 수학](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B를 사용한 UI 감지](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - 수학 샘플 - - Phi-4-미니-플래시-추론-지침 샘플 [Phi-4-미니-플래시-추론-지침 수학 데모](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-Mini-Flash-Reasoning-Instruct 샘플 [Phi-4-Mini-Flash-Reasoning-Instruct 수학 데모](./md/02.Application/09.Math/MathDemo.ipynb) - 오디오 샘플 - - Phi-4 샘플 - - [📓] [Phi-4-multimodal을 활용한 오디오 기록 추출](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - Phi-4 샘플 + - [📓] [Phi-4-multimodal을 사용한 오디오 녹취 추출](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Phi-4-multimodal 오디오 샘플](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Phi-4-multimodal 음성 번역 샘플](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET 콘솔 앱을 사용한 Phi-4-multimodal 오디오 분석 및 기록 생성](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [Phi-4-multimodal을 사용하여 오디오 파일을 분석하고 녹취록을 생성하는 .NET 콘솔 애플리케이션](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE 샘플 - Phi-3 / 3.5 샘플 - - [📓] [Phi-3.5 전문가 혼합 모델(MoEs) 소셜 미디어 샘플](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search 및 LlamaIndex를 사용한 RAG(검색 보강 생성) 파이프라인 구축](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 전문가 혼합 모델 (MoEs) 소셜 미디어 샘플](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search, LlamaIndex와 함께 검색 기반 생성 (RAG) 파이프라인 빌드](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - 함수 호출 샘플 - Phi-4 샘플 🆕 - [📓] [Phi-4-mini와 함께 함수 호출 사용하기](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [함수 호출을 사용하여 Phi-4-mini로 다중 에이전트 만들기](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [함수 호출을 사용하여 Phi-4-mini 다중 에이전트 생성하기](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Ollama와 함께 함수 호출 사용하기](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [ONNX와 함께 함수 호출 사용하기](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - 멀티모달 믹싱 샘플 + - 멀티모달 혼합 샘플 - Phi-4 샘플 🆕 - [📓] [기술 저널리스트로서 Phi-4-multimodal 사용하기](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET 콘솔 앱을 사용하여 Phi-4-multimodal로 이미지 분석](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [.NET 콘솔 애플리케이션을 사용하여 Phi-4-multimodal로 이미지 분석](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Phi 미세 조정 샘플 +- Phi 미세 조정 - [미세 조정 시나리오](./md/03.FineTuning/FineTuning_Scenarios.md) - [미세 조정 vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Phi-3을 산업 전문가로 만드는 미세 조정](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [VS Code용 AI 도구 키트를 사용한 Phi-3 미세 조정](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Azure Machine Learning 서비스를 사용한 Phi-3 미세 조정](./md/03.FineTuning/Introduce_AzureML.md) - - [Lora를 사용한 Phi-3 미세 조정](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora를 사용한 Phi-3 미세 조정](./md/03.FineTuning/FineTuning_Qlora.md) - - [Microsoft Foundry를 사용한 Phi-3 미세 조정](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Azure ML CLI/SDK를 사용한 Phi-3 미세 조정](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Microsoft Olive를 사용한 미세 조정](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive 실습 랩을 통한 미세 조정](./md/03.FineTuning/olive-lab/readme.md) - - [Weights and Bias를 사용한 Phi-3-vision 미세 조정](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Apple MLX 프레임워크를 사용한 Phi-3 미세 조정](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3를 업계 전문가로 만드는 미세 조정](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [AI Toolkit for VS Code로 Phi-3 미세 조정](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure Machine Learning 서비스로 Phi-3 미세 조정](./md/03.FineTuning/Introduce_AzureML.md) + - [Lora로 Phi-3 미세 조정](./md/03.FineTuning/FineTuning_Lora.md) + - [QLora로 Phi-3 미세 조정](./md/03.FineTuning/FineTuning_Qlora.md) + - [Microsoft Foundry로 Phi-3 미세 조정](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Azure ML CLI/SDK로 Phi-3 미세 조정](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Microsoft Olive로 미세 조정](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Microsoft Olive 실습 랩으로 미세 조정](./md/03.FineTuning/olive-lab/readme.md) + - [Weights and Bias로 Phi-3-vision 미세 조정](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Apple MLX 프레임워크로 Phi-3 미세 조정하기](./md/03.FineTuning/FineTuning_MLX.md) - [Phi-3-vision 미세 조정 (공식 지원)](./md/03.FineTuning/FineTuning_Vision.md) - - [Kaito AKS, Azure 컨테이너(공식 지원)로 Phi-3 미세 조정하기](./md/03.FineTuning/FineTuning_Kaito.md) + - [Kaito AKS, Azure 컨테이너로 Phi-3 미세 조정 (공식 지원)](./md/03.FineTuning/FineTuning_Kaito.md) - [Phi-3 및 3.5 Vision 미세 조정](https://github.com/2U1/Phi3-Vision-Finetune) - 실습 랩 - - [최첨단 모델 탐색: LLM, SLM, 로컬 개발 등](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP 잠재력 열기: Microsoft Olive로 미세 조정하기](https://github.com/azure/Ignite_FineTuning_workshop) + - [최신 모델 탐색: LLM, SLM, 로컬 개발 등](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP 잠재력 발휘: Microsoft Olive로 미세 조정](https://github.com/azure/Ignite_FineTuning_workshop) - 학술 연구 논문 및 출판물 - [Textbooks Are All You Need II: phi-1.5 기술 보고서](https://arxiv.org/abs/2309.05463) - - [Phi-3 기술 보고서: 휴대폰에서 로컬로 실행하는 고성능 언어 모델](https://arxiv.org/abs/2404.14219) + - [Phi-3 기술 보고서: 휴대폰에서 실행하는 고성능 언어 모델](https://arxiv.org/abs/2404.14219) - [Phi-4 기술 보고서](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini 기술 보고서: 혼합 LoRA를 통한 컴팩트하면서도 강력한 멀티모달 언어 모델](https://arxiv.org/abs/2503.01743) - - [차량 내 함수 호출을 위한 소형 언어 모델 최적화](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) 다지선다형 질문 답변을 위한 PHI-3 미세 조정: 방법론, 결과 및 과제](https://arxiv.org/abs/2501.01588) + - [Phi-4-Mini 기술 보고서: Mixture-of-LoRAs를 통한 컴팩트하지만 강력한 멀티모달 언어 모델](https://arxiv.org/abs/2503.01743) + - [차량 내 기능 호출을 위한 소형 언어 모델 최적화](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) PHI-3의 다중 선택 질문 응답 미세 조정: 방법론, 결과 및 과제](https://arxiv.org/abs/2501.01588) - [Phi-4-추론 기술 보고서](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Phi-4-mini-추론 기술 보고서](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Phi 모델 사용하기 -### Microsoft Foundry의 Phi +### Microsoft Foundry에서 Phi 사용하기 -Microsoft Phi를 사용하는 방법과 다양한 하드웨어 장치에서 E2E 솔루션을 구축하는 방법을 배울 수 있습니다. 직접 Phi를 경험하려면, [Microsoft Foundry Azure AI 모델 카탈로그](https://aka.ms/phi3-azure-ai)에서 모델을 시험하고 시나리오에 맞게 Phi를 맞춤화하는 것부터 시작하세요. [Microsoft Foundry 시작하기](/md/02.QuickStart/AzureAIFoundry_QuickStart.md)에서 자세히 알아볼 수 있습니다. +Microsoft Phi를 사용하고 다양한 하드웨어 장치에서 E2E 솔루션을 구축하는 방법을 배울 수 있습니다. Phi를 직접 경험하려면 모델을 사용해 보고 여러분의 시나리오에 맞게 Phi를 맞춤 설정해 보세요. [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai)에서 시작할 수 있습니다. 자세한 내용은 [Microsoft Foundry 시작하기](/md/02.QuickStart/AzureAIFoundry_QuickStart.md)에서 확인하세요. 플레이그라운드 -각 모델별 전용 플레이그라운드에서 모델을 테스트할 수 있습니다: [Azure AI Playground](https://aka.ms/try-phi3). +각 모델마다 전용 테스트 공간인 [Azure AI Playground](https://aka.ms/try-phi3)가 있습니다. -### GitHub 모델의 Phi +### GitHub 모델에서 Phi 사용하기 -Microsoft Phi를 사용하는 방법과 다양한 하드웨어 장치용 E2E 솔루션 구축 방법을 배울 수 있습니다. 직접 Phi를 경험하려면 [GitHub 모델 카탈로그](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo)를 통해 모델을 시험하고 시나리오에 맞게 Phi를 맞춤화하세요. [GitHub 모델 카탈로그 시작하기](/md/02.QuickStart/GitHubModel_QuickStart.md)에서 자세히 알아볼 수 있습니다. +Microsoft Phi를 사용하고 여러분의 다양한 하드웨어 장치에서 E2E 솔루션을 구축하는 방법을 배울 수 있습니다. Phi를 직접 경험하려면 모델을 사용해 보고 시나리오에 맞게 Phi를 맞춤 설정하세요. [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo)에서 시작하며, 자세한 내용은 [GitHub Model Catalog 시작하기](/md/02.QuickStart/GitHubModel_QuickStart.md)에서 확인하세요. 플레이그라운드 -각 모델별 전용 [플레이그라운드에서 모델 테스트](/md/02.QuickStart/GitHubModel_QuickStart.md) 가능합니다. +각 모델에는 전용 [모델 테스트 플레이그라운드](/md/02.QuickStart/GitHubModel_QuickStart.md)가 있습니다. -### Hugging Face의 Phi +### Hugging Face에서 Phi 사용하기 -모델은 [Hugging Face](https://huggingface.co/microsoft)에서도 찾을 수 있습니다. +또한 [Hugging Face](https://huggingface.co/microsoft)에서 모델을 찾을 수 있습니다. 플레이그라운드 -[Hugging Chat 플레이그라운드](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) + [Hugging Chat 플레이그라운드](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 기타 강좌 + ## 🎒 기타 강좌 -우리 팀은 다른 강좌들도 제작합니다! 확인해 보세요: +저희 팀에서 제공하는 다른 강좌들도 확인해 보세요! ### LangChain @@ -263,62 +266,64 @@ Microsoft Phi를 사용하는 방법과 다양한 하드웨어 장치용 E2E 솔 [![초보자를 위한 AI 에이전트](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### 생성 AI 시리즈 -[![초보자를 위한 생성 AI](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![생성 AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![생성 AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![생성 AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### 생성형 AI 시리즈 +[![초보자를 위한 생성형 AI](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![생성형 AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![생성형 AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![생성형 AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - + ### 핵심 학습 [![초보자를 위한 ML](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![초보자를 위한 데이터 과학](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![초보자를 위한 AI](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![초보자를 위한 사이버 보안](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![초보자를 위한 사이버보안](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![초보자를 위한 웹 개발](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![초보자를 위한 IoT](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![초보자를 위한 XR 개발](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### 코파일럿 시리즈 -[![AI 협업 프로그래밍용 코파일럿](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![AI 페어 프로그래밍을 위한 코파일럿](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![C#/.NET용 코파일럿](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![코파일럿 어드벤처](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## 책임 있는 AI -마이크로소프트는 고객이 당사의 AI 제품을 책임감 있게 사용할 수 있도록 지원하며, 학습 내용을 공유하고 투명성 노트 및 영향 평가와 같은 도구를 통해 신뢰 기반의 파트너십을 구축하는 데 최선을 다하고 있습니다. 이러한 자료는 [https://aka.ms/RAI](https://aka.ms/RAI)에서 찾을 수 있습니다. -마이크로소프트의 책임 있는 AI 접근 방식은 공정성, 신뢰성 및 안전성, 개인정보 보호 및 보안, 포용성, 투명성, 그리고 책임성에 관한 AI 원칙에 기반합니다. +Microsoft는 고객들이 AI 제품을 책임감 있게 사용할 수 있도록 돕고, 학습 내용을 공유하며, 투명성 노트 및 영향 평가 같은 도구를 통해 신뢰 기반의 파트너십을 구축하는 데 헌신하고 있습니다. 이러한 자료는 [https://aka.ms/RAI](https://aka.ms/RAI)에서 확인할 수 있습니다. +Microsoft의 책임 있는 AI 접근법은 공정성, 신뢰성 및 안전성, 개인정보 보호 및 보안, 포용성, 투명성, 그리고 책무성이라는 AI 원칙에 기반을 두고 있습니다. + +이 샘플에서 사용된 것과 같은 대규모 자연어, 이미지 및 음성 모델은 불공정하거나 신뢰할 수 없거나 공격적일 수 있는 방식으로 동작할 가능성이 있어 피해를 줄 수 있습니다. 위험과 한계에 대해 알고 싶다면 [Azure OpenAI 서비스 투명성 노트](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text)를 참고해 주세요. + -이 샘플에 사용된 것과 같은 대규모 자연어, 이미지, 음성 모델은 부당하거나 신뢰할 수 없거나 불쾌감을 줄 수 있는 행동을 할 가능성이 있어 피해를 줄 수 있습니다. 위험과 한계에 대한 정보를 얻으려면 [Azure OpenAI 서비스 투명성 노트](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text)를 참조하시기 바랍니다. -위험을 완화하는 권장 방법은 아키텍처에 유해한 동작을 감지하고 방지할 수 있는 안전 시스템을 포함하는 것입니다. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview)는 독립적인 보호 계층을 제공하여 애플리케이션과 서비스에서 유해한 사용자 생성 및 AI 생성 콘텐츠를 감지할 수 있습니다. Azure AI Content Safety에는 유해한 자료를 감지할 수 있는 텍스트 및 이미지 API가 포함되어 있습니다. Microsoft Foundry 내에서 Content Safety 서비스는 다양한 양식의 유해 콘텐츠를 감지하는 샘플 코드를 보고, 탐색하고, 시도해 볼 수 있도록 합니다. 다음 [빠른 시작 문서](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest)는 서비스에 요청하는 방법을 안내합니다. +이러한 위험을 완화하는 권장 방법은 유해한 행위를 감지하고 방지할 수 있는 안전 시스템을 아키텍처에 포함하는 것입니다. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview)는 독립적인 보호 계층을 제공하여, 애플리케이션과 서비스에서 유해한 사용자 생성 콘텐츠 및 AI 생성 콘텐츠를 감지할 수 있습니다. Azure AI Content Safety는 유해한 자료를 감지할 수 있는 텍스트 및 이미지 API를 포함합니다. Microsoft Foundry 내에서 Content Safety 서비스는 다양한 모달리티에서 유해 콘텐츠를 감지하기 위한 샘플 코드를 보고 탐색하고 시험해볼 수 있게 합니다. 다음 [빠른 시작 문서](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest)가 서비스에 요청하는 과정을 안내합니다. -또 다른 고려 사항은 전체 애플리케이션 성능입니다. 다중 양식 및 다중 모델 애플리케이션의 경우, 성능은 시스템이 사용자와 개발자가 기대하는 대로, 유해한 출력을 생성하지 않는 것을 포함하여 수행하는 것을 의미합니다. [성능 및 품질 및 위험 및 안전 평가자](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)를 사용하여 전체 애플리케이션의 성능을 평가하는 것이 중요합니다. 또한 [사용자 지정 평가자](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators)를 생성하고 평가할 수 있는 기능도 제공합니다. +또 고려해야 할 점은 전반적인 애플리케이션 성능입니다. 다중 모달 및 다중 모델 애플리케이션의 경우 성능은 시스템이 사용자와 귀하가 기대하는 대로 작동함을 의미하며, 여기에는 유해한 출력을 생성하지 않는 것도 포함됩니다. [성능 및 품질, 위험 및 안전 평가자](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)를 사용하여 전체 애플리케이션의 성능을 평가하는 것이 중요합니다. 또한 [사용자 지정 평가자](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators)를 생성하고 평가할 수도 있습니다. -[Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html)를 사용하여 개발 환경에서 AI 애플리케이션을 평가할 수 있습니다. 테스트 데이터 세트 또는 목표가 주어지면, 생성 AI 애플리케이션 생성물은 내장된 평가자 또는 선택한 사용자 지정 평가자를 통해 정량적으로 측정됩니다. 시스템을 평가하기 위한 azure ai evaluation sdk를 시작하려면 [빠른 시작 가이드](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)를 참조할 수 있습니다. 평가 실행을 완료하면 [Microsoft Foundry에서 결과를 시각화](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)할 수 있습니다. +개발 환경에서 [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html)를 사용하여 AI 애플리케이션을 평가할 수 있습니다. 테스트 데이터 세트나 대상으로부터 귀하의 생성형 AI 애플리케이션 생성물을 내장 평가자 또는 선택한 사용자 지정 평가자로 정량적으로 측정합니다. 시스템 평가를 시작하려면 [빠른 시작 가이드](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)를 따라 하십시오. 평가 실행을 완료하면 [Microsoft Foundry에서 결과를 시각화](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)할 수 있습니다. ## 상표 -이 프로젝트에는 프로젝트, 제품 또는 서비스의 상표 또는 로고가 포함될 수 있습니다. Microsoft 상표 또는 로고의 권한 있는 사용은 [Microsoft의 상표 및 브랜드 지침](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)을 준수해야 합니다. -이 프로젝트의 수정된 버전에서 Microsoft 상표 또는 로고를 사용하는 것은 혼동을 초래하거나 Microsoft 후원을 시사해서는 안 됩니다. 제3자 상표 또는 로고의 모든 사용은 해당 제3자의 정책에 따릅니다. +이 프로젝트에는 프로젝트, 제품 또는 서비스에 대한 상표나 로고가 포함될 수 있습니다. Microsoft 상표 또는 로고의 승인된 사용은 [Microsoft 상표 및 브랜드 가이드라인](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)을 준수해야 합니다. +수정된 버전의 이 프로젝트에서 Microsoft 상표 또는 로고를 사용하는 경우 혼동을 일으키거나 Microsoft 후원을 암시하지 않아야 합니다. 타사 상표 또는 로고의 모든 사용은 해당 타사의 정책에 따릅니다. -## 도움말 받기 +## 도움 받기 -AI 앱 개발 중에 문제가 발생하거나 질문이 있는 경우 다음에 참여하세요: +AI 앱 구축 중에 막히거나 질문이 있을 경우 다음에 참여하세요: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -제품 피드백이나 빌드 중 오류가 있는 경우 방문하세요: +제품 피드백이나 빌드 중 오류가 있을 경우 방문하세요: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**면책 조항**: -이 문서는 AI 번역 서비스 [Co-op Translator](https://github.com/Azure/co-op-translator)를 사용하여 번역되었습니다. 정확성을 위해 최선을 다하고 있으나, 자동 번역은 오류나 부정확한 내용이 포함될 수 있습니다. 원문은 해당 언어의 원본 문서가 권위 있는 출처로 간주되어야 합니다. 중요한 정보의 경우 전문적인 인간 번역을 권장합니다. 이 번역 사용으로 인해 발생하는 모든 오해나 오역에 대해 당사는 책임을 지지 않습니다. +**면책 조항**: +이 문서는 AI 번역 서비스 [Co-op Translator](https://github.com/Azure/co-op-translator)를 사용하여 번역되었습니다. 정확성을 기하기 위해 노력하고 있으나, 자동 번역은 오류나 부정확한 부분이 있을 수 있음을 유의하시기 바랍니다. 원본 문서의 원어본이 권위 있는 자료로 간주되어야 합니다. 중요한 정보의 경우, 전문가의 인간 번역을 권장합니다. 이 번역 사용으로 인해 발생하는 오해나 잘못된 해석에 대해 당사는 책임을 지지 않습니다. \ No newline at end of file diff --git a/translations/ko/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/ko/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..4f216d9c0 --- /dev/null +++ b/translations/ko/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Microsoft phi-4-mini 및 SQLite FTS5를 이용한 하이브리드 RAG (제로-클라우드 SLM)\n", + "\n", + "> **저자:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **중점 분야:** 소형 언어 모델(SLM), SQLite FTS5 BM25, 밀집 임베딩, 상호 순위 융합(RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. 동기: 로컬 SLM의 키워드 검색 문제\n", + "순수 밀집 벡터 임베딩에만 의존하는 표준 RAG 아키텍처는 종종 정확한 숫자 토큰(예: `2,340,000 TL`, 계약 코드, 계좌 번호)을 검색하지 못합니다. \n", + "반면, 희소한 어휘 검색(BM25)은 의미상의 동의어와 바뀐 질문을 놓칩니다.\n", + "\n", + "이 조리법에서는 다음을 결합한 **고속 인메모리 하이브리드 검색 엔진** 구현 방법을 보여줍니다:\n", + "1. **밀집 벡터** (코사인 유사도)\n", + "2. **희소 어휘 검색** (SQLite FTS5 BM25)\n", + "3. **상호 순위 융합 (RRF, $k=60$)**\n", + "4. Microsoft `phi-4-mini`를 사용한 **근거 인용 생성 (`[1]`, `[2]`)**\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. 이중 SQLite 스키마 (조밀 벡터 + 가상 FTS5 BM25 테이블)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. 샘플 수집 및 실행 벤치마크\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Microsoft phi-4-mini에 대한 근거 있는 프롬프트 작성\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**면책 조항**:\n이 문서는 AI 번역 서비스 [Co-op Translator](https://github.com/Azure/co-op-translator)를 사용하여 번역되었습니다. 정확성을 기하기 위해 노력하고 있으나, 자동 번역은 오류나 부정확한 부분이 있을 수 있음을 유의하시기 바랍니다. 원본 문서의 원어본이 권위 있는 자료로 간주되어야 합니다. 중요한 정보의 경우, 전문가의 인간 번역을 권장합니다. 이 번역 사용으로 인해 발생하는 오해나 잘못된 해석에 대해 당사는 책임을 지지 않습니다.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/lt/.co-op-translator.json b/translations/lt/.co-op-translator.json index 0f68ed53c..47e2217c7 100644 --- a/translations/lt/.co-op-translator.json +++ b/translations/lt/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "lt" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:25:44+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T16:49:36+00:00", "source_file": "README.md", "language_code": "lt" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "lt" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T16:50:44+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "lt" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-09-12T15:01:52+00:00", diff --git a/translations/lt/README.md b/translations/lt/README.md index 983bcf6c5..e78de3d36 100644 --- a/translations/lt/README.md +++ b/translations/lt/README.md @@ -1,42 +1,42 @@ -# Phi Receptų Knyga: Praktiniai Pavyzdžiai su „Microsoft“ Phi Modeliais +# Phi Virtuvė: Praktiniai Pavyzdžiai su Microsoft Phi Modeliais [![Atidarykite ir naudokite pavyzdžius GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Atidaryti Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub bendradarbiai](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub problemos](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub užklausos traukiant (pull requests)](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Laukiami](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub pull-užklausos](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Sveiki](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![GitHub stebėtojai](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub šakos (forks)](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub žvaigždės (stars)](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub šakos](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub žvaigždės](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi yra atvirojo kodo DI modelių serija, sukurta „Microsoft“. +Phi yra atvirojo kodo DI modelių serija, sukurta Microsoft. -Šiuo metu Phi yra galingiausias ir ekonomiškiausias mažas kalbos modelis (SLM), kuris turi labai gerus etaloninius rodiklius daugikalbiškumo, samprotavimo, teksto/pokalbių generavimo, kodavimo, vaizdų, garso ir kitose srityse. +Phi šiuo metu yra galingiausias ir ekonomiškiausias mažas kalbos modelis (SLM), turintis labai gerus rezultatus keliomis kalbomis, sprendžiant uždavinius, generuojant tekstą/pokalbį, kodą, vaizdus, garsą ir kitose situacijose. -Jūs galite diegti Phi debesyje arba kraštiniuose įrenginiuose, taip pat lengvai kurti generatyvios DI programas su ribotais skaičiavimo ištekliais. +Phi galima diegti debesyje arba krašto įrenginiuose, ir lengvai kurti generatyvias DI programas su apribota skaičiavimo galia. Sekite šiuos žingsnius, kad pradėtumėte naudotis šiais ištekliais: -1. **Padarykite "Fork" šiam Paskirstymui**: Spauskite [![GitHub šakos (forks)](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Klonuokite Paskirstymą**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Prisijunkite prie Microsoft AI Discord bendruomenės ir susipažinkite su ekspertais ir kitais kūrėjais**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +1. **Sukurkite šaką (Fork)**: Spauskite [![GitHub šakos](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Klonuokite saugyklą**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Prisijunkite prie Microsoft AI Discord bendruomenės ir susipažinkite su ekspertais bei kitais kūrėjais**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/lt/cover.eb18d1b9605d754b.webp) +![dengiamasis](../../translated_images/lt/cover.eb18d1b9605d754b.webp) ### 🌐 Daugiakalbė Parama -#### Palaikoma per GitHub Action (automatizuota ir visada atnaujinama) +#### Palaikoma per GitHub Action (Automatizuota ir Visada Atnaujinta) -[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](./README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) +[Arabų](../ar/README.md) | [Bengalų](../bn/README.md) | [Bulgarų](../bg/README.md) | [Birmos (Mianmaro)](../my/README.md) | [Kinų (supaprastinta)](../zh-CN/README.md) | [Kinų (tradicinė, Honkongas)](../zh-HK/README.md) | [Kinų (tradicinė, Makao)](../zh-MO/README.md) | [Kinų (tradicinė, Taivanas)](../zh-TW/README.md) | [Kroatų](../hr/README.md) | [Čekų](../cs/README.md) | [Danų](../da/README.md) | [Olandų](../nl/README.md) | [Estų](../et/README.md) | [Suomių](../fi/README.md) | [Prancūzų](../fr/README.md) | [Vokiečių](../de/README.md) | [Graikų](../el/README.md) | [Hebrajų](../he/README.md) | [Hindi](../hi/README.md) | [Vengrų](../hu/README.md) | [Indoneziečių](../id/README.md) | [Italų](../it/README.md) | [Japonų](../ja/README.md) | [Kannados](../kn/README.md) | [Khmerų](../km/README.md) | [Korėjiečių](../ko/README.md) | [Lietuvių](./README.md) | [Malajiečių](../ms/README.md) | [Malajalamų](../ml/README.md) | [Maračių](../mr/README.md) | [Nepaliečių](../ne/README.md) | [Nigerijos Pidžino](../pcm/README.md) | [Norvegų](../no/README.md) | [Persų (Farsi)](../fa/README.md) | [Lenkų](../pl/README.md) | [Portugalių (Brazilija)](../pt-BR/README.md) | [Portugalių (Portugalija)](../pt-PT/README.md) | [Pandžabių (Gurmukhi)](../pa/README.md) | [Rumunų](../ro/README.md) | [Rusų](../ru/README.md) | [Serbų (Kirilikos)](../sr/README.md) | [Slovakų](../sk/README.md) | [Slovėnų](../sl/README.md) | [Ispanų](../es/README.md) | [Svahilių](../sw/README.md) | [Švedų](../sv/README.md) | [Tagalogų (Filipinų)](../tl/README.md) | [Tamilų](../ta/README.md) | [Telugų](../te/README.md) | [Tailando](../th/README.md) | [Turkų](../tr/README.md) | [Ukrainiečių](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamiečių](../vi/README.md) -> **Teikiate pirmenybę klonuoti lokaliai?** +> **Norite klonuoti vietoje?** > -> Šiame paskirstyme yra daugiau nei 50 kalbų vertimų, kurie ženkliai padidina parsisiuntimo dydį. Norėdami klonuoti be vertimų, naudokite ribotą atsisiuntimą: +> Ši saugykla apima daugiau nei 50 kalbų vertimų, kas žymiai padidina atsisiuntimo dydį. Norėdami klonuoti be vertimų, naudokite retinimo parinktį: > > **Bash / macOS / Linux:** > ```bash @@ -52,7 +52,7 @@ Sekite šiuos žingsnius, kad pradėtumėte naudotis šiais ištekliais: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Tai suteiks jums viską, ko reikia kursui užbaigti, daug greičiau atsisiunčiant. +> Tai suteiks jums viską, ko reikia kursui atlikti, su daug greitesniu atsisiuntimu. ## Turinys @@ -60,259 +60,264 @@ Sekite šiuos žingsnius, kad pradėtumėte naudotis šiais ištekliais: - Įvadas - [Sveiki atvykę į Phi šeimą](./md/01.Introduction/01/01.PhiFamily.md) - [Jūsų aplinkos nustatymas](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Pagrindinių technologijų supratimas](./md/01.Introduction/01/01.Understandingtech.md) + - [Svarbiausių technologijų supratimas](./md/01.Introduction/01/01.Understandingtech.md) - [DI saugumas Phi modeliams](./md/01.Introduction/01/01.AISafety.md) - - [Phi aparatinės įrangos palaikymas](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi modeliai ir jų prieinamumas platformose](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Naudojant Guidance-ai ir Phi](./md/01.Introduction/01/01.Guidance.md) - - [GitHub Marketplace modeliai](https://github.com/marketplace/models) - - [Azure DI modelių katalogas](https://ai.azure.com) + - [Phi aparatūros palaikymas](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Phi modeliai ir pasiekiamumas platformose](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Naudojimasis Guidance-ai ir Phi](./md/01.Introduction/01/01.Guidance.md) + - [GitHub Marketplace Modeliai](https://github.com/marketplace/models) + - [Azure DI Modelių katalogas](https://ai.azure.com) -- Phi spėjimas skirtingose aplinkose +- Phi išvedimas skirtingose aplinkose - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [GitHub modeliai](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry modelių katalogas](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [GitHub Modeliai](./md/01.Introduction/02/02.GitHubModel.md) + - [Microsoft Foundry Modelių katalogas](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [DI įrankių rinkinys VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [AI Įrankių rinkinys VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry vietinis](./md/01.Introduction/02/07.FoundryLocal.md) - -- Phi šeimos spėjimas - - [Phi spėjimas iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Phi spėjimas Android](./md/01.Introduction/03/Android_Inference.md) - - [Phi spėjimas Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Phi spėjimas DI kompiuteryje](./md/01.Introduction/03/AIPC_Inference.md) - - [Phi spėjimas su Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Phi spėjimas vietiniame serveryje](./md/01.Introduction/03/Local_Server_Inference.md) - - [Phi spėjimas nuotoliniame serveryje naudojant DI įrankių rinkinį](./md/01.Introduction/03/Remote_Interence.md) - - [Phi spėjimas su Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Phi – regos spėjimas vietoje](./md/01.Introduction/03/Vision_Inference.md) - - [Phi spėjimas su Kaito AKS, Azure konteineriais (oficialus palaikymas)](./md/01.Introduction/03/Kaito_Inference.md) -- [Phi šeimos kiekybinimas](./md/01.Introduction/04/QuantifyingPhi.md) - - [Phi-3.5 / 4 kiekybinimas naudojant llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Phi-3.5 / 4 kiekybinimas naudojant Generatyvios DI plėtinius for onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Phi-3.5 / 4 kiekybinimas naudojant Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Phi-3.5 / 4 kiekybinimas naudojant Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Foundry lokalus](./md/01.Introduction/02/07.FoundryLocal.md) + +- Phi šeimos išvedimas + - [Phi išvedimas iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Phi išvedimas Android](./md/01.Introduction/03/Android_Inference.md) + - [Phi išvedimas Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Phi išvedimas AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Phi išvedimas su Apple MLX sistema](./md/01.Introduction/03/MLX_Inference.md) + - [Phi išvedimas vietiniame serveryje](./md/01.Introduction/03/Local_Server_Inference.md) + - [Phi išvedimas nuotoliniame serveryje naudojant AI Įrankių rinkinį](./md/01.Introduction/03/Remote_Interence.md) + - [Phi išvedimas su Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Phi – vizijos išvedimas vietoje](./md/01.Introduction/03/Vision_Inference.md) + - [Phi išvedimas su Kaito AKS, Azure konteineriais (oficialus palaikymas)](./md/01.Introduction/03/Kaito_Inference.md) +- [Phi šeimos kiekinimas](./md/01.Introduction/04/QuantifyingPhi.md) + - [Phi-3.5 / 4 kiekinimas naudojant llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Phi-3.5 / 4 kiekinimas naudojant Generative AI plėtinius onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Phi-3.5 / 4 kiekinimas naudojant Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Phi-3.5 / 4 kiekinimas naudojant Apple MLX sistemą](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Phi vertinimas - - [Atsakinga DI](./md/01.Introduction/05/ResponsibleAI.md) + - [Atsakingas DI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry vertinimui](./md/01.Introduction/05/AIFoundry.md) - [Naudojant Promptflow vertinimui](./md/01.Introduction/05/Promptflow.md) - RAG su Azure DI paieška - [Kaip naudoti Phi-4-mini ir Phi-4-multimodal (RAG) su Azure DI paieška](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Nulinės debesies vietinis hibridinis RAG su SQLite FTS5 ir phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi programėlių kūrimo pavyzdžiai - - Teksto ir pokalbių programėlės - - Phi-4 pavyzdžiai +- Phi programų kūrimo pavyzdžiai + - Teksto ir pokalbių programos + - Phi-4 pavyzdžiai - [📓] [Pokalbis su Phi-4-mini ONNX modeliu](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Pokalbis su Phi-4 vietiniu ONNX modeliu .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Pokalbis .NET konsolėje su Phi-4 ONNX, naudojant Semantinį branduolį](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 pavyzdžiai + - [Pokalbių .NET konsolės programa su Phi-4 ONNX naudojant Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Pavyzdžiai - [Vietinis pokalbių robotas naršyklėje naudojant Phi3, ONNX Runtime Web ir WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Daugiamodelinis - Interaktyvus Phi-3-mini ir OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Apvyniojimo kūrimas ir Phi-3 naudojimas su MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Modelio optimizavimas - Kaip optimizuoti Phi-3-min modelį ONNX Runtime Web naudojant Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [OpenVino Pokalbis](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Daugelio modelių - Interaktyvus Phi-3-mini ir OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Sukuriant apvalkalą ir naudojant Phi-3 su MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Modelio optimizavimas - Kaip optimizuoti Phi-3-min modelį ONNX Runtime Web su Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [WinUI3 programa su Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 daugiamodelio DI varoma užrašų programėlės pavyzdys](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Tobulinti ir integruoti individualius Phi-3 modelius su Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Tobulinti ir integruoti individualius Phi-3 modelius su Prompt flow Microsoft Foundry aplinkoje](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Įvertinti tobulintą Phi-3 / Phi-3.5 modelį Microsoft Foundry, sutelkiant dėmesį į Microsoft atsakingos DI principus](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + -[WinUI3 Daugelio modelių AI palaikoma pastabų programa Pavyzdys](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Papildomas apmokymas ir pasirinktinių Phi-3 modelių integravimas su Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Papildomas apmokymas ir pasirinktinių Phi-3 modelių integravimas su Prompt flow Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Įvertinkite papildomai apmokytą Phi-3 / Phi-3.5 modelį Microsoft Foundry, orientuotą į Microsoft atsakingo DI principus](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Phi-3.5-mini-instruct kalbos prognozavimo pavyzdys (kinų/anglų)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG pokalbių robotas](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Naudojant Windows GPU sukurti Prompt flow sprendimą su Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Naudojant Microsoft Phi-3.5 tflite sukurti Android programėlę](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Naudojant Windows GPU kūrti Prompt flow sprendimą su Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Naudojant Microsoft Phi-3.5 tflite kurti Android programėlę](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - [Klausimai ir atsakymai .NET pavyzdys naudojant vietinį ONNX Phi-3 modelį su Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Konsolinė pokalbių .NET programa su Semantic Kernel ir Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure DI objektų atpažinimo SDK kodo pavyzdžiai - - Phi-4 pavyzdžiai - - [📓] [Projekto kodo generavimas naudojant Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 pavyzdžiai - - [Sukurti savo Visual Studio Code GitHub Copilot pokalbį su Microsoft Phi-3 šeima](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Sukurti savo Visual Studio Code pokalbių Copilot agentą su Phi-3.5 pagal GitHub modelius](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - Sudėtingo mąstymo pavyzdžiai - - Phi-4 pavyzdžiai - - [📓] [Phi-4-mini-mąstymo arba Phi-4-mąstymo pavyzdžiai](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Phi-4-mini-mąstymo tobulinimas su Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-mąstymo tobulinimas su Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-mąstymas su GitHub modeliais](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-mąstymas su Microsoft Foundry modeliais](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [Konsolės pokalbių .NET programa su Semantic Kernel ir Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI Inference SDK Kodo pagrindu pavyzdžiai + - Phi-4 Pavyzdžiai + - [📓] [Generuoti projekto kodą naudojant Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Phi-3 / 3.5 Pavyzdžiai + - [Sukurkite savo Visual Studio Code GitHub Copilot Pokalbių su Microsoft Phi-3 šeima](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Sukurkite savo Visual Studio Code Chat Copilot agentą su Phi-3.5 pagal GitHub Modelius](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - Pažangių samprotavimų pavyzdžiai + - Phi-4 Pavyzdžiai + - [📓] [Phi-4-mini-samprotavimų arba Phi-4-samprotavimų pavyzdžiai](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Papildomas Phi-4-mini-samprotavimų apmokymas su Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Papildomas Phi-4-mini-samprotavimų apmokymas su Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-samprotavimai su GitHub Modeliais](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-samprotavimai su Microsoft Foundry Modeliais](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demonstracijos - [Phi-4-mini demonstracijos talpinamos Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demonstracijos talpinamos Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Vaizdų pavyzdžiai - - Phi-4 pavyzdžiai - - [📓] [Naudokite Phi-4-multimodal vaizdams skaityti ir kodo generavimui](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 pavyzdžiai - - [📓][Phi-3-vaizdų Tekstas į tekstą](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vaizdų CLIP įterpimas](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 perdirbimas](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Vaizdinis kalbos asistentas - su Phi3-Vision ir OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision daugialygis arba daugiavaizdinis pavyzdys](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision vietinis ONNX modelis naudojant Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Meniu pagrindu Phi-3 Vision vietinis ONNX modelis naudojant Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Mąstymo ir vaizdų pavyzdžiai - - Phi-4-Mąstymas-Vaizdas-15B - - [📓] [Naudojant Phi-4-Mąstymas-Vaizdas-15B avarinių perėjimų aptikimui](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Naudojant Phi-4-Mąstymas-Vaizdas-15B matematikai](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Naudojant Phi-4-Mąstymas-Vaizdas-15B UI aptikimui](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [Phi-4-multimodal demonstracijos talpinamos Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Vizijos pavyzdžiai + - Phi-4 Pavyzdžiai + - [📓] [Naudokite Phi-4-multimodal vaizdams skaityti ir kodui generuoti](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 Pavyzdžiai + - [📓][Phi-3-vizija - vaizdo tekstas į tekstą](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vizija-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vizija CLIP įterpimas](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [DEMONSTRACIJA: Phi-3 Perdirbimas](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vizija - Vaizdinis kalbos asistentas - su Phi3-Vision ir OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vizija Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vizija OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vizijos daugiasluoksnis arba daugelio vaizdų pavyzdys](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vizija Vietinis ONNX modelis naudojant Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Meniu pagrindu Phi-3 Vizija Vietinis ONNX modelis naudojant Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Samprotavimo-Vizijos pavyzdžiai + - Phi-4-Samprotavimo-Vizija-15B + - [📓] [Naudojant Phi-4-Samprotavimo-Vizija-15B nepilnamečių priežastims nustatyti](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Naudojant Phi-4-Samprotavimo-Vizija-15B matematiniams uždaviniams](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Naudojant Phi-4-Samprotavimo-Vizija-15B UI nustatymui](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Matematikos pavyzdžiai - - Phi-4-Mini-Flash-Mąstymo-Instruktavimo pavyzdžiai [Matematikos demonstracija su Phi-4-Mini-Flash-Mąstymo-Instruktavimu](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-Mini-Flash-Samprotavimo-Instruct pavyzdžiai [Matematikos demonstracija su Phi-4-Mini-Flash-Samprotavimo-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - Garso pavyzdžiai - - Phi-4 pavyzdžiai - - [📓] [Garso transkriptų išgijimas naudojant Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - Phi-4 Pavyzdžiai + - [📓] [Garso transkriptų išskyrimas naudojant Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Phi-4-multimodal garso pavyzdys](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Phi-4-multimodal kalbos vertimo pavyzdys](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET konsolinė programa naudojant Phi-4-multimodal garsui analizuoti ir transkripcijai generuoti](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [.NET konsolės programa naudojant Phi-4-multimodal garso failui analizuoti ir transkripcijai generuoti](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - MoE pavyzdžiai - - Phi-3 / 3.5 pavyzdžiai - - [📓] [Phi-3.5 ekspertų mišiniai (MoEs) socialinės žiniasklaidos pavyzdys](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Retrieval-Augmented Generation (RAG) grandinės kūrimas su NVIDIA NIM Phi-3 MOE, Azure AI Search ir LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - MOE Pavyzdžiai + - Phi-3 / 3.5 Pavyzdžiai + - [📓] [Phi-3.5 Ekspertų mišinio modelių (MoEs) socialinių tinklų pavyzdys](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Kūriant Retrieval-Augmented Generation (RAG) sistemą su NVIDIA NIM Phi-3 MOE, Azure AI Search ir LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Funkcijų iškvietimo pavyzdžiai - - Phi-4 pavyzdžiai 🆕 - - [📓] [Funkcijų iškvietimo naudojimas su Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Funkcijų iškvietimo naudojimas kuriant daugiaprieinagius su Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Funkcijų iškvietimo naudojimas su Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Funkcijų iškvietimo naudojimas su ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Daugiamodelinis mišrinimas pavyzdžiai - - Phi-4 pavyzdžiai 🆕 - - [📓] [Phi-4-multimodal naudojimas kaip technologijų žurnalistas](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET konsolinė programa naudojant Phi-4-multimodal vaizdams analizuoti](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Phi tobulinimo pavyzdžiai - - [Tobulinimo scenarijai](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Tobulinimas vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Leiskite Phi-3 tapti pramonės ekspertu](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Phi-3 tobulinimas su AI Toolkit for VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Phi-3 tobulinimas su Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Phi-3 tobulinimas su Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Phi-3 tobulinimas su QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Phi-3 tobulinimas su Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Phi-3 tobulinimas su Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Tobulinimas su Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Tobulinimas su Microsoft Olive praktinis laboratorinis darbas](./md/03.FineTuning/olive-lab/readme.md) - - [Phi-3-vision tobulinimas su Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Phi-3 tobulinimas su Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-vision tobulinimas (oficiali parama)](./md/03.FineTuning/FineTuning_Vision.md) - - [Phi-3 tobulinimas su Kaito AKS, Azure konteineriai (oficiali palaikymas)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Phi-3 ir 3.5 Vision tobulinimas](https://github.com/2U1/Phi3-Vision-Finetune) - -- Praktinė laboratorija - - [Pažangiausių modelių tyrinėjimas: LLM, SLM, vietinė plėtra ir daugiau](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP potencialo atrakcionas: Tobulinimas su Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - Funkcijų kvietimo pavyzdžiai + - Phi-4 Pavyzdžiai 🆕 + - [📓] [Funkcijų kvietimo naudojimas su Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Funkcijų kvietimo naudojimas kuriant daugiaagentines sistemas su Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Funkcijų kvietimo naudojimas su Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Funkcijų kvietimo naudojimas su ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Daugiavaičių mišinio pavyzdžiai + - Phi-4 Pavyzdžiai 🆕 + - [📓] [Naudojant Phi-4-multimodal kaip technologijų žurnalistą](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET konsolės programa naudojant Phi-4-multimodal vaizdams analizuoti](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Phi modelių papildomas apmokymas + - [Papildomo apmokymo scenarijai](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Papildomas apmokymas prieš RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Papildomas apmokymas: leiskite Phi-3 tapti pramonės ekspertu](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Phi-3 papildomas apmokymas su AI įrankių rinkiniu VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Phi-3 papildomas apmokymas su Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Phi-3 papildomas apmokymas su Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Phi-3 papildomas apmokymas su QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Phi-3 papildomas apmokymas su Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Phi-3 papildomas apmokymas su Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Papildomas apmokymas su Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Praktinis papildomas apmokymas su Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) + - [Phi-3-vizija papildomas apmokymas su Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Phi-3 smulkus derinimas naudojant Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision smulkus derinimas (oficiali parama)](./md/03.FineTuning/FineTuning_Vision.md) + - [Phi-3 smulkus derinimas su Kaito AKS, Azure konteineriais (oficiali parama)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Phi-3 ir 3.5 Vision smulkus derinimas](https://github.com/2U1/Phi3-Vision-Finetune) + +- Praktinis užsiėmimas + - [Ištyrinėti pažangiausius modelius: LLM, SLM, vietinė plėtra ir daugiau](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Atrakinti NLP potencialą: smulkus derinimas su Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - Akademiniai moksliniai straipsniai ir publikacijos - - [Vadovėliai yra viskas, ko reikia II: phi-1.5 techninis pranešimas](https://arxiv.org/abs/2309.05463) - - [Phi-3 techninis pranešimas: itin gabus kalbos modelis tiesiog jūsų telefone](https://arxiv.org/abs/2404.14219) + - [Textbooks Are All You Need II: phi-1.5 techninis pranešimas](https://arxiv.org/abs/2309.05463) + - [Phi-3 techninis pranešimas: itin pajėgus kalbos modelis vietoje jūsų telefone](https://arxiv.org/abs/2404.14219) - [Phi-4 techninis pranešimas](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini techninis pranešimas: kompaktiški, bet galingi multimodaliniai kalbos modeliai per mišrų LoRA](https://arxiv.org/abs/2503.01743) - - [Mažų kalbos modelių optimizavimas automobilių funkcijų iškvietimui](https://arxiv.org/abs/2501.02342) - - [(Kodėl PHI) PHI-3 tobulinimas daugelio pasirinkimų klausimų atsakymams: metodologija, rezultatai ir iššūkiai](https://arxiv.org/abs/2501.01588) - - [Phi-4 sprendimo techninis pranešimas](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini sprendimo techninis pranešimas](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Phi-4-Mini techninis pranešimas: kompaktiški ir galingi multimodaliniai kalbos modeliai per Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Mažų kalbos modelių optimizavimas funkcijų kvietimams transporto priemonėse](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) PHI-3 smulkus derinimas daugialypiams klausimams atsakyti: metodologija, rezultatai ir iššūkiai](https://arxiv.org/abs/2501.01588) + - [Phi-4-argumentavimo techninis pranešimas](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-argumentavimo techninis pranešimas](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Phi modelių naudojimas ### Phi Microsoft Foundry platformoje -Galite sužinoti, kaip naudoti Microsoft Phi ir kaip kurti E2E sprendimus savo įvairiuose įrenginiuose. Norėdami patirti Phi patys, pradėkite žaisti su modeliais ir pritaikyti Phi savo scenarijams naudodami [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Daugiau sužinosite skiltyje „Pradžia su [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md)“. +Galite išmokti, kaip naudoti Microsoft Phi ir kaip kurti E2E sprendimus skirtinguose jūsų įrenginiuose. Norėdami patys išbandyti Phi, pradėkite žaisti su modeliais ir pritaikyti Phi jūsų scenarijams naudojant [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Daugiau sužinokite per Pradžiamokslį su [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Bandomoji aplinka** -Kiekvienas modelis turi specialią bandomąją aplinką, kurioje galite išbandyti modelį: [Azure AI Playground](https://aka.ms/try-phi3). +**Žaidimų aikštelė** +Kiekvienas modelis turi skirtą žaidimų aikštelę modeliui išbandyti [Azure AI Playground](https://aka.ms/try-phi3). ### Phi GitHub modeliuose -Galite sužinoti, kaip naudoti Microsoft Phi ir kurti E2E sprendimus savo įvairiuose įrenginiuose. Norėdami patirti Phi patys, pradėkite žaisti su modeliu ir pritaikyti Phi savo scenarijams naudodami [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Daugiau sužinosite skiltyje „Pradžia su [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md)“. +Galite išmokti, kaip naudoti Microsoft Phi ir kaip kurti E2E sprendimus skirtinguose jūsų įrenginiuose. Norėdami patys išbandyti Phi, pradėkite žaisti su modeliu ir pritaikyti Phi jūsų scenarijams naudojant [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Daugiau sužinokite per Pradžiamokslį su [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Bandomoji aplinka** -Kiekvienas modelis turi specialią [bandomąją aplinką modelio testavimui](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Žaidimų aikštelė** +Kiekvienas modelis turi skirtą [žaidimų aikštelę modeliui išbandyti](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi Hugging Face platformoje -Modelį taip pat galite rasti [Hugging Face](https://huggingface.co/microsoft) svetainėje. +Modelį taip pat galite rasti [Hugging Face](https://huggingface.co/microsoft) -**Bandomoji aplinka** -[Hugging Chat bandomoji aplinka](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Žaidimų aikštelė** + [Hugging Chat žaidimų aikštelė](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Kiti kursai + ## 🎒 Kiti kursai -Mūsų komanda kuria ir kitus kursus! Peržiūrėkite: +Mūsų komanda rengia kitus kursus! Pažiūrėkite: -### LangChain -[![LangChain4j pradedantiesiems](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js pradedantiesiems](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain pradedantiesiems](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +### LangChain +[![LangChain4j pradedantiesiems](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js pradedantiesiems](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain pradedantiesiems](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agentai -[![AZD pradedantiesiems](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI pradedantiesiems](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP pradedantiesiems](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI agentai pradedantiesiems](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agentai +[![AZD pradedantiesiems](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI pradedantiesiems](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP pradedantiesiems](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI agentai pradedantiesiems](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Generatyvinis DI serija -[![Generatyvinis DI pradedantiesiems](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generatyvinis DI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generatyvinis DI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generatyvinis DI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### Generatyvinio AI serija +[![Generatyvinis AI pradedantiesiems](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generatyvinis AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generatyvinis AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generatyvinis AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Pagrindinis mokymasis -[![ML pradedantiesiems](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Duomenų mokslas pradedantiesiems](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![DI pradedantiesiems](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Kibernetinis saugumas pradedantiesiems](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Tinklapių kūrimas pradedantiesiems](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT pradedantiesiems](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR plėtra pradedantiesiems](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) + +### Pagrindinis mokymasis +[![ML pradedantiesiems](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Duomenų mokslas pradedantiesiems](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI pradedantiesiems](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Kibernetinis saugumas pradedantiesiems](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Tinklalapių kūrimas pradedantiesiems](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT pradedantiesiems](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR kūrimas pradedantiesiems](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Copilot serija -[![Copilot AI suporuotam programavimui](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot nuotykiai](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + +### Copilot serija +[![Copilot dirbančiam poroje su AI programavime](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot C#/.NET programuotojams](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot nuotykis](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Atsakingas DI +## Atsakingas dirbtinis intelektas + +Microsoft siekia padėti savo klientams atsakingai naudoti mūsų DI produktus, dalintis savo patirtimi ir kurti pasitikėjimu pagrįstas partnerystes per priemones, tokias kaip Transparency Notes ir Impact Assessments. Daug šių išteklių galima rasti adresu [https://aka.ms/RAI](https://aka.ms/RAI). +Microsoft atsakingo DI požiūris remiasi mūsų DI principais: sąžiningumu, patikimumu ir saugumu, privatumu ir saugumu, įtrauktimi, skaidrumu bei atskaitomybe. + +Didelio masto natūralios kalbos, vaizdų ir balso modeliai – kaip šio pavyzdžio atveju – gali elgtis nesąžiningai, nepatikimai ar įžeidžiančiai, kas gali sukelti žalos. Prašome pasitikrinti [Azure OpenAI paslaugos skaidrumo pastabą](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), kad būtumėte informuoti apie rizikas ir apribojimus. -Microsoft įsipareigoja padėti savo klientams atsakingai naudoti mūsų DI produktus, dalintis mūsų patirtimi ir kurti pasitikėjimu grįstus partnerystes per įrankius, tokius kaip Skaidrumo Užrašai ir Poveikio Vertinimai. Daugelį šių išteklių rasite adresu [https://aka.ms/RAI](https://aka.ms/RAI). -Microsoft atsakingo DI požiūris grindžiamas mūsų DI principais: teisingumo, patikimumo ir saugumo, privatumo ir saugumo, įtraukties, skaidrumo ir atsakomybės. -Didelio masto natūralios kalbos, vaizdų ir balso modeliai – tokie, kokie naudojami šiame pavyzdyje – gali elgtis neteisingai, nepatikimai arba galimai įžeidžiančiai, sukeldami žalą. Prašome susipažinti su [Azure OpenAI paslaugos Skaidrumo užrašu](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), kad būtumėte informuoti apie rizikas ir apribojimus. -Rekomenduojamas šių rizikų mažinimo būdas yra įtraukti saugumo sistemą į jūsų architektūrą, kuri gali aptikti ir užkirsti kelią žalingam elgesiui. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) suteikia nepriklausomą apsaugos sluoksnį, gebantį aptikti žalingą naudotojų ir dirbtinio intelekto generuojamą turinį programose ir paslaugose. Azure AI Content Safety apima teksto ir vaizdų API, leidžiančias aptikti žalingą medžiagą. Microsoft Foundry aplinkoje Content Safety paslauga leidžia peržiūrėti, tyrinėti ir išbandyti pavyzdinį kodą, skirtą žalingo turinio aptikimui įvairiose modalitetuose. Toliau pateikta [greito starto dokumentacija](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) padės jums atlikti užklausas šiai paslaugai. +Rekomenduojamas būdas mažinti šias rizikas yra įtraukti saugumo sistemą į savo architektūrą, kuri galėtų aptikti ir užkirsti kelią žalingam elgesiui. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) teikia nepriklausomą apsaugos sluoksnį, gebantį aptikti žalingą vartotojų ar dirbtinio intelekto sukurtą turinį programėlėse ir paslaugose. Azure AI Content Safety apima teksto ir vaizdų API, kurios leidžia aptikti žalingą medžiagą. Microsoft Foundry aplinkoje Content Safety paslauga leidžia peržiūrėti, tyrinėti ir išbandyti pavyzdinį kodą, skirtą žalingo turinio aptikimui skirtingose modalumuose. Ši [greito pradžios dokumentacija](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) padės jums atlikti užklausas šiai paslaugai. -Kitas aspektas, kurį reikia apsvarstyti, yra bendras programos našumas. Su daugiapločiais ir daugiamodeliais sprendimais, pagal našumą suprantame, kad sistema veikia taip, kaip tikitės jūs ir jūsų naudotojai, įskaitant ir nesugeneruojant kenksmingų rezultatų. Svarbu įvertinti bendrą programos našumą naudojant [Veiklos ir kokybės bei rizikos ir saugos vertinimo priemones](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Taip pat turite galimybę kurti ir vertinti naudodami [individualias vertinimo priemones](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Kitas svarbus aspektas yra bendra programėlės veikimo sparta. Daugiamodalėse ir daugimodelinėse programėlėse veikimo sparta reiškia, kad sistema veikia taip, kaip tikitės jūs ir jūsų vartotojai, įskaitant tai, jog nesukuria žalingo turinio. Svarbu įvertinti visos programėlės veikimą naudojant [Veikimo ir Kokybės bei Rizikos ir Saugumo vertintojus](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Taip pat turite galimybę kurti ir vertinti naudojant [pasirinktinius vertintojus](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Galite įvertinti savo DI programą kūrimo aplinkoje naudodami [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Turėdami testinį duomenų rinkinį arba tikslą, jūsų generuojami DI rezultatai kiekybiškai matuojami naudojant įmontuotas arba pasirinktas individualias vertinimo priemones. Norėdami pradėti naudotis azure ai evaluation sdk savo sistemos vertinimui, galite sekti [greito starto vadovą](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Užbaigę vertinimo ciklą, galite [vizualizuoti rezultatus Microsoft Foundry aplinkoje](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Galite vertinti savo DI programėlę kūrimo aplinkoje naudodami [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Turint testų duomenų rinkinį arba tikslą, jūsų generatyvios DI programėlės generacijos yra kiekybiškai matuojamos naudojant įmontuotus arba pasirinktinius vertintojus pagal jūsų pasirinkimą. Norėdami pradėti naudotis Azure AI Evaluation SDK savo sistemos vertinimui, galite sekti [greito pradžios vadovą](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Kai atliksite vertinimo vykdymą, [Microsoft Foundry aplinkoje galite vizualizuoti rezultatus](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Prekių ženklai -Šiame projekte gali būti prekių ženklai arba logotipai projektams, produktams ar paslaugoms. Leidžiama naudoti Microsoft prekių ženklus ar logotipus tik laikantis ir vadovaujantis [Microsoft prekių ženklų ir prekės ženklo gairėmis](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Modifikuotose šio projekto versijose Microsoft prekių ženklų ar logotipų naudojimas neturi sukelti painiavos ar nurodyti Microsoft rėmimą. Bet koks trečiųjų šalių prekių ženklų ar logotipų naudojimas yra reglamentuojamas atitinkamų trečiųjų šalių taisyklių. +Šiame projekte gali būti naudojami prekių ženklai arba logotipai, susiję su projektais, produktais ar paslaugomis. Leidžiamas Microsoft prekių ženklų ar logotipų naudojimas turi atitikti ir laikytis [Microsoft prekių ženklų ir prekės ženklo gairių](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Microsoft prekių ženklų ar logotipų naudojimas šio projekto modifikuotose versijose neturi sukelti painiavos ar reikšti Microsoft rėmimą. Bet koks trečiųjų šalių prekių ženklų ar logotipų naudojimas yra reglamentuojamas tų trečiųjų šalių politikos. ## Pagalba -Jei sutriktumėte arba turite klausimų dėl DI programų kūrimo, prisijunkite prie: +Jei susiduriate su sunkumais ar turite klausimų apie DI programėlių kūrimą, prisijunkite prie: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Jei turite produktų atsiliepimų arba aptikote klaidų kūrimo metu, apsilankykite: +Jei turite produktų atsiliepimų ar pastebite klaidų kūrimo metu, apsilankykite: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) @@ -320,5 +325,5 @@ Jei turite produktų atsiliepimų arba aptikote klaidų kūrimo metu, apsilankyk **Atsakomybės apribojimas**: -Šis dokumentas buvo išverstas naudojant dirbtinio intelekto vertimo paslaugą [Co-op Translator](https://github.com/Azure/co-op-translator). Nors siekiame tikslumo, prašome atkreipti dėmesį, kad automatizuoti vertimai gali turėti klaidų ar netikslumų. Originalus dokumentas jo gimtąja kalba laikomas autoritetingu šaltiniu. Svarbiai informacijai rekomenduojama naudoti profesionalų žmogaus vertimą. Mes neatsakome už jokius nesusipratimus ar netinkamus aiškinimus, kilusius naudojantis šiuo vertimu. +Šis dokumentas buvo išverstas naudojant dirbtinio intelekto vertimo paslaugą [Co-op Translator](https://github.com/Azure/co-op-translator). Nors siekiame tikslumo, prašome atkreipti dėmesį, kad automatiniai vertimai gali turėti klaidų ar netikslumų. Originalus dokumentas jo gimtąja kalba laikomas autoritetingu šaltiniu. Svarbiai informacijai rekomenduojama naudoti profesionalų žmogiškąjį vertimą. Mes neatsakome už jokius nesusipratimus ar neteisingą interpretaciją, kilusią naudojantis šiuo vertimu. \ No newline at end of file diff --git a/translations/lt/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/lt/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..8d381569c --- /dev/null +++ b/translations/lt/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hibridinis RAG su Microsoft phi-4-mini ir SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Autorius:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Dėmesys:** Maži kalbos modeliai (SLM), SQLite FTS5 BM25, tankios įterptinės reprezentacijos, atvirkštinė reitingo fuzija (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motyvacija: Raktažodžių atsiminimo dilema vietiniuose SLM\n", + "Standartinės RAG architektūros, kurios remiasi tik tankiomis vektorinėmis įterptinėmis reprezentacijomis, dažnai nepajėgia atkurti tikslių skaitinių žetonų (pvz., `2,340,000 TL`, sutarties kodai, sąskaitų numeriai). \n", + "Priešingai, retas leksinis paieškos būdas (BM25) nepastebi semantinių sinonimų ir perrašytų klausimų.\n", + "\n", + "Ši praktinė knyga demonstruoja, kaip įgyvendinti **aukštos spartos, atmintyje veikiantį hibridinį paieškos variklį**, jungiantį:\n", + "1. **Tankius vektorius** (kosininis panašumas)\n", + "2. **Retą leksinę paiešką** (SQLite FTS5 BM25)\n", + "3. **Atvirkštinę reitingo fuziją (RRF, $k=60$)**\n", + "4. **Pagrįstą citatų generavimą (`[1]`, `[2]`)** su Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Dvigubas SQLite schemos (tankūs vektoriai + virtuali FTS5 BM25 lentelė)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Pavyzdžio įkėlimo ir vykdymo etalonas\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Grindžiamų užklausų formavimas Microsoft phi-4-mini modeliui\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Atsakomybės apribojimas**:\nŠis dokumentas buvo išverstas naudojant dirbtinio intelekto vertimo paslaugą [Co-op Translator](https://github.com/Azure/co-op-translator). Nors siekiame tikslumo, prašome atkreipti dėmesį, kad automatiniai vertimai gali turėti klaidų ar netikslumų. Originalus dokumentas jo gimtąja kalba laikomas autoritetingu šaltiniu. Svarbiai informacijai rekomenduojama naudoti profesionalų žmogiškąjį vertimą. Mes neatsakome už jokius nesusipratimus ar neteisingą interpretaciją, kilusią naudojantis šiuo vertimu.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/ml/.co-op-translator.json b/translations/ml/.co-op-translator.json index 5a66e4a83..705fb36c2 100644 --- a/translations/ml/.co-op-translator.json +++ b/translations/ml/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "ml" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T08:06:37+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:00:41+00:00", "source_file": "README.md", "language_code": "ml" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "ml" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:04:45+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "ml" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-12-21T15:51:00+00:00", diff --git a/translations/ml/README.md b/translations/ml/README.md index 5e341b4da..687e6557a 100644 --- a/translations/ml/README.md +++ b/translations/ml/README.md @@ -1,42 +1,42 @@ -# ഫി കുക്ക്ബുക്ക്: Microsoft-യുടെ Phi മോഡലുകളുമായുള്ള ഹാൻഡ്‌സോൺ ഉദാഹരണങ്ങൾ +# Phi പാചക പുസ്‌തകം: Microsoft-യുടെ Phi മോഡലുകളുമായി കൈകാര്യം ചെയ്യുന്ന ഉദാഹരണങ്ങൾ -[![GitHub Codespaces-ൽ സാമ്പിളുകൾ തുറന്ന് ഉപയോഗിക്കുക](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![GitHub Codespaces-ൽ സാമ്പിൾ ഒപ്പുവെച്ച് ഉപയോഗിക്കുക](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Dev Containers-ൽ തുറക്കുക](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub സംഭാവകർ](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub പ്രശ്നങ്ങൾ](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub പുൾ അഭ്യർത്ഥനകൾ](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs സ്വാഗതം ചെയ്യുന്നു](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub പുൾ-റിക്വസ്റ്റ്](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![പുൾ-റിക്വെസ്റ്റുകൾ സ്വാഗതം](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub വാച്ചേഴ്സ്](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub വാച്ചർമാർ](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub ഫോർക്കുകൾ](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub സ്റ്റാർസ്](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi Microsoft വികസിപ്പിച്ച ഒരു ഓപ്പൺ സോഴ്‌സ് AI മോഡലുകളുടെ പരമ്പരയാണ്. +Phi Microsoft വികസിപ്പിച്ചിട്ടുള്ള ഓപ്പൺ സോഴ്സ് AI മോഡലുകളുടെ ഒരു പരമ്പരയാണ്. -Phi നിലവിൽ വളരെ ശക്തവും ചെലവുകുറഞ്ഞതുമായ ചെറിയ ഭാഷാ മോഡലാണ് (SLM), മൾട്ടി-ഭാഷ, ബോധവത്കരണം, എഴുത്ത്/ചാറ്റ് üret generation, കോഡിംഗ്, ചിത്രം, ഓഡിയോ എന്നിവയടക്കമുള്ള വിവിധ പ്രയോഗങ്ങളിൽ മികച്ച പ്രകടനം കാഴ്ചവെക്കുന്നു. +Phi നിലവിൽ ഏറ്റവും ശക്തിയും ചെലവ് കാര്യക്ഷമതയുമായി ബന്ധപ്പെട്ടു ചെറിയ ഭാഷാ മോഡലുകളിലേതാണ് (SLM), ബഹുഭാഷാ, ആയോജനം, ടെക്സ്റ്റ്/ചാറ്റ് നിർമ്മാണം, കോഡിംഗ്, ചിത്രങ്ങൾ, ഓഡിയോ എന്നിവയിൽ മികച്ച ബെൻച്മാർക്കുകൾ കൈവരിച്ചിരിക്കുന്നു. -നിങ്ങൾ Phi ക്ലൗഡിലും എഡ്ജ് ഉപകരണങ്ങളിലും മൊരിയ്ക്കാം, കൂടാതെ കുറഞ്ഞ കംപ്യൂട്ടിംഗ് ശേഷിയുള്ള ജനന AI അപേക്ഷകൾ आसानीയോടെ നിർമ്മിക്കാം. +നിങ്ങൾ Phi ക്ലൗഡിലോ എഡ്ജ് ഉപകരണങ്ങളിലോ വിന്യസിക്കാനാകും, കൂടാതെ പരിമിത കംപ്യൂട്ടിങ്ങ് ശേഷിയുള്ളതിനും വളരെ എളുപ്പത്തിൽ ജനറേറ്റീവ് AI ആപ്ലിക്കേഷനുകൾ നിർമ്മിക്കാനാകും. -ഈ ഘട്ടങ്ങൾ പിന്തുടർന്ന് ഈ സ്രോതസ്സുകൾ ഉപയോഗിച്ച് തുടക്കം കുറിക്കുക: -1. **റിപ്പോസിറ്ററി ഫോർക്കുചെയ്യുക**: ക്ലിക്ക് ചെയ്യുക [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **റിപ്പോസിറ്ററി ക്ലോൺ ചെയ്യുക**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Microsoft AI Discord കമ്മ്യൂണിറ്റിയിൽ ചേരുക, വിദഗ്ദ്ധരെയും സഹപ്രവർത്തകരെയും കാണുക**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +ഈ റിസോഴ്സുകൾ ഉപയോഗിക്കാൻ തുടങ്ങുന്നതിന് താഴെ കൊടുത്തിട്ടുള്ള ചുവടുകൾ പിന്തുടരുക: +1. **റിപ്പോസിറ്ററി ഫോർക്കുചെയ്യുക**: [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) ക്ലിക്ക് ചെയ്യുക +2. **റിപ്പോസിറ്ററി ക്ലോണുചെയ്യുക**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Microsoft AI Discord സമൂഹത്തിൽ ചേർന്ന് വിദഗ്ധരുമായും മറ്റ് ഡെവലപ്പർമാരുമായും കണ്ടുമുട്ടുക**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/ml/cover.eb18d1b9605d754b.webp) -### 🌐 മൾട്ടി-ഭാഷാ പിന്തുണ +### 🌐 ബഹുഭാഷാ പിന്തുണ -#### GitHub Action വഴി പിന്തുണ (തത്സമയം & എല്ലായ്പ്പോഴും പുതുക്കപ്പെട്ടത്) +#### GitHub ആക്ഷൻ വഴിയുള്ള പിന്തുണ (സ്വയമേവ പ്രവർത്തിക്കുന്നതും എല്ലായ്പ്പോഴും പുതുക്കപ്പെടുന്നതും) [Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](./README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **സ്ഥലീയമായി ക്ലോൺ ചെയ്യാൻ ഇഷ്ടപ്പെടുന്നുണ്ടോ?** +> **പ്രാദേശികമായി ക്ലോൺ ചെയ്യണമോ?** > -> ഈ റിപോസിറ്ററിയിൽ 50-ലധികം ഭാഷാ പരിഭാഷകൾ ഉൾപ്പെടുത്തിയിട്ടുണ്ട്, ഇത് ഡൗൺലോഡ് വലുപ്പം ഗണ്യമായി വർദ്ധിപ്പിക്കുന്നു. പരിഭാഷകൾ ഇല്ലാതെ ക്ലോൺ ചെയ്യാൻ sparse checkout ഉപയോഗിക്കുക: +> ഈ റിപ്പോസിറ്ററിയിൽ 50-ലധികം ഭാഷാ പരിഭാഷകൾ ഉള്ളതിനാൽ ഡൗൺലോഡ് സൈസ് വളരെ വലുതാണ്. പരിഭാഷകളില്ലാതെ ക്ലോൺ ചെയ്യാൻ sparse checkout ഉപയോഗിക്കുക: > > **Bash / macOS / Linux:** > ```bash @@ -52,270 +52,278 @@ Phi നിലവിൽ വളരെ ശക്തവും ചെലവുകു > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> ഇത് കോഴ്‌സ് പൂര്‍ത്തിയാക്കാൻ ആവശ്യമായ എല്ലാ ഘടകങ്ങളും വേഗത്തിൽ ഡൗൺലോഡ് ചെയ്യാൻ സഹായിക്കും. +> ഇത് നിങ്ങൾക്ക് വളരെ വേഗത്തിലുള്ള ഡൗൺലോഡോടെ കോഴ്‌സ് പൂർത്തിയാക്കാൻ ആവശ്യമുള്ള എല്ലാ ഫയലുകളും നൽകും. ## ഉള്ളടക്ക പട്ടിക -- പരിചയപ്പെടുത്തൽ +- പരിചയം - [Phi കുടുംബത്തിലേക്ക് സ്വാഗതം](./md/01.Introduction/01/01.PhiFamily.md) - [നിങ്ങളുടെ പരിസ്ഥിതി ക്രമീകരിക്കൽ](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [പ്രധാന സാങ്കേതികവിദ്യകൾ മനസ്സിലാക്കുക](./md/01.Introduction/01/01.Understandingtech.md) + - [പ്രധാന സാങ്കേതികവിദ്യകൾ മനസിലാക്കുക](./md/01.Introduction/01/01.Understandingtech.md) - [Phi മോഡലുകൾക്കുള്ള AI സുരക്ഷ](./md/01.Introduction/01/01.AISafety.md) - - [Phi ഹാർഡ്വെയർ പിന്തുണ](./md/01.Introduction/01/01.Hardwaresupport.md) - - [വിവിധ പ്ലാറ്റ്ഫോമുകളിൽ Phi മോഡലുകളും ലഭ്യതയും](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Guidance-ai ഉം Phi ഉം ഉപയോഗിക്കുക](./md/01.Introduction/01/01.Guidance.md) - - [GitHub മാർക്കറ്റ്‌പ്ലേസ് മോഡലുകൾ](https://github.com/marketplace/models) + - [Phi ഹാർഡ്‌വെയർ പിന്തുണ](./md/01.Introduction/01/01.Hardwaresupport.md) + - [ഫ്ലാറ്റ്ഫോം വിപുലമായി Phi മോഡലുകളും ലഭ്യതയും](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Guidance-ai և Phi ഉപയോഗിക്കൽ](./md/01.Introduction/01/01.Guidance.md) + - [GitHub മാർക്കറ്റ് പ്ലേസ് മോഡലുകൾ](https://github.com/marketplace/models) - [Azure AI മോഡൽ കാറ്റലോഗ്](https://ai.azure.com) -- വ്യത്യസ്ത പരിസ്ഥിതികളിൽ Phi ഇൻഫറൻസ് +- വ്യത്യസ്ത പരിസ്ഥിതികളിൽ Phi-ന്റെ ഇൻഫെറൻസ് - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub മോഡലുകൾ](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry Model Catalog](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Microsoft Foundry മോഡൽ കാറ്റലോഗ്](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI ടൂൾകിറ്റ് VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry ലോക്കൽ](./md/01.Introduction/02/07.FoundryLocal.md) - -- Phi കുടുംബത്തിന്റെ ഇൻഫറൻസ് - - [iOS-ൽ Phi ഇൻഫറൻസ്](./md/01.Introduction/03/iOS_Inference.md) - - [Android-ൽ Phi ഇൻഫറൻസ്](./md/01.Introduction/03/Android_Inference.md) - - [Jetson-ൽ Phi ഇൻഫറൻസ്](./md/01.Introduction/03/Jetson_Inference.md) - - [AI PC-ൽ Phi ഇൻഫറൻസ്](./md/01.Introduction/03/AIPC_Inference.md) - - [Apple MLX Framework ഉപയോഗിച്ച് Phi ഇൻഫറൻസ്](./md/01.Introduction/03/MLX_Inference.md) - - [ലോക്കൽ സെർവറിൽ Phi ഇൻഫറൻസ്](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI ടൂൾകിറ്റ് ഉപയോഗിച്ച് റിമോട്ട് സെർവറിൽ Phi ഇൻഫറൻസ്](./md/01.Introduction/03/Remote_Interence.md) - - [Rust ഉപയോഗിച്ച് Phi ഇൻഫറൻസ്](./md/01.Introduction/03/Rust_Inference.md) - - [ലോക്കലിൽ Phi-യുടെ വിഷൻ ഇൻഫറൻസ്](./md/01.Introduction/03/Vision_Inference.md) - - [Kaito AKS, Azure Containers (അധികൃത പിന്തുണ) ഉപയോഗിച്ച് Phi ഇൻഫറൻസ്](./md/01.Introduction/03/Kaito_Inference.md) -- [Phi കുടുംബത്തിന്റെ ക്വാണ്ടിഫിക്കേഷൻ](./md/01.Introduction/04/QuantifyingPhi.md) - - [llama.cpp ഉപയോഗിച്ച് Phi-3.5 / 4 ക്വാണ്ടൈസിംഗ്](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime-നുള്ള Generative AI വിപുലീകരണങ്ങൾ ഉപയോഗിച്ച് Phi-3.5 / 4 ക്വാണ്ടൈസിംഗ്](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Intel OpenVINO ഉപയോഗിച്ച് Phi-3.5 / 4 ക്വാണ്ടൈസിംഗ്](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Apple MLX Framework ഉപയോഗിച്ച് Phi-3.5 / 4 ക്വാണ്ടൈസിംഗ്](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Phi மதிப்பீடு - - [ഉത്തരം നൽകുന്ന AI](./md/01.Introduction/05/ResponsibleAI.md) - - [മൈക്രോസോഫ്റ്റ് ഫൗണ്ട്രി വിലയിരുത്തലിനായി](./md/01.Introduction/05/AIFoundry.md) - - [Promptflow ഉപയോഗിച്ച് വിലയിരുത്തൽ](./md/01.Introduction/05/Promptflow.md) + - [Foundry ലൊക്കൽ](./md/01.Introduction/02/07.FoundryLocal.md) + +- Phi കുടുംബത്തിലെ ഇൻഫെറൻസ് + - [iOS-ൽ ഇൻഫെറൻസ് Phi](./md/01.Introduction/03/iOS_Inference.md) + - [Android-ൽ ഇൻഫെറൻസ് Phi](./md/01.Introduction/03/Android_Inference.md) + - [Jetson-ൽ ഇൻഫെറൻസ് Phi](./md/01.Introduction/03/Jetson_Inference.md) + - [AI PC-യിൽ ഇൻഫെറൻസ് Phi](./md/01.Introduction/03/AIPC_Inference.md) + - [Apple MLX ഫ്രെയിംവർക്ക് ഉപയോഗിച്ച് Phi ഇൻഫെറൻസ്](./md/01.Introduction/03/MLX_Inference.md) + - [ലൊക്കൽ സെർവറിൽ ഇൻഫെറൻസ് Phi](./md/01.Introduction/03/Local_Server_Inference.md) + - [AI ടൂൾകിറ്റ് ഉപയോഗിച്ച് റിമോട്ട് സെർവറിൽ ഇൻഫെറൻസ് Phi](./md/01.Introduction/03/Remote_Interence.md) + - [Rust ഉപയോഗിച്ച് ഇൻഫെറൻസ് Phi](./md/01.Introduction/03/Rust_Inference.md) + - [ലൊക്കലിൽ ഇൻഫെറൻസ് Phi--Vision](./md/01.Introduction/03/Vision_Inference.md) + - [Kaito AKS, Azure Containers (ഓഫിഷ്യൽ പിന്തുണ) ഉപയോഗിച്ച് ഇൻഫെറൻസ് Phi](./md/01.Introduction/03/Kaito_Inference.md) +- [Phi കുടുംബം ക്വാണ്ടിഫൈ ചെയ്ത് പരിശോധിക്കൽ](./md/01.Introduction/04/QuantifyingPhi.md) + - [llama.cpp ഉപയോഗിച്ച് Phi-3.5 / 4 ക്വാണ്ടൈസ് ചെയ്യൽ](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntime-നുള്ള ജനറേറ്റീവ് AI എക്സ്റ്റൻസ് വഴി Phi-3.5 / 4 ക്വാണ്ടൈസ് ചെയ്യൽ](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Intel OpenVINO ഉപയോഗിച്ച് Phi-3.5 / 4 ക്വാണ്ടൈസ് ചെയ്യൽ](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Apple MLX ഫ്രെയിംവർക്ക് ഉപയോഗിച്ച് Phi-3.5 / 4 ക്വാണ്ടൈസ് ചെയ്യൽ](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- Phi-ന്റെ മൂല്യനിർണയം + - [Responsibile AI](./md/01.Introduction/05/ResponsibleAI.md) + - [Microsoft Foundry മൂല്യനിർണയത്തിന്](./md/01.Introduction/05/AIFoundry.md) + - [Promptflow ഉപയോഗിച്ച് മൂല്യനിർണയം](./md/01.Introduction/05/Promptflow.md) -- Azure AI Search ഉപയോഗിച്ച് RAG - - [Phi-4-mini, Phi-4-multimodal(RAG) Azure AI Search-ലുമായി എങ്ങനെ ഉപയോഗിക്കാമെന്നത്](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- Azure AI സെർച്ച് ഉപയോഗിച്ച് RAG + - [Phi-4-miniയും Phi-4-multimodal (RAG)യും Azure AI സെർച്ച് ഉപയോഗിച്ച് എങ്ങനെ ഉപയോഗിക്കാമെന്ന്](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud ലൊക്കൽ ഹൈബ്രിഡ് RAG SQLite FTS5 സജ്ജീകരിച്ച് phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi ആപ്ലിക്കേഷൻ ഡവലപ്പ്മെന്റ് സാമ്പിളുകൾ - - എഴുത്ത് & ചാറ്റ് ആപ്ലിക്കേഷനുകൾ +- Phi ആപ്ലിക്കേഷൻ ഡെവലപ്മെന്റ് സാമ്പിളുകൾ + - ടെക്സ്റ്റ് & ചാറ്റ് ആപ്ലിക്കേഷനുകൾ - Phi-4 സാമ്പിളുകൾ - [📓] [Phi-4-mini ONNX മോഡലുമായി ചാറ്റ്](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 ലോക്കൽ ONNX മോഡലുമായി ചാറ്റ് .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [സെമെന്റിക് കെർണൽ ഉപയോഗിച്ച് Phi-4 ONNX ഉപയോഗിക്കുന്ന ചാറ്റ് .NET കൺസോൾ ആപ്പ്](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - [Phi-4 ലൊക്കൽ ONNX മോഡലുമായി ചാറ്റ് .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Sementic Kernel ഉപയോഗിച്ച് Phi-4 ONNX-ഉള്ള .NET കൺസോൾ ആപ്പ്](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 സാമ്പിളുകൾ - - [Phi3, ONNX Runtime Web, WebGPU ഉപയോഗിച്ച് ബ്രൗസറായി ലോക്കൽ ചാറ്റ്‌ബോട്ടോ](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Phi3, ONNX Runtime Web, WebGPU ഉപയോഗിച്ച് ബ്രൗസറിൽ ലോക്കൽ ചാറ്റ്ബോട്ട്](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino ചാറ്റ്](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [മൾട്ടി മോഡൽ - ഇൻററാക്ടീവ് Phi-3-മിനി ആൻഡ് OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - ഒരു റാപ്പർ നിർമ്മിക്കുകയും Phi-3 MLFlow ഉപയോഗിച്ച് ഉപയോഗിക്കുകയും ചെയ്യുന്നത്](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [മോഡൽsgiving - ONNX റൺടൈം വെബിനായി Phi-3-മിനി മോഡൽ ഒപ്റ്റിമൈസ് ചെയ്യുന്നത് Olive ഉപയോഗിച്ച് എങ്ങനെ](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 ആപ്പ് Phi-3 മിനി-4k-ഇൻസ്ട്രക്ട്-onnx ഉളളത്](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 മൾട്ടി മോഡൽ AI പവേർഡ് നോട്ട്സ് ആപ്പ് സാമ്പിൾ](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [കസ്റ്റം Phi-3 മോഡലുകൾ ഫൈൻ-ട്യൂൺ ചെയ്ത് പ്രോമ്പ്റ്റ് ഫ്ലോവിനൊപ്പം എങ്ങനെയാണ് സംയോജിപ്പിക്കുന്നത്](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [മൈക്രോസോഫ്റ്റ് ഫൗണ്ടറിയിൽ പ്രോമ്പ്റ്റ് ഫ്ലോവിനൊപ്പം കസ്റ്റം Phi-3 മോഡലുകൾ ഫൈൻ-ട്യൂൺ ചെയ്ത് സംയോജിപ്പിക്കുന്നത്](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [മൈക്രോസോഫ്റ്റിന്റെ ഉത്തരവാദിത്വമുള്ള AI സിദ്ധാന്തങ്ങളിൽ പോടി വെച്ച് ഫൈൻ-ട്യൂൺ ചെയ്ത Phi-3 / Phi-3.5 മോഡൽ മൈക്രോസോഫ്റ്റ് ഫൗണ്ടറിയിൽ ഫലപ്രദമായി പരിശോധിക്കുക](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-മിനി-ഇൻസ്ട്രക്ട് ലാംഗ്വേജ് പ്രവചന സാമ്പിൾ (ചൈനീസ്/ഇംഗ്ലീഷ്)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-ഇൻസ്ട്രക്ട് WebGPU RAG ചാറ്റ്‌ബോട്ട്](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Phi-3.5-ഇൻസ്ട്രക്ട് ONNX ഉപയോഗിച്ച് പ്രോമ്പ്റ്റ് ഫ്ലോ സൊല്യൂഷൻ നിർമ്മിക്കുന്നതിനായി Windows GPU ഉപയോഗിക്കുന്നത്](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [മൈക്രോസോഫ്റ്റ് Phi-3.5 tflite ഉപയോഗിച്ച് ആൻഡ്രോയ്ഡ് ആപ്പ് നിർമ്മിക്കുക](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [ക്യു & എ .NET ഉദാഹരണം കണിക ONNX Phi-3 മോഡൽ ഉപയോഗിച്ച് Microsoft.ML.OnnxRuntime ഉപയോഗിക്കുന്നത്](../../md/04.HOL/dotnet/src/LabsPhi301) - - [സെമാന്റിക് കർണൽ Phi-3 ഉപയോഗിച്ച് കൺസോൾ ചാറ്റ് .NET ആപ്പ്](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI ഇൻഫറൻസ് SDK കോഡ് അടിസ്ഥാനമാക്കിയ സാമ്പിളുകൾ - - Phi-4 സാമ്പിളുകൾ - - [📓] [Phi-4-മൾട്ടിമോഡൽ ഉപയോഗിച്ച് പ്രോജക്ട് കോഡ് സൃഷ്ടിക്കുക](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [മൾട്ടി മോഡൽ - ഇന്ററാക്ടീവ് Phi-3-mini, OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - വ്രാപ്പർ നിർമ്മിക്കൽ, Phi-3 MLFlow ഉപയോഗിച്ച്](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [മോഡൽ ഓപ്റ്റിമൈസേഷൻ - ONNX Runtime Web നിൽക്കാൻ Phi-3-mini മോഡൽ ഓപ്റ്റിമൈസ് ചെയ്യുക Olive ഉപയോഗിച്ച്](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 ആപ്പ് Phi-3 mini-4k-instruct-onnx ഉപയോഗിച്ച്](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 മൾട്ടി മോഡൽ AI പ്രചോദിത നോട്ട്സ് ആപ്പ് സാമ്പിൾ](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Fine-tune ചെയ്ത് Prompt flow ഉപയോഗിച്ച് കസ്റ്റം Phi-3 മോഡലുകൾ ഒരു പൂർണ്ണ സംയോജനം](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Microsoft Foundry-ൽ Prompt flow ഉപയോഗിച്ച് കസ്റ്റം Phi-3 മോഡലുകൾ Fine-tune ചെയ്ത് സംയോജനം](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Microsoft Foundry-ൽ Phi-3 / Phi-3.5 മോഡലിന്റെ ഫൈൻ-ട്യൂണിങ് വിലയിരുത്തൽ Responsible AI ചട്ടങ്ങൾക്കൊപ്പം](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct ഭാഷാ പ്രവചനം സാമ്പിൾ (ചൈനീസ്/ഇംഗ്ലീഷ്)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG ചാറ്റ്ബോട്ട്](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Windows GPU ഉപയോഗിച്ച് Phi-3.5-Instruct ONNX പ്രോമ്പ്റ്റ് ഫ്ലോ സൊല്യൂഷൻ നിർമ്മിക്കൽ](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Microsoft Phi-3.5 tflite ഉപയോഗിച്ച് ആൻഡ്രോയിഡ് ആപ്പ് സൃഷ്ടിക്കൽ](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Microsoft.ML.OnnxRuntime ഉപയോഗിച്ച് ലോക്കൽ ONNX Phi-3 മോഡൽ ഉപയോഗിക്കുന്ന Q&A .NET ഉദാഹരണം](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Semantic Kernel, Phi-3 ഉപയോഗിച്ച് കൺസോൾ ചാറ്റ് .NET ആപ്പ്](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI ഇൻഫെറൻസ് SDK കോഡ് അടിസ്ഥാനമുള്ള സാമ്പിളുകൾ + - Phi-4 സാമ്പിളുകൾ + - [📓] [Phi-4-multimodal ഉപയോഗിച്ച് പ്രോജക്ട് കോഡ് ജനറേറ്റ് ചെയ്യുക](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 സാമ്പിളുകൾ - - [മൈക്രോസോഫ്റ്റ് Phi-3 കുടുംബവുമായി നിങ്ങളുടെ സ്വന്തം Visual Studio Code GitHub Copilot ചാറ്റ് നിർമ്മിക്കുക](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub മോഡലുകളുമായി Phi-3.5 ഉപയോഗിച്ച് നിങ്ങളുടെ സ്വന്തം Visual Studio Code ചാറ്റ് കോപിലറ്റ് ഏജന്റ് സൃഷ്ടിക്കുക](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - പുരോഗമിച്ച ലജ്ജാപൂർവക സാമ്പിളുകൾ - - Phi-4 സാമ്പിളുകൾ - - [📓] [Phi-4-മിനി-ലജ്ജാപൂർവക അല്ലെങ്കിൽ Phi-4-ലജ്ജാപൂർവക സാമ്പിളുകൾ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [മൈക്രോസോഫ്റ്റ് ഓലീവ് ഉപയോഗിച്ച് Phi-4-മിനി-ലജ്ജാപൂർവക ഫൈൻ-ട്യൂൺ ചെയ്യുന്നത്](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [ആപ്പിൾ MLX ഉപയോഗിച്ച് Phi-4-മിനി-ലജ്ജാപൂർവക ഫൈൻ-ട്യൂൺ ചെയ്യുന്നത്](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [GitHub മോഡലുകളുമായി Phi-4-മിനി-ലജ്ജാപൂർവക](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [മൈക്രോസോഫ്റ്റ് ഫൗണ്ടറി മോഡലുകളുമായി Phi-4-മിനി-ലജ്ജാപൂർവക](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [Microsoft Phi-3 ശൃംഖല ഉപയോഗിച്ച് നിങ്ങളുടെ സ്വന്തം Visual Studio Code GitHub Copilot Chat നിർമ്മിക്കുക](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub മോഡലുകൾ ഉപയോഗിച്ച് Phi-3.5 ഉപയോഗിച്ച് Visual Studio Code Chat Copilot ഏജന്റ് സൃഷ്ടിക്കുക](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - അഡ്വാൻസ് ചെയ്ത റീഫീസനിംഗ് സാമ്പിളുകൾ + - Phi-4 സാമ്പിളുകൾ + - [📓] [Phi-4-mini-reasoning അല്ലെങ്കിൽ Phi-4-reasoning സാമ്പിളുകൾ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Microsoft Olive ഉപയോഗിച്ച് Phi-4-mini-reasoning ഫൈൻ-ട്യൂണിങ്](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLX ഉപയോഗിച്ച് Phi-4-mini-reasoning ഫൈൻ-ട്യൂണിങ്](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [GitHub മോഡലുകളോടൊപ്പം Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Microsoft Foundry മോഡലുകളോടൊപ്പം Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - ഡെമോകൾ - - [Phi-4-മിനി ഡെമോകൾ Hugging Face Spaces-ൽ ഹോസ്റ്റ് ചെയ്തു](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-മൾട്ടിമോഡൽ ഡെമോകൾ Hugginge Face Spaces-ൽ ഹോസ്റ്റ് ചെയ്തു](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - കാഴ്ച സാമ്പിളുകൾ - - Phi-4 സാമ്പിളുകൾ - - [📓] [ചിത്രങ്ങൾ വായിക്കാന്‍ Phi-4-മൾട്ടിമോഡൽ ഉപയോഗിച്ച് കോഡ് സൃഷ്ടിക്കുക](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [Hugging Face Spaces-ൽ ഹോസ്റ്റ് ചെയ്ത Phi-4-mini ഡെമോകൾ](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Hugging Face Spaces-ൽ ഹോസ്റ്റ് ചെയ്ത Phi-4-multimodal ഡെമോകൾ](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - വിഷൻ സാമ്പിളുകൾ + - Phi-4 സാമ്പിളുകൾ + - [📓] [Phi-4-multimodal ഉപയോഗിച്ച് ചിത്രം വായിക്കുക, കോഡ് ജനറേറ്റ് ചെയ്യുക](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 സാമ്പിളുകൾ - - [📓][Phi-3-കാഴ്ച-ചിത്രം എഴുത്തിൽ നിന്ന് എഴുത്തിലേക്ക്](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-കാഴ്ച-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-കാഴ്ച CLIP എംബെഡ്ഡിങ്ങ്](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [ഡെമോ: Phi-3 റീസൈക്ലിങ്](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-കാഴ്ച - ദൃശ്യ ഭാഷാ സഹായി - Phi3-കാഴ്ചയും OpenVINO-യും ഉപയോഗിച്ച്](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 ദൃശ്യം Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 ദൃശ്യം OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 ദൃശ്യം മൾട്ടി-ഫ്രെയിം അല്ലെങ്കിൽ മൾട്ടി-ചിത്രം സാമ്പിൾ](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 ദൃശ്യം ഒരു ലൊക്കൽ ONNX മോഡൽ Microsoft.ML.OnnxRuntime .NET ഉപയോഗിച്ച്](../../md/04.HOL/dotnet/src/LabsPhi303) - - [മെനു അടിസ്ഥാനമാക്കിയ Phi-3 ദൃശ്യം ലൊക്കൽ ONNX മോഡൽ Microsoft.ML.OnnxRuntime .NET ഉപയോഗിച്ച്](../../md/04.HOL/dotnet/src/LabsPhi304) - - - ലജ്ജാപൂർവക-കാഴ്ച സാമ്പിളുകൾ - - Phi-4-ലജ്ജാപൂർവക-കാഴ്ച-15B - - [📓] [Phi-4-ലജ്ജാപൂർവക-കാഴ്ച-15B ഉപയോഗിച്ച് ജെവാക്കിങ് കണ്ടെത്തൽ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Phi-4-ലജ്ജാപൂർവക-കാഴ്ച-15B ഉപയോഗിച്ച് ഗണിതം](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-ലജ്ജാപൂർവക-കാഴ്ച-15B ഉപയോഗിച്ച് UI കണ്ടെത്തലും](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓][Phi-3-vision-ചിത്രം എഴുത്തായി മാറ്റൽ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vision CLIP എംബഡ്ഡിങ്](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [ഡെമോ: Phi-3 റീസൈക്ലിംഗ്](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - ദൃശ്യഭാഷാ അസിസ്റ്റന്റ് - Phi3-vision, OpenVINO ഉപയോഗിച്ചത്](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision മൾട്ടി-ഫ്രेम് അല്ലെങ്കിൽ മൾട്ടി-ഇമേജ് സാമ്പിൾ](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Microsoft.ML.OnnxRuntime .NET ഉപയോഗിച്ച് Phi-3 Vision ലോക്കൽ ONNX മോഡൽ](../../md/04.HOL/dotnet/src/LabsPhi303) + - [മെനു അടിസ്ഥാനത്തിൽ Phi-3 Vision ലോക്കൽ ONNX മോഡൽ Microsoft.ML.OnnxRuntime .NET ഉപയോഗിച്ച്](../../md/04.HOL/dotnet/src/LabsPhi304) + + - റീഫീസനിംഗ്-വിഷൻ സാമ്പിളുകൾ + - Phi-4-റീഫീസനിംഗ്-വിഷൻ-15B + - [📓] [Phi-4-റീഫീസനിംഗ്-വിഷൻ-15B ജേ-വാല്കിംഗ് കണ്ടെത്താൻ ഉപയോഗിക്കുന്നു](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-റീഫീസനിംഗ്-വിഷൻ-15B കൊണ്ട് ഗണിതം ചെയ്യുന്നു](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Phi-4-റീഫീസനിംഗ്-വിഷൻ-15B ഉപയോഗിച്ച് UI കണ്ടെത്തൽ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - ഗണിത സാമ്പിളുകൾ - - Phi-4-മിനി-ഫ്ലാഷ്-ലജ്ജാപൂർവക-ഇൻസ്ട്രക്ട് സാമ്പിളുകൾ [Phi-4-മിനി-ഫ്ലാഷ്-ലജ്ജാപൂർവക-ഇൻസ്ട്രക്ട് ഉപയോഗിച്ചുള്ള ഗണിത ഡെമോ](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-മിനി-ഫ്ലാഷ്-റീഫീസനിംഗ്-ഇൻസ്ട്രക്റ്റ് സാമ്പിളുകൾ [Phi-4-മിനി-ഫ്ലാഷ്-റീഫീസനിംഗ്-ഇൻസ്ട്രക്റ്റ് ഗണിത ഡെമോ](./md/02.Application/09.Math/MathDemo.ipynb) - - ശബ്ദ സാമ്പിളുകൾ - - Phi-4 സാമ്പിളുകൾ - - [📓] [Phi-4-മൾട്ടിമോഡൽ ഉപയോഗിച്ച് ശബ്ദ ട്രാൻസ്‌ക്രിപ്റ്റുകൾ എടുക്കുന്നു](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-മൾട്ടിമോഡൽ ശബ്ദ സാമ്പിൾ](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-മൾട്ടിമോഡൽ ശബ്ദ വിവർത്തനം സാമ്പിൾ](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET കൺസോൾ ആപ്ലിക്കേഷൻ Phi-4-മൾട്ടിമോഡൽ ശബ്ദം വിശകലനം ചെയ്യുന്നതിനും ട്രാൻസ്‌ക്രിപ്റ്റ് സൃഷ്ടിക്കുന്നതിനും ഉപയോഗിക്കുന്നു](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - ഓഡിയോ സാമ്പിളുകൾ + - Phi-4 സാമ്പിളുകൾ + - [📓] [Phi-4-multimodal ഉപയോഗിച്ച് ഓഡിയോ ട്രാൻസ്ക്രിപ്റ്റുകൾ എടുക്കുന്നു](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal ഓഡിയോ സാമ്പിൾ](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal സ്പീച്ച് ട്രാൻസ്ലേഷൻ സാമ്പിൾ](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET കൺസോൾ ആപ്പ് Phi-4-multimodal ഉപയോഗിച്ച് ഓഡിയോ ഫയൽ വിശകലനം ചെയ്ത് ട്രാൻസ്ക്രിപ്റ്റ് നിർമ്മിക്കുന്നു](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE സാമ്പിളുകൾ - Phi-3 / 3.5 സാമ്പിളുകൾ - - [📓] [Phi-3.5 സ്പെഷ്യലിസ്റ്റ് മിശ്രിത മോഡലുകൾ (MoEs) സോഷ്യൽ മീഡിയ സാമ്പിൾ](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search, LlamaIndex ഉപയോഗിച്ചുള്ള Retrieval-Augmented Generation (RAG) പൈപ്പ്‌ലൈൻ നിർമ്മാണം](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 എക്സ്പർട്സ് മിക്‌ഷർ (MoEs) സോഷ്യൽ മീഡിയ സാമ്പിൾ](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search, LlamaIndex ഉപയോഗിച്ച് റിട്ട്രീവൽ-ഓഗ്മെന്റഡ് ജനറേഷൻ (RAG) പൈപ്പ്‌ലൈൻ നിർമ്മിക്കൽ](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - ഫംഗ്ഷൻ കോളിംഗ് സാമ്പിളുകൾ + - ഫംഗ്ഷൻ കോൾ ചെയ്യൽ സാമ്പിളുകൾ - Phi-4 സാമ്പിളുകൾ 🆕 - - [📓] [Phi-4-മിനി ഉപയോഗിച്ച് ഫംഗ്ഷൻ കോളിംഗ് ഉപയോഗിക്കുന്നു](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Phi-4-മിനി ഉപയോഗിച്ച് മൾട്ടി-ഏജന്റുമാർ സൃഷ്ടിക്കാൻ ഫംഗ്ഷൻ കോളിംഗ് ഉപയോഗിക്കുന്നു](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Ollama-വുമായുള്ള ഫംഗ്ഷൻ കോളിംഗ് ഉപയോഗിക്കുന്നത്](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [ONNX-വുമായുള്ള ഫംഗ്ഷൻ കോളിംഗ് ഉപയോഗിക്കുന്നു](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - മൾട്ടിമോഡൽ മിക്സിങ് സാമ്പിളുകൾ + - [📓] [Phi-4-mmini ഉപയോഗിച്ച് ഫംഗ്ഷൻ കോൾ ചെയ്യൽ](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Phi-4-mmini ഉപയോഗിച്ച് മൾട്ടി-ഏജന്റുകൾ സൃഷ്ടിക്കാൻ ഫംഗ്ഷൻ കോൾ ചെയ്യൽ](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Ollama ഉപയോഗിച്ച് ഫംഗ്ഷൻ കോൾ ചെയ്യൽ](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [ONNX ഉപയോഗിച്ച് ഫംഗ്ഷൻ കോൾ ചെയ്യൽ](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - മൾട്ടിമോഡൽ മിശ്രിത സാമ്പിളുകൾ - Phi-4 സാമ്പിളുകൾ 🆕 - - [📓] [ടെക്നോളജി പത്രകാരൻ ആയി Phi-4-മൾട്ടിമോഡൽ ഉപയോഗിക്കുന്നത്](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET കൺസോൾ ആപ്ലിക്കേഷൻ Phi-4-മൾട്ടിമോഡൽ ഉപയോഗിച്ച് ചിത്രങ്ങൾ വിശകലനം ചെയ്യുന്നതിന്](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [📓] [സാങ്കേതിക മാധ്യമപ്രതിനിധിയായി Phi-4-multimodal ഉപയോഗിക്കൽ](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET കൺസോൾ ആപ്പ് ഇമേജുകൾ വിശകലനം ചെയ്യാൻ Phi-4-multimodal ഉപയോഗിക്കുന്നു](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - ഫൈൻ-ട്യൂണിംഗ് Phi സാമ്പിളുകൾ - - [ഫൈൻ-ട്യൂണിംഗ് സീനാരിയോകൾ](./md/03.FineTuning/FineTuning_Scenarios.md) + - [ഫൈൻ-ട്യൂണിംഗ് സീനാറിയോകൾ](./md/03.FineTuning/FineTuning_Scenarios.md) - [ഫൈൻ-ട്യൂണിംഗ് vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Phi-3 വ്യവസായ വിദഗ്ധനാകാൻ അനുവദിക്കുക ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [VS കോഡ് എഐ ടൂൾകിറ്റ് ഉപയോഗിച്ച് Phi-3 ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [അസ്യൂർ മെഷീൻ ലേണിംഗ് സർവീസ് ഉപയോഗിച്ച് Phi-3 ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/Introduce_AzureML.md) + - [Phi-3 ഒരു വ്യവസായ വിദഗ്ധരാകാൻ അനുവദിക്കുന്ന ഫൈൻ-ട്യൂൺ](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [AI ടൂൾകിറ്റ് വഴി Phi-3 ഫൈൻ-ട്യൂണിംഗ് VS കോഡിൽ](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure മെഷീൻ ലേണിംഗ് സേവനം ഉപയോഗിച്ച് Phi-3 ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/Introduce_AzureML.md) - [Lora ഉപയോഗിച്ച് Phi-3 ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/FineTuning_Lora.md) - [QLora ഉപയോഗിച്ച് Phi-3 ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/FineTuning_Qlora.md) - [Microsoft Foundry ഉപയോഗിച്ച് Phi-3 ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/FineTuning_AIFoundry.md) - [Azure ML CLI/SDK ഉപയോഗിച്ച് Phi-3 ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/FineTuning_MLSDK.md) - [Microsoft Olive ഉപയോഗിച്ച് ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive ഹാൻഡ്‌സ്-ഓൺ ലാബ് ഉപയോഗിച്ച് ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/olive-lab/readme.md) + - [Microsoft Olive ഹാൻസ്-ഓൺ ലാബ് ഉപയോഗിച്ച് ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/olive-lab/readme.md) - [Weights and Bias ഉപയോഗിച്ച് Phi-3-vision ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Apple MLX ഫ്‌റെയിംവർക്ക് ഉപയോഗിച്ച് Phi-3 ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-vision ഫൈൻ-ട്യൂണിംഗ് (അധികൃത പിന്തുണ)](./md/03.FineTuning/FineTuning_Vision.md) - - [കൈറ്റോ AKS, അസ്യൂർ കന്റെയ്‌നർമാർ (അധികൃത പിന്തുണ) ഉപയോഗിച്ച് ഫൈ-3 ഫൈൻ-ട്യൂണിംഗ്](./md/03.FineTuning/FineTuning_Kaito.md) - - [ഫൈ-3 ആൻഡ് 3.5 വിഷൻ ഫൈൻ-ട്യൂണിംഗ്](https://github.com/2U1/Phi3-Vision-Finetune) + + - [ആപ്പിള്‍ MLX ഫ്രെയിംവര്‍ക്കുമായി Phi-3 ഫൈന്‍-ട്യൂണിംഗ്](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision ഫൈന്‍-ട്യൂണിംഗ് (അധिकारिक പിന്തുണ)](./md/03.FineTuning/FineTuning_Vision.md) + - [Kaito AKS, Azure Containers ഉപയോഗിച്ച് Phi-3 ഫൈന്‍-ട്യൂണിംഗ് (അധिकारिक പിന്തുണ)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Phi-3 and 3.5 Vision ഫൈന്‍-ട്യൂണിംഗ്](https://github.com/2U1/Phi3-Vision-Finetune) - ഹാൻഡ്സ് ഓൺ ലാബ് - - [അത്യാധുനിക മോഡലുകൾ പഠനം: LLMകൾ, SLMകൾ, പ്രാദേശിക ഡെവലപ്പ്മെന്റ് ഉൾപ്പെടെ](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP ശേഷി തുറക്കൽ: Microsoft Olive ഉപയോഗിച്ച് ഫൈൻ-ട്യൂണിംഗ്](https://github.com/azure/Ignite_FineTuning_workshop) + - [ആദ്യ നിര മോഡലുകള്‍ പരീക്ഷിക്കുന്നത്: LLMs, SLMs, ലോക്കൽ ഡെവലപ്മെന്റ് എന്നിവ](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP ശേഷി തെളിയിക്കൽ: Microsoft Olive ഉപയോഗിച്ച് ഫൈന്‍-ട്യൂണിംഗ്](https://github.com/azure/Ignite_FineTuning_workshop) -- അക്കാദമിക് ഗവേഷണ പേപ്പറുകളും പ്രസിദ്ധീകരണങ്ങളും - - [Textbooks Are All You Need II: phi-1.5 സാങ്കേതിക റിപ്പോർട്ട്](https://arxiv.org/abs/2309.05463) - - [Phi-3 സാങ്കേതിക റിപ്പോർട്ട്: നിങ്ങളുടെ ഫോണിൽ ലോക്കലായി എല്ലാ കഴിവുള്ള ഭാഷാ മോഡൽ](https://arxiv.org/abs/2404.14219) - - [Phi-4 സാങ്കേതിക റിപ്പോർട്ട്](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini സാങ്കേതിക റിപ്പോർട്ട്: കംപാക്റ്റും ശക്തിയേറിയ മൾട്ടിമോഡൽ ഭാഷാ മോഡലുകൾ Mixture-of-LoRAs വഴി](https://arxiv.org/abs/2503.01743) - - [കാറിനായി ചെറിയ ഭാഷാ മോഡലുകൾ ഫംഗ്ഷൻ-കോൾക്കായി ഓപ്റ്റിമൈസ് ചെയ്യുന്നു](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) PHI-3-ന്റെ मल्टिൾ ചോയ്സ് ക്വഷൻ ആൻസറിംഗിനുള്ള ഫൈൻ-ട്യൂണിംഗ്: രീതിശാസ്ത്രം, ഫലങ്ങൾ, വെല്ലുവിളികൾ](https://arxiv.org/abs/2501.01588) - - [Phi-4-റീസണിംഗ് സാങ്കേതിക റിപ്പോർട്ട്](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mMini-റീസണിംഗ് സാങ്കേതിക റിപ്പോർട്ട്](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) +- അക്കാദമിക് ഗവേഷണ पेപ്പർസും പ്രസിദ്ധീകരണങ്ങളും + - [Textbooks Are All You Need II: phi-1.5 സാങ്കേതിക റിപ്പോര്‍ട്ട്](https://arxiv.org/abs/2309.05463) + - [Phi-3 സാങ്കേതിക റിപ്പോര്‍ട്ട്: നിങ്ങളുടെ ഫോണില്‍ highly capable language model](https://arxiv.org/abs/2404.14219) + - [Phi-4 സാങ്കേതിക റിപ്പോര്ട്ട്](https://arxiv.org/abs/2412.08905) + - [Phi-4-Mini സാങ്കേതിക റിപ്പോര്ട്ട്: മിക്സ്ചർ ഓഫ് ലോറാസിലൂടെ ചെറുതും ശക്തിയുള്ള മൾട്ടിമോഡല്‍ ലാംഗ്വേജ് മോഡലുകള്‍](https://arxiv.org/abs/2503.01743) + - [ഇൻ-വണ്ടിയിലുള്ള ഫംഗ്ഷൻ-കൊല്ലിംഗിനായി ചെറിയ ഭാഷാ മോഡലുകൾ ഓപ്റ്റിമൈസ് ചെയ്യുന്നു](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) ഫൈന്‍-ട്യൂണിംഗ് PHI-3 മൾട്ടിപ്പിൾ-ചോയ്‌സ് ചോദ്യോത്തരത്തിനായി: പ്രക്രിയ, ഫലങ്ങളും വെല്ലുവിളികളും](https://arxiv.org/abs/2501.01588) + - [Phi-4-reasoning സാങ്കേതിക റിപ്പോര്ട്ട്](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-reasoning സാങ്കേതിക റിപ്പോര്ട്ട്](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Phi മോഡലുകൾ ഉപയോഗിക്കൽ +## Phi മോഡലുകൾ ഉപയോഗിക്കുന്നത് -### Microsoft Foundry-യിൽ Phi +### Microsoft Foundry ല്‍ Phi -നിങ്ങളുടെ വ്യത്യസ്ത ഹാർഡ്വെയർ ഉപകരണങ്ങളിൽ Microsoft Phi എങ്ങനെ ഉപയോഗിക്കാമെന്ന്, എ2ഇ പരിഹാരങ്ങൾ എങ്ങനെ നിർമ്മിക്കാമെന്ന് നിങ്ങൾക്ക് പഠിക്കാം. Phi സ്വയം അനുഭവിക്കാനായി, മോഡലുകൾ ഉപയോഗിച്ച് കളിച്ചുകൊണ്ട് നിങ്ങളുടെ സ്ഥിതിഗതികൾക്കായി Phi കസ്റ്റമൈസ് ചെയ്യാൻ [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) ഉപയോഗിക്കാമെന്ന് നിങ്ങൾക്ക് പഠിക്കാം. കൂടുതൽ അറിവിന് [Microsoft Foundry]-ന്റെ തുടക്കം / QuickStart കാണുക (/md/02.QuickStart/AzureAIFoundry_QuickStart.md). +Microsoft Phi ഉപയോഗിക്കുന്നത് പരിശീലിക്കുകയും വിവിധ ഹാർഡ്വെയർ ഉപകരണങ്ങളിൽ end-to-end പരിഹാരങ്ങൾ നിർമ്മിക്കുകയും ചെയ്യാൻ നിങ്ങൾക്ക് ഇതിൽ പഠിക്കാം. Phi നേരിട്ട് അനുഭവിക്കാൻ, മോഡലുകളില്‍ കളിച്ച് നിങ്ങളുടെ സാഹചര്യങ്ങൾക്കനുസരിച്ചു Phi ഇഷ്ടാനുസൃതമാക്കുക [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) ഉപയോഗിക്കുക, കൂടാതെ [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) എന്നത് ആരംഭിക്കുക. -**പ്ലേഗ്രൗണ്ട്** -ഓരോ മോഡലിനും പരീക്ഷിക്കാൻ സങ്കല്പിതമായൊരു [Azure AI Playground](https://aka.ms/try-phi3) ഉണ്ട്. +**പ്ലേഗ്രൗണ്ട്** +ഓരോ മോഡലിനും പരീക്ഷിക്കാൻ സമർപ്പിച്ച പ്ലേഗ്രൗണ്ട് ഉണ്ട് [Azure AI പ്ലേഗ്രൗണ്ട്](https://aka.ms/try-phi3). ### GitHub മോഡലുകളിൽ Phi -നിങ്ങളുടെ വിവിധ ഹാർഡ്വെയർ ഉപകരണങ്ങളിൽ Microsoft Phi എങ്ങനെ ഉപയോഗിക്കാമെന്ന്, എ2ഇ പരിഹാരങ്ങൾ എങ്ങനെ നിർമ്മിക്കാമെന്ന് നിങ്ങൾക്ക് പഠിക്കാം. Phi സ്വയം അനുഭവിക്കാനായി, മോഡൽ ഉപയോഗിച്ച് കളിച്ചുകൊണ്ട് നിങ്ങളുടെ സ്ഥിതിഗതികൾക്കായി Phi കസ്റ്റമൈസ് ചെയ്യാൻ [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ഉപയോഗിക്കാവുന്നതാണ്. കൂടുതൽ അറിവിന് [GitHub Model Catalog]-ന്റെ തുടക്കം / QuickStart കാണുക (/md/02.QuickStart/GitHubModel_QuickStart.md). +Microsoft Phi ഉപയോഗിച്ച് വേറെയും end-to-end പരിഹാരങ്ങൾ നിങ്ങളുടെ ഹാർഡ്വെയർ ഉപകരണങ്ങളില്‍ എങ്ങിനെയാണ് നിർമ്മിക്കേണ്ടതെന്ന് പഠിക്കാൻ നിങ്ങൾക്ക് കഴിയും. Phi നേരിട്ട് അനുഭവിക്കാനായി മോഡലുമായി കളിച്ച് നിങ്ങളുടെ സാഹചര്യങ്ങൾക്കനുസരിച്ചു Phi ഇഷ്ടാനുസൃതമാക്കുക [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ഉപയോഗിക്കുക, കൂടാതെ [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) ആരംഭിക്കുക. -**പ്ലേഗ്രൗണ്ട്** -ഓരോ മോഡലിനും പരീക്ഷിക്കാൻ പ്രത്യേക [പ്ലേഗ്രൗണ്ട്](/md/02.QuickStart/GitHubModel_QuickStart.md) ഉണ്ട്. +**പ്ലേഗ്രൗണ്ട്** +ഓരോ മോഡലിനും [പരീക്ഷിക്കാൻ സമർപ്പിച്ച പ്ലേഗ്രൗണ്ട്](/md/02.QuickStart/GitHubModel_QuickStart.md) ഉണ്ട്. -### Hugging Face-ൽ Phi +### Hugging Face ല്‍ Phi -മോഡൽ [Hugging Face](https://huggingface.co/microsoft) ൽ ലഭ്യമാണ്. +മോഡൽ [Hugging Face](https://huggingface.co/microsoft) ല്‍ കണ്ടെത്താനും കഴിയും -**പ്ലേഗ്രൗണ്ട്** -[Hugging Chat പ്ലേഗ്രൗണ്ട്](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**പ്ലേഗ്രൗണ്ട്** + [Hugging Chat പ്ലേ ഗ്രൗണ്ട്](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 മറ്റ് കോഴ്‌സുകൾ + ## 🎒 മറ്റ് കോഴ്‌സുകൾ -നമ്മുടെ ടീം മറ്റ് കോഴ്‌സുകളും ഒരുക്കുന്നു! പരിശോധിക്കുക: +ഞങ്ങളുടെ ടീമിന് മറ്റ് കോഴ്‌സുകളും ഉണ്ട്! പരിശോധിക്കുക: -### LangChain -[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain for Beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +### LangChain +[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain for Beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / ഏജന്റുമാർ -[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) ---- +### Azure / Edge / MCP / ഏജന്‍റുകൾ +[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) -### ജനറേറ്റീവ് AI സീരീസ് -[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- + +### ജനറേറ്റീവ് AI പരമ്പര +[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) -### കോർ ലേണിങ് -[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- + +### കോർ പഠനം +[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) -### കോപൈലറ്റ് സീരീസ് -[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +--- + +### കോപൈലറ്റ് പരമ്പര +[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## ഉത്തരവാദിത്വമുള്ള AI +## ഉത്തരവാദിത്തമുള്ള AI + +മൈക്രോസോഫ്റ്റ് നമ്മുടെ AI ഉൽപ്പന്നങ്ങൾ ഉത്തരവാദിത്തത്തോടെ ഉപയോഗിക്കാൻ ഉപഭോക്താക്കളെ സഹായിക്കാനും, പഠനങ്ങൾ പങ്കുവെക്കാനും, ട്രാൻസ്പാരൻസി നോട്ട്‌സ്, ഇംപാക്റ്റ് അസസ്‌മെന്റുകൾ പോലുള്ള ഉപകരണങ്ങളിലൂടെ വിശ്വാസപാരമായ പങ്കാളിത്തങ്ങൾ നിര്‍മിക്കാനും പ്രതിജ്ഞാബദ്ധമാണ്. ഈ സ്രോതസ്സുകളിൽ പലതും [https://aka.ms/RAI](https://aka.ms/RAI) ല്‍ ലഭ്യമാണ്. +മൈക്രോസോഫ്റ്റിന്റെ ഉത്തരവാദിത്തമുള്ള AI സമീപനം, ന്യായത്വം, വിശ്വാസ്യതയും സുരക്ഷയും, സ്വകാര്യതയും സുരക്ഷയും, ഉൾപ്പെടുത്തലും, പരദർശിത്വവും ഉത്തരവാദിത്തവും എന്ന ഞങ്ങളുടെ AI 원칙ങ്ങളിലാണ് അടിസ്ഥിതമായിരിക്കുന്നത്. + +വലിയതലത്തിലുള്ള പ്രകൃതി ഭാഷ, ചിത്രം, സംഭാഷണ മോഡലുകൾ - ഈ സാമ്പിളിൽ ഉപയോഗിക്കുന്നവ പോലുള്ള - അനീതിയുള്ള, വിശ്വാസയോഗ്യമല്ലാത്ത, അല്ലെങ്കിൽ അപമാനകരമായ രീതിയിൽ പ്രവർത്തിക്കാനും നുഴഞ്ഞുകയറാനും സാധ്യതയുണ്ട്, ഇത് ഹാനിക്കരമായിരിക്കാം. ദയവായി അപകടങ്ങളും പരിധികളും അറിയാൻ [Azure OpenAI സർവീസ് ട്രാൻസ്പാരൻസി നോട്ട്](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) കാണുക. -Microsoft നമ്മുടെ ഉപഭോക്താക്കൾ AI ഉൽപ്പന്നങ്ങൾ ഉത്തരവാദിത്വത്തോടെ ഉപയോഗിക്കാൻ സഹായിക്കാനും, നമുക്ക് നിന്ന് പഠനങ്ങൾ പങ്കുവെക്കാനും, ട്രാൻസ്‌പറൻസി നോട്ടുകളും പ്രഭാവം വിലയിരുത്തലുകളും പോലുള്ള ഉപകരണങ്ങളിലൂടെ വിശ്വാസം അടങ്ങിയ പങ്കാളിത്തങ്ങൾ സൃഷ്ടിക്കാനും പ്രതിജ്ഞാബദ്ധമാണ്. ഈ പിന്തുണകൾ [https://aka.ms/RAI](https://aka.ms/RAI) എന്നത് കാണാവുന്നതാണ്. -Microsoft‍റെ ഉത്തരവാദിത്വ AI സമീപനം നീതിമാന്മാരാകൽ, വിശ്വാസ്യതയും സുരക്ഷയും, സ്വകാര്യതയും സുരക്ഷയും, ഉൾക്കൊള്ളൽ, സുതാര്യത, ഉത്തരവാദിത്വം തുടങ്ങിയ AI സ 원 ത്തുകളിലടിസ്ഥാനമാണ്. -ഈ ഉദാഹരണത്തിൽ ഉപയോഗിച്ച വലിയ പരിധിയിലെ സ്വാഭാവിക ഭാഷ, ചിത്രങ്ങൾ, ശബ്ദ മോഡലുകൾ അനീതി, വിശ്വാസ്യതകളില്ലായ്ന്ന് അല്ലെങ്കിൽ അപമാനകരമായി പെരുമാറാൻ സാധ്യതയുള്ളതിനാൽ കേടുപാടുകൾ ഉണ്ടാക്കും. ദയവായി അപകടങ്ങളും പരിമിതികളും അറിയാൻ [Azure OpenAI സേവന ട്രാൻസ്‌പറൻസി നോട്ടും](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) പരിശോധിക്കുക. -ഈ അപകടങ്ങൾ കുറയ്ക്കാൻ ശിപാർസുചെയ്ത സമീപനം നിങ്ങളുടെ വാസ്തുവിദ്യയിൽ സുരക്ഷാ സംവിധാനം ഉൾപ്പെടുത്തുകയാണ്, ഇത് ഹാനികരമായ പെരുമാറ്റം കണ്ടെത്തിയും തടഞ്ഞു നിർത്തിയും കഴിയും. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) സ്വതന്ത്രമായ ഒറ്റ പാളി സംരക്ഷണം നൽകുന്നു, ആപ്ലിക്കേഷനുകളും സേവനങ്ങളിലും ഹാനികരമായ ഉപയോക്താവ് സൃഷ്ടിച്ചും AI സൃഷ്ടിച്ചും ഉള്ള ഉള്ളടക്കം കണ്ടെത്താൻ കഴിയും. Azure AI Content Safety ല് ടെക്സ്റ് ಮತ್ತು ഇമേജ് API-കൾ ഉൾക്കൊള്ളുന്നു, അവ ഹാനികരമായ ഉള്ളടക്കം കണ്ടെത്താൻ സഹായിക്കുന്നു. Microsoft Foundry-യിൽ, Content Safety സേവനം വിവിധ രൂപത്തിലുള്ള ഹാനികരമായ ഉള്ളടക്കം കണ്ടെത്തുന്നതിനുള്ള സാമ്പിൾ കോഡ് കാണാനും പരീക്ഷിക്കാനും സഹായിക്കുന്നു. താഴെ ഡോക്യുമെന്റേഷനിലുള്ള [ക്വിക്ക്‌സ്റ്റാർട്ട്](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) സേവനത്തിന് അഭ്യർത്ഥനകൾ പടിഞ്ഞിരിക്കുന്നു. +ഈ അപകടങ്ങൾ നിയന്ത്രിക്കുന്നതിനുള്ള ശുപാർശ ചെയ്യുന്ന സമീപനം, നിങ്ങളുടെ ആർക്കിടെക്ചറിൽ ഒരു സുരക്ഷാ സംവിധാനം ഉൾപ്പെടുത്തുകയാണ്, ഇത് ഹാനികരമായ പെരുമാറ്റം കണ്ടെത്താനും തടയാനും കഴിയും. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) അവിടെയുള്ള ആപ്പ്‌ളിക്കേഷനുകളിലും സേവനങ്ങളിലും ഹാനികരമായ ഉപയോക്താവ്-സമർപ്പിച്ചും AI-സൃഷ്ടിച്ചും ഉള്ള ഉള്ളടക്കം കണ്ടെത്താനും തടയാനും കഴിയുന്ന സ്വതന്ത്രഏകക സുരക്ഷാ അടിസ്ഥാനമാണ്. Azure AI Content Safety ഹാനികരമായ വിഷയങ്ങൾ കണ്ടെത്താനുള്ള ടെക്സ്റ്റ്‌യും ഇമേജ് API-കളും ഉൾക്കൊള്ളുന്നു. Microsoft Foundry-യിൽ ഉള്ള Content Safety സേവനം, വ്യത്യസ്ത മോഡാലിറ്റികളിലെ ഹാനികരമായ ഉള്ളടക്കം കണ്ടെത്തുന്നതിനുള്ള സാംപിൾ കോഡ് കാണുവാനും പരിശോധന നടത്തുവാനും അനുവദിക്കുന്നു. താഴെപ്പറയുന്ന [ക്വിക്‌സ്റ്റാർട്ട് ഡോക്യുമെന്റേഷൻ](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) സേവനത്തിനായി നിർദ്ദേശങ്ങൾ നൽകുന്നു. -ആനുകൂല്യമായി പരിഗണിക്കേണ്ട മറ്റൊരു ഘടകം ആപ്ലിക്കേഷൻ സമഗ്ര പ്രകടനമാണ്. മൾട്ടി-മോഡൽ, മൾട്ടി-മോഡൽ ആപ്ലിക്കേഷനുകളുടെ സാഹചര്യത്തിൽ, പ്രകടനം എന്ന് അർത്ഥമാക്കുന്നത് നിങ്ങൾക്കും നിങ്ങളുടെ ഉപയോക്താക്കളും പ്രതീക്ഷിക്കുന്ന പോലെ സിസ്റ്റം പ്രവർത്തിക്കണമെന്നും, ഹാനികരമായ ഔട്ട്പുട്ടുകൾ സൃഷ്ടിക്കാതിരിക്കണമെന്നും ആണ്. നിങ്ങളുടെ ആപ്ലിക്കേഷന്റെ സമഗ്ര പ്രകടനം [പ്രകടനവും ഗുണനിലവാരവും അപകടം സുരക്ഷാ വിവരങ്ങളും കുറിച്ച് സ്റ്റാൻഡേർഡ് വിലയിരുത്തലുകളിലൂടെ](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) വിലയിരുത്തുന്നത് അത്യന്താപേക്ഷിതമാണ്. നിങ്ങൾക്ക് അതോടൊപ്പം [കസ്റ്റം വിലയിരുത്തലുകൾ](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) സൃഷ്ടിച്ചും അവലോകനം നടത്തിയും ചെയ്യാനുള്ള കഴിവുണ്ട്. +മറ്റൊരു പരിഗണിക്കേണ്ട ഘടകം ആകെ ആപ്‌ളിക്കേഷൻ പ്രകടനമാണ്. ബഹുമോദാലിട്ടും ബഹുമോഡൽ ആപ്ലിക്കേഷനുകളിലും, സിസ്റ്റം നിങ്ങളുടെ ഉദ്ദേശപ്രകാരം പ്രവർത്തിക്കുന്നതായും ഉപയോക്താക്കളുടെ പ്രതീക്ഷകൾക്കനുസരിച്ച്, ഹാനികരമായ ഔട്ട്പുട്ടുകൾ സൃഷ്ടിക്കപ്പെടാതിരിക്കുകയും ചെയ്യുന്നതായും പ്രകടനം ഗണിക്കുന്നു. നിങ്ങളുടെ ആകെ ആപ്‌ളിക്കേഷന്റെ പ്രകടനം [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) ഉപയോഗിച്ച് വിലയിരുത്തുന്നത് നിർണായകമാണ്. [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) സൃഷ്ടിക്കുകയും അവ ഉപയോഗിച്ചു വിലയിരുത്തുകയും ചെയ്യുന്നതും കഴിയും. -നിങ്ങളുടെ AI ആപ്ലിക്കേഷൻ വികസന പരിസ്ഥിതിയിൽ [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ഉപയോഗിച്ച് നിങ്ങൾക്ക് വിലയിരുത്താം. ഒരു ടെസ്റ്റ് ഡാറ്റാസെറ്റ് അല്ലെങ്കിൽ ലക്ഷ്യം നൽകിയാൽ, നിങ്ങളുടെ ജനറേറ്റീവ് AI ആപ്ലിക്കേഷൻ ജനറേഷനുകൾ ഉൾപ്പെടുത്തിയിട്ടുള്ളവയോ നിങ്ങൾ തെരഞ്ഞെടുത്ത കസ്റ്റം വിലയിരുത്തലുകളോ ഉപയോഗിച്ച് കണക്ക് കൂട്ടി അളക്കപ്പെടും. നിങ്ങളുടെ സിസ്റ്റം വിലയിരുത്താൻ azure ai evaluation sdk തുടങ്ങി തുടങ്ങി വേണ്ടി, നിങ്ങൾക്ക് താഴെപ്പറയുന്ന [ക്വിക്ക്‌സ്റ്റാർട്ട് ഗൈഡ്](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) പിന്തുടരാം. ഒരു വിലയിരുത്തൽ റൺ പ്രവർത്തിപ്പിച്ചശേഷം, നിങ്ങൾക്ക് [Microsoft Foundry-യിൽ ഫലങ്ങൾ ദൃശ്യമാക്കാം](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +[Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ഉപയോഗിച്ച് നിങ്ങൾക്ക് നിങ്ങളുടെ AI ആപ്ലിക്കേഷൻ വികസന പരിസ്ഥിതിയിൽ തന്നെ വിലയിരുത്താം. ടെസ്റ്റ് ഡാറ്റാസെറ്റ് അല്ലെങ്കിൽ ടാർഗെറ്റ് നൽകിയാൽ, നിങ്ങളുടെ ജനറേറ്റീവ് AI ആപ്ലിക്കേഷന്റെ സൃഷ്ടികൾ സ്‌ഥിതിഗതികൾ ഉൾക്കൊള്ളുന്ന ഇൻ-ബിൽറ്റ് ഇവല്യുവേറ്ററുകൾ അല്ലെങ്കിൽ നിങ്ങളുടെ തെരഞ്ഞെടുക്കുന്ന കസ്റ്റം ഇവല്യുവേറ്ററുകൾ ഉപയോഗിച്ച് സംഖ്യാത്മകമായി അളക്കപ്പെടുന്നു. നിങ്ങളുടെ സിസ്റ്റം വിലയിരുത്താൻ ആഴ്യൂർ AI ഇവല്യുവേഷൻ SDK ഉപയോഗിച്ച് തുടങ്ങുവാൻ [ക്വിക്‌സ്റ്റാർട്ട് ഗൈഡ്](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) പിന്തുടരാം. ഒരു ഇവല്യുവേഷൻ റൺ നടത്തുമ്പോൾ, നിങ്ങൾക്ക് [Microsoft Foundry-ൽ ഫലം ദൃശ്യവത്കരിക്കാൻ](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) കഴിയും. -## ട്രേഡ്മാർക്കുകൾ +## ട്രേഡ്മാർക്ക്‌ಗಳು -ഈ പ്രോജക്ടിൽ പ്രോജക്ടുകൾ, ഉൽപ്പന്നങ്ങൾ, സേവനങ്ങൾ എന്നിവയുടെ ട്രേഡ്മാർക്കുകൾ അല്ലെങ്കിൽ ലോഗോകൾ ഉണ്ടാകാവുന്നതാണ്. മൈക്രോസോഫ്റ്റ് ട്രേഡ്മാർക്കുകൾ അല്ലെങ്കിൽ ലോഗോകൾ അനുമതിപ്‌തു ഉപയോഗിക്കുന്നത് [Microsoft-ന്റെ ട്രേഡ്മാർക് & ബ്രാൻഡ് മാർഗനിർദ്ദേശങ്ങൾ](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) അനുസരിച്ച് ആയിരിക്കണം. -മൈക്രോസോഫ്റ്റ് ട്രേഡ്മാർക്കുകൾ അല്ലെങ്കിൽ ലോഗോകൾ ഉപയോഗം കൊണ്ടുള്ള തിരുത്തിയ പതിപ്പുകളിൽ ആശയം അല്ലെങ്കിൽ Microsoft-ന്റെ കായികത ഉറപ്പില്ലാതാക്കരുത്. മൂന്നാർക്കുള്ള ട്രേഡ്മാർക്കുകൾ അല്ലെങ്കിൽ ലോഗോകൾ ഉപയോഗിക്കുമ്പോൾ ആ തൃतीयരുടെ നയങ്ങൾ പ്രധാനം ചെയ്യും. +ഈ പ്രോജക്ടിൽ പ്രോജക്റ്റുകൾ, ഉൽപ്പന്നങ്ങൾ, അല്ലെങ്കിൽ സേവനങ്ങൾക്കുമുള്ള ട്രേഡ്മാർക്ക്‌കളും ലോഗോകളും থাকতে পারে. Microsoft-ന്റെ ട്രേഡ്മാർക്ക് അല്ലെങ്കിൽ ലോഗോകളുടെ അനുമതിയുള്ള ഉപയോഗം [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) പാലിക്കേണ്ടതാണ്. +ഈ പ്രോജക്ടിന്റെ തിരുത്തിയ പതിപ്പുകളിലെ Microsoft ട്രേഡ്മാർക്ക് അല്ലെങ്കിൽ ലോഗോകളുടെ ഉപയോഗം Microsoft പിന്തുണ ഉറപ്പുവരുത്തുകയോ ആശങ്ക സൃഷ്ടിക്കുകയോ ചെയ്യരുത്. മൂന്നാം പാർട്ടി ട്രേഡ്മാർക്ക് അല്ലെങ്കിൽ ലോഗോകളുടെ ഉപയോഗം ആ മൂന്നാം പാർട്ടി നയങ്ങൾക്കു വിധേയമാണ്. -## സഹായം നേടുക +## സഹായം നേടൽ -AI ആപ്പുകൾ നിർമ്മിക്കുന്നതിൽ നിങ്ങൾ കുടുങ്ങുകയോ ചോദ്യങ്ങളുണ്ടെങ്കിലോ, ചേരുക: +AI ആപ്ലിക്കേഷനുകള്‍ സൃഷ്ടിക്കുന്നതിനിടെ നിങ്ങൾക്ക് പ്രശ്നങ്ങളോ ചോദ്യങ്ങളോ ഉണ്ടെങ്കിൽ, ചേരുക: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -നിഗമനസമ്മതം നൽകാൻ അല്ലെങ്കിൽ പിഴവുകളുണ്ടെങ്കിൽ: +ഉൽപ്പന്ന ഫീഡ്‌ബാക്കോ പിഴവുകളോ ഉണ്ടെങ്കിൽ സന്ദർശിക്കുക: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**അസ്വീകാരം**: -ഈ പ്രമാണം AI പരിഭാഷ സേവനമായ [Co-op Translator](https://github.com/Azure/co-op-translator) ഉപയോഗിച്ച് വിവർത്തനം ചെയ്തതാണ്. നാം ശരിയായ വിവർത്തനത്തിനായി ശ്രമിക്കുമ്പോഴും, യന്ത്രം ചെയ്ത വിവർത്തനങ്ങളിൽ പിശകുകൾ അല്ലെങ്കിൽ തെറ്റായ വിവരങ്ങൾ ഉണ്ടാകാൻ സാധ്യതയുള്ളതിനാൽ ദയവായി ശ്രദ്ധിക്കുക. സ്വതന്ത്രമായ ഭാഷയിലുള്ള പ്രാഥമിക പ്രമാണം പ്രാമാണികമായ ഉറവിടമായി പരിഗണിക്കണം. നിർണായക വിവരങ്ങൾക്ക്, പ്രൊഫഷണൽ മനുഷ്യ വിവർത്തനം ശിപാർശ ചെയ്യപ്പെടുന്നു. ഈ വിവർത്തനം ഉപയോഗിക്കുന്നതിൽനിന്ന് വരുന്ന എന്തു തെറ്റിദ്ധാരണകളെയും കൃത്യമായ വ്യാഖ്യാനങ്ങളെയും കാരണം ഞങ്ങൾ ഉത്തരവാദിത്വം ഏറ്റെടുക്കുന്നില്ല. +**അറിയിപ്പ്**: +ഈ രേഖ AI പരിഭാഷാ സേവനം [Co-op Translator](https://github.com/Azure/co-op-translator) ഉപയോഗിച്ച് പരിഭാഷപ്പെടുത്തിയതാണ്. ഞങ്ങൾ കൃത്യതയ്ക്കായി ശ്രമിക്കുന്നുവെങ്കിലും, ഓട്ടോമേറ്റഡ് പരിഭാഷകളിൽ പിഴവുകൾ അല്ലെങ്കിൽ തെറ്റായ വിവരങ്ങൾ ഉണ്ടാകാൻ സാധ്യതയുണ്ട്. അതിന്റെ സ്വാഭാവിക ഭാഷയിലുള്ള അസൽ രേഖയാണ് പ്രാമാണികമായ ഉറവിടമായി പരിഗണിക്കേണ്ടത്. നിർണായകമായ വിവരങ്ങൾക്ക്, പ്രൊഫഷണൽ മനുഷ്യ പരിഭാഷ ശുപാർശ ചെയ്യുന്നു. ഈ പരിഭാഷ ഉപയോഗിച്ച് ഉണ്ടാകുന്ന തെറ്റിദ്ധാരണകൾ അല്ലെങ്കിൽ തെറ്റായ വ്യാഖ്യാനങ്ങൾക്കായി ഞങ്ങൾ ഉത്തരവാദികളല്ല. \ No newline at end of file diff --git a/translations/ml/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/ml/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..052e7989f --- /dev/null +++ b/translations/ml/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 മൈക്രോസോഫ്റ്റ് phi-4-mini & SQLite FTS5 (സീറോ-ക്ലൗഡ് SLM) ഉപയോഗിച്ച ഹൈബ്രിഡ് RAG\n", + "\n", + "> **രചയിതാവ്:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **ഫോകസ്:** ചെറുകിട ഭാഷ മാടൽസുകൾ (SLMs), SQLite FTS5 BM25, ഡെൻസ്സ് എംബെഡിംഗ്‌സ്, റിസിപ്രോകൽ റാങ്ക് ഫ്യൂഷൻ (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. പ്രേരണ: പ്രാദേശിക SLM ലെ കീവേഡ് റീകാൾ പ്രശ്നം\n", + "ഡെൻസ്സ് വെക്ടർ എംബെഡിംഗ്‌സ് മാത്രമേ ആശ്രയിക്കുന്ന സ്റ്റാൻഡേർഡ് RAG ആർക്കിടെക്ചറുകൾ സാധാരണയായി കൃത്യമായ സംഖ്യാത്മക ടോകണുകൾ (ഉദാഹരണത്തിന്, `2,340,000 TL`, കരാർ കോഡുകൾ, അക്കൗണ്ട് നമ്പറുകൾ) റിട്ട്ഷീവ് ചെയ്യാൻ പരാജയപ്പെടുന്നു. \n", + "മറുവായി, സ്‌പാർസ് ലെക്സിക്കൽ സെർച്ച് (BM25) സിമാന്റിക് പദപര്യായങ്ങൾക്കും പാരാഫ്രേസ് ചെയ്ത ചോദ്യങ്ങൾക്കും മിസ്സ് ചെയ്യുന്നു.\n", + "\n", + "ഈ കുക്ക്ബുക്ക് എങ്ങനെ നടപ്പിലാക്കാമെന്ന് കാണിക്കുന്നു **ഉയർന്ന വേഗതയുള്ള, ഇൻ-മെമ്മറി ഹൈബ്രിഡ് റിട്രീവൽ എഞ്ചിൻ** സംയോജിപ്പിക്കുന്നത്:\n", + "1. **ഡെൻസ്സ് വെക്ടർമാർ** (കോസൈൻ സമാനത)\n", + "2. **സ്‌പാർസ് ലെക്സിക്കൽ സെർച്ച്** (SQLite FTS5 BM25)\n", + "3. **റിസിപ്രോകൽ റാങ്ക് ഫ്യൂഷൻ (RRF, $k=60$)**\n", + "4. മൈക്രോസോഫ്റ്റ് `phi-4-mini` ഉപയോഗിച്ച് **സ്ഥാനപദ്ധതിയുള്ള സൈറ്റേഷൻ ജനറേഷൻ (`[1]`, `[2]`)**.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. ഇരട്ട SQLite സ്‌കീമ (സാന്ദ്ര വിക്ടറുകളും വെർച്വൽ FTS5 BM25 ടേബിളും)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. സാമ്പിൾ ഇൻജെക്ഷൻ & എക്സിക്യൂഷൻ ബെഞ്ച്മാർക്ക്\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. മൈക്രോസോഫ്റ്റ് phi-4-mini-യ്ക്കുള്ള ഗ്രൗണ്ടഡ് പ്രോംപ്റ്റ് രൂപീകരണം\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**അറിയിപ്പ്**:\nഈ രേഖ AI പരിഭാഷാ സേവനം [Co-op Translator](https://github.com/Azure/co-op-translator) ഉപയോഗിച്ച് പരിഭാഷപ്പെടുത്തിയതാണ്. ഞങ്ങൾ കൃത്യതയ്ക്കായി ശ്രമിക്കുന്നുവെങ്കിലും, ഓട്ടോമേറ്റഡ് പരിഭാഷകളിൽ പിഴവുകൾ അല്ലെങ്കിൽ തെറ്റായ വിവരങ്ങൾ ഉണ്ടാകാൻ സാധ്യതയുണ്ട്. അതിന്റെ സ്വാഭാവിക ഭാഷയിലുള്ള അസൽ രേഖയാണ് പ്രാമാണികമായ ഉറവിടമായി പരിഗണിക്കേണ്ടത്. നിർണായകമായ വിവരങ്ങൾക്ക്, പ്രൊഫഷണൽ മനുഷ്യ പരിഭാഷ ശുപാർശ ചെയ്യുന്നു. ഈ പരിഭാഷ ഉപയോഗിച്ച് ഉണ്ടാകുന്ന തെറ്റിദ്ധാരണകൾ അല്ലെങ്കിൽ തെറ്റായ വ്യാഖ്യാനങ്ങൾക്കായി ഞങ്ങൾ ഉത്തരവാദികളല്ല.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/mr/.co-op-translator.json b/translations/mr/.co-op-translator.json index 43fc2b24d..f4585c59d 100644 --- a/translations/mr/.co-op-translator.json +++ b/translations/mr/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "mr" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T23:58:45+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:40:39+00:00", "source_file": "README.md", "language_code": "mr" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "mr" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:43:39+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "mr" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:22+00:00", diff --git a/translations/mr/README.md b/translations/mr/README.md index 6f7a86166..6e14b5c82 100644 --- a/translations/mr/README.md +++ b/translations/mr/README.md @@ -1,4 +1,4 @@ -# Phi कुकबुक: Microsoft च्या Phi मॉडेल्ससह प्रत्यक्ष उदाहरणे +# Phi कुकबुक: Microsoft च्या Phi मॉडेल्ससह हँड्स-ऑन उदाहरणे [![GitHub Codespaces मध्ये नमुने उघडा आणि वापरा](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Dev Containers मध्ये उघडा](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) @@ -8,35 +8,35 @@ [![GitHub पुल-रिक्वेस्ट](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![PRs स्वागतार्ह](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub वॉचर्स](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub फोर्क्स](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub पहारेकरी](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub फोर्क](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub स्टार्स](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi हा Microsoft द्वारे विकसित केलेल्या ओपन सोर्स AI मॉडेल्सची एक मालिका आहे. +Phi हा Microsoft द्वारे विकसित केलेल्या मुक्त स्त्रोत AI मॉडेल्सची मालिका आहे. -Phi सध्या सर्वात शक्तिशाली आणि किफायतशीर लहान भाषा मॉडेल (SLM) असून, मल्टि-भाषा, तर्कशक्ति, मजकूर/चैट जनरेशन, कोडिंग, प्रतिमा, ऑडिओ आणि इतर परिस्थितींमध्ये अत्युत्तम परफॉर्मन्स दाखवतो. +Phi सध्या सर्वात शक्तिशाली आणि किफायतशीर लहान भाषा मॉडेल (SLM) आहे, ज्याला बहुभाषिक, तर्क, मजकूर/चॅट तयार करणे, कोडिंग, प्रतिमा, ऑडिओ आणि इतर अनेक परिदृश्यांमध्ये उत्कृष्ट कामगिरी मिळाली आहे. -आपण Phi ला क्लाऊड किंवा एज डिव्हायसेसवर डिप्लॉय करू शकता आणि कमी कम्प्यूटिंग पॉवरसह सहजपणे जनरेटिव्ह AI ऍप्लिकेशन्स तयार करू शकता. +तुम्ही Phi क्लाउड किंवा एज डिव्हायसेसवर डिप्लॉय करू शकता, आणि मर्यादित संगणकीय शक्तीसह सहजपणे सर्जनशील AI अनुप्रयोग तयार करू शकता. -या संसाधनांचा वापर सुरू करण्यासाठी खालील पावले अनुसरा: -1. **रिपॉझिटरी फोर्क करा**: क्लिक करा [![GitHub फोर्क्स](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **रिपॉझिटरी क्लोन करा**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Microsoft AI Discord कम्युनिटीमध्ये सहभागी व्हा आणि तज्ञ व सहकारी विकसकांशी भेटा**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +या संसाधनांचा वापर सुरू करण्यासाठी खालील चरणांचे अनुसरण करा: +1. **रेपो फोर्क करा**: क्लिक करा [![GitHub फोर्क](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **रेपो क्लोन करा**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Microsoft AI Discord समुदायात सहभागी व्हा आणि तज्ञ व इतर विकसकांशी भेटा**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/mr/cover.eb18d1b9605d754b.webp) ### 🌐 बहुभाषिक समर्थन -#### GitHub Action द्वारे समर्थित (स्वयंचलित आणि सदैव अद्ययावत) +#### GitHub Action द्वारे समर्थित (स्वयंचलित आणि नेहमी अद्ययावत) -[अरबी](../ar/README.md) | [बंगाली](../bn/README.md) | [बुल्गेरियन](../bg/README.md) | [बर्मी (म्यानमार)](../my/README.md) | [चिनी (सोप्या)][(./translations/zh-CN/README.md)] | [चिनी (परंपरागत, हॉंगकॉग)](../zh-HK/README.md) | [चिनी (परंपरागत, मकाउ)](../zh-MO/README.md) | [चिनी (परंपरागत, तैवान)](../zh-TW/README.md) | [क्रोएशियन](../hr/README.md) | [चेक](../cs/README.md) | [डॅनिश](../da/README.md) | [डच](../nl/README.md) | [एस्टोनियन](../et/README.md) | [फिनिश](../fi/README.md) | [फ्रेंच](../fr/README.md) | [जर्मन](../de/README.md) | [ग्रीक](../el/README.md) | [हिब्रू](../he/README.md) | [हिंदी](../hi/README.md) | [हंगेरीयन](../hu/README.md) | [इंडोनेशियन](../id/README.md) | [इटालियन](../it/README.md) | [जपानी](../ja/README.md) | [कन्नड](../kn/README.md) | [ख्मेर](../km/README.md) | [कोरियन](../ko/README.md) | [लिथुव्हेनियन](../lt/README.md) | [मलय](../ms/README.md) | [मल्याळम](../ml/README.md) | [मराठी](./README.md) | [नेपाली](../ne/README.md) | [नायजेरियन पिजिन](../pcm/README.md) | [नॉर्वेजियन](../no/README.md) | [फारसी (पर्शियन)](../fa/README.md) | [पोलिश](../pl/README.md) | [पोर्तुगिझ (ब्राझील)](../pt-BR/README.md) | [पोर्तुगिझ (पोर्तुगाल)](../pt-PT/README.md) | [पंजाबी (गुरुमुखी)](../pa/README.md) | [रोमानियन](../ro/README.md) | [रशियन](../ru/README.md) | [सर्बियन (सिरिलिक)](../sr/README.md) | [स्लोव्हाक](../sk/README.md) | [स्लोव्हेनियन](../sl/README.md) | [स्पॅनिश](../es/README.md) | [स्वाहिली](../sw/README.md) | [स्वीडिश](../sv/README.md) | [टागालॉग (फिलिपिनो)](../tl/README.md) | [तमिळ](../ta/README.md) | [तेलुगू](../te/README.md) | [थाई](../th/README.md) | [तुर्की](../tr/README.md) | [युक्रेनियन](../uk/README.md) | [उर्दू](../ur/README.md) | [व्हिएतनामी](../vi/README.md) +[अरबी](../ar/README.md) | [बंगाली](../bn/README.md) | [बुल्गेरियन](../bg/README.md) | [बर्मी (म्यानमार)](../my/README.md) | [चिनी (सरलीकृत)](../zh-CN/README.md) | [चिनी (परंपरागत, हॉंग कॉंग)](../zh-HK/README.md) | [चिनी (परंपरागत, मकाऊ)](../zh-MO/README.md) | [चिनी (परंपरागत, तैवान)](../zh-TW/README.md) | [क्रोएशियन](../hr/README.md) | [चेक](../cs/README.md) | [डॅनिश](../da/README.md) | [डच](../nl/README.md) | [एस्टोनियन](../et/README.md) | [फिनिश](../fi/README.md) | [फ्रेंच](../fr/README.md) | [जर्मन](../de/README.md) | [ग्रीक](../el/README.md) | [हिब्रू](../he/README.md) | [हिंदी](../hi/README.md) | [हंगेरीयन](../hu/README.md) | [इंडोनेशियन](../id/README.md) | [इटालियन](../it/README.md) | [जपानी](../ja/README.md) | [कन्नड](../kn/README.md) | [खमेर](../km/README.md) | [कोरियन](../ko/README.md) | [लिथुआनियन](../lt/README.md) | [मलय](../ms/README.md) | [मलयाळम](../ml/README.md) | [मराठी](./README.md) | [नेपाली](../ne/README.md) | [नायजेरियन पिडगिन](../pcm/README.md) | [नॉर्वेजियन](../no/README.md) | [फारसी (पर्शियन)](../fa/README.md) | [पोलिश](../pl/README.md) | [पोर्तुगीज (ब्राझील)](../pt-BR/README.md) | [पोर्तुगीज (पोर्तुगाल)](../pt-PT/README.md) | [पंजाबी (गुरमुखी)](../pa/README.md) | [रोमानियन](../ro/README.md) | [रशियन](../ru/README.md) | [सर्बियन (सिरिलिक)](../sr/README.md) | [स्लोव्हाक](../sk/README.md) | [स्लोव्हेनियन](../sl/README.md) | [स्पॅनिश](../es/README.md) | [स्वाहिली](../sw/README.md) | [स्वीडिश](../sv/README.md) | [टगालॉग (फिलिपिनो)](../tl/README.md) | [तामिळ](../ta/README.md) | [तेलुगू](../te/README.md) | [थाई](../th/README.md) | [तुर्किश](../tr/README.md) | [युक्रेनियन](../uk/README.md) | [उर्दू](../ur/README.md) | [व्हिएतनामीज](../vi/README.md) -> **स्थानिक पद्धतीने क्लोन करायला प्राधान्य देता?** +> **स्थानिक क्लोन प्राधान्य देता का?** > -> या रिपॉझिटरीमध्ये ५०+ भाषांमध्ये अनुवाद आहेत ज्यामुळे डाउनलोडचा आकार खूप वाढतो. भाषांतरांशिवाय क्लोन करण्यासाठी sparse checkout वापरा: +> या रेपॉसिटरीमध्ये ५०+ भाषा भाषांतरांचा समावेश आहे ज्यामुळे डाउनलोड आकार लक्षणीयरीतीने वाढतो. भाषांतरांशिवाय क्लोन करण्यासाठी sparse checkout वापरा: > > **Bash / macOS / Linux:** > ```bash @@ -45,209 +45,212 @@ Phi सध्या सर्वात शक्तिशाली आणि क > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` > -> **CMD (Windows):** +> **CMD (विंडोज):** > ```cmd > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> यामुळे तुम्हाला कोर्स पूर्ण करण्यासाठी आवश्यक सर्व काही मिळेल आणि डाउनलोडही खूप लवकर होईल. +> यामुळे कोर्स पूर्ण करण्यासाठी आवश्यक असलेले सर्व काही आपल्याला जलद डाउनलोडसह मिळेल. -## सामग्रीची सूची +## विषय सूची - परिचय - - [Phi कुटुंबात आपले स्वागत आहे](./md/01.Introduction/01/01.PhiFamily.md) + - [Phi कुटुंबात स्वागत आहे](./md/01.Introduction/01/01.PhiFamily.md) - [आपले वातावरण सेट करणे](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [महत्त्वाच्या तंत्रज्ञानांचे समजणे](./md/01.Introduction/01/01.Understandingtech.md) + - [मुख्य तंत्रज्ञान समजून घेणे](./md/01.Introduction/01/01.Understandingtech.md) - [Phi मॉडेल्ससाठी AI सुरक्षा](./md/01.Introduction/01/01.AISafety.md) - [Phi हार्डवेअर समर्थन](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi मॉडेल्स आणि प्लॅटफॉर्मवरील उपलब्धता](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Guidance-ai आणि Phi चा वापर](./md/01.Introduction/01/01.Guidance.md) + - [फळा आणि क्लाऊडवर Phi मॉडेल्स आणि उपलब्धता](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Guidance-ai आणि Phi वापरणे](./md/01.Introduction/01/01.Guidance.md) - [GitHub मार्केटप्लेस मॉडेल्स](https://github.com/marketplace/models) - [Azure AI मॉडेल कॅटलॉग](https://ai.azure.com) -- वेगवेगळ्या वातावरणात Phi चे प्रेरणा घेतलेले (Inference) - - [हगींग फेस](./md/01.Introduction/02/01.HF.md) +- वेगवेगळ्या वातावरणात Phi चे अनुमान लावणे + - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub मॉडेल्स](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry मॉडेल कॅटलॉग](./md/01.Introduction/02/03.AzureAIFoundry.md) - - [ओल्लामा](./md/01.Introduction/02/04.Ollama.md) + - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI टूलकिट VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - - [एनव्हीआयडीआयए NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - -- Phi कुटुंबाचा प्रेरणा (Inference) - - [iOS मध्ये Phi प्रेरणा](./md/01.Introduction/03/iOS_Inference.md) - - [Android मध्ये Phi प्रेरणा](./md/01.Introduction/03/Android_Inference.md) - - [Jetson मध्ये Phi प्रेरणा](./md/01.Introduction/03/Jetson_Inference.md) - - [AI PC मध्ये Phi प्रेरणा](./md/01.Introduction/03/AIPC_Inference.md) - - [Apple MLX फ्रेमवर्कसह Phi प्रेरणा](./md/01.Introduction/03/MLX_Inference.md) - - [स्थानिक सर्व्हरमध्ये Phi प्रेरणा](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI टूलकिट वापरून रिमोट सर्व्हरवर Phi प्रेरणा](./md/01.Introduction/03/Remote_Interence.md) - - [Rust सह Phi प्रेरणा](./md/01.Introduction/03/Rust_Inference.md) - - [स्थानिक Vision मध्ये Phi प्रेरणा](./md/01.Introduction/03/Vision_Inference.md) - - [Kaito AKS, Azure कंटेनर्ससह Phi प्रेरणा (अधिकृत समर्थन)](./md/01.Introduction/03/Kaito_Inference.md) + - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) + - [Foundry लोकल](./md/01.Introduction/02/07.FoundryLocal.md) + +- Phi कुटुंबाचे अनुमान लावणे + - [iOS मध्ये Phi चे अनुमान लावणे](./md/01.Introduction/03/iOS_Inference.md) + - [Android मध्ये Phi चे अनुमान लावणे](./md/01.Introduction/03/Android_Inference.md) + - [Jetson मध्ये Phi चे अनुमान लावणे](./md/01.Introduction/03/Jetson_Inference.md) + - [AI PC मध्ये Phi चे अनुमान लावणे](./md/01.Introduction/03/AIPC_Inference.md) + - [Apple MLX फ्रेमवर्कसह Phi चे अनुमान लावणे](./md/01.Introduction/03/MLX_Inference.md) + - [लोकल सर्व्हरमध्ये Phi चे अनुमान लावणे](./md/01.Introduction/03/Local_Server_Inference.md) + - [AI टूलकिट वापरून रिमोट सर्व्हरवर Phi चे अनुमान लावणे](./md/01.Introduction/03/Remote_Interence.md) + - [Rust सह Phi चे अनुमान लावणे](./md/01.Introduction/03/Rust_Inference.md) + - [लोकलमध्ये Phi--Vision चे अनुमान लावणे](./md/01.Introduction/03/Vision_Inference.md) + - [Kaito AKS, Azure कंटेनर्ससह Phi चे अनुमान लावणे (अधिकृत समर्थन)](./md/01.Introduction/03/Kaito_Inference.md) - [Phi कुटुंबाचे क्वांटिफायिंग](./md/01.Introduction/04/QuantifyingPhi.md) - - [llama.cpp वापरून Phi-3.5 / 4 चे क्वांटायझिंग](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime साठी जनरेटिव्ह AI एक्सटेंशन्स वापरून Phi-3.5 / 4 चे क्वांटायझिंग](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Intel OpenVINO वापरून Phi-3.5 / 4 चे क्वांटायझिंग](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Apple MLX फ्रेमवर्क वापरून Phi-3.5 / 4 चे क्वांटायझिंग](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Phi चे मूल्यांकन - - [उत्तरदायी AI](./md/01.Introduction/05/ResponsibleAI.md) - - [मूल्यमापनासाठी Microsoft Foundry](./md/01.Introduction/05/AIFoundry.md) + - [llama.cpp वापरून Phi-3.5 / 4 चे क्वांटायझेशन](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntime साठी Generative AI विस्तार वापरून Phi-3.5 / 4 चे क्वांटायझेशन](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Intel OpenVINO वापरून Phi-3.5 / 4 चे क्वांटायझेशन](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Apple MLX फ्रेमवर्क वापरून Phi-3.5 / 4 चे क्वांटायझेशन](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- Phi चे मूल्यमापन + - [जवाबदारी AI](./md/01.Introduction/05/ResponsibleAI.md) + - [Microsoft Foundry मूल्यमापनासाठी](./md/01.Introduction/05/AIFoundry.md) - [मूल्यमापनासाठी Promptflow वापरणे](./md/01.Introduction/05/Promptflow.md) - Azure AI Search सह RAG - - [Phi-4-mini आणि Phi-4-multimodal (RAG) सह Azure AI Search कसे वापरावे](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Phi-4-mini आणि Phi-4-multimodal(RAG) Azure AI Search सह कसे वापरायचे](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud लोकल Hybrid RAG SQLite FTS5 आणि phi-4-mini सह](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi अ‍ॅप्लिकेशन विकास नमुने - - मजकूर आणि चैट अ‍ॅप्लिकेशन्स +- Phi अनुप्रयोग विकास नमुने + - मजकूर आणि चॅट अनुप्रयोग + - Phi-4 नमुने + - [📓] [Phi-4-mini ONNX मॉडेल सह चॅट](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Phi-4 लोकल ONNX मॉडेल सह चॅट .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Sementic Kernel वापरून Phi-4 ONNX सह .NET कन्सोल अ‍ॅप चॅट](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 नमुने + - [Phi3, ONNX Runtime Web आणि WebGPU वापरून ब्राउझरमधील स्थानिक चॅटबोट](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [OpenVino चॅट](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [मल्टी मॉडेल - इंटरऍक्टिव्ह Phi-3-mini आणि OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - रॅपर तयार करणे आणि Phi-3 वापरून MLFlow सह वापरणे](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [मॉडेल ऑप्टिमायझेशन - Olive सह ONNX Runtime Web साठी Phi-3-min मॉडेल कसे ऑप्टिमाईज करावे](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Phi-3 mini-4k-instruct-onnx सह WinUI3 अॅप](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 मल्टी मॉडेल AI पॉवर्ड नोट्स अॅप नमुना](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Prompt flow सह कस्टम Phi-3 मॉडेल्सचा फाइन-ट्यून आणि एकत्रीकरण](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Microsoft Foundry मध्ये Prompt flow सह कस्टम Phi-3 मॉडेल्स फाइन-ट्यून आणि एकत्रीकरण](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Microsoft च्या जबाबदार AI तत्त्वांवर लक्ष केंद्रीत करून Microsoft Foundry मध्ये फाइन-ट्यून केलेल्या Phi-3 / Phi-3.5 मॉडेलचे मूल्यांकन](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct भाषा भाकीत नमुना (चिनी/इंग्रजी)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG चॅटबोट](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Phi-3.5-Instruct ONNX सह Windows GPU वापरून Prompt flow सोल्यूशन तयार करणे](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Microsoft Phi-3.5 tflite वापरून Android अॅप तयार करणे](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Microsoft.ML.OnnxRuntime वापरून स्थानिक ONNX Phi-3 मॉडेल वापरणारे Q&A .NET उदाहरण](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Semantic Kernel आणि Phi-3 सह Console चॅट .NET अॅप](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI Inference SDK कोड आधारित नमुने - Phi-4 नमुने - - [📓] [Phi-4-mini ONNX मॉडेलसह चैट करा](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 स्थानिक ONNX मॉडेलसह NET मध्ये चैट करा](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [सामांतिक कर्नेल वापरून Phi-4 ONNX सह .NET कन्सोल अ‍ॅपमध्ये चैट करा](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - [📓] [Phi-4-multimodal वापरून प्रोजेक्ट कोड तयार करणे](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 नमुने - - [Phi3, ONNX Runtime Web आणि WebGPU वापरून ब्राउझरमधील स्थानिक चैटबॉट](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - इंटरऐक्टिव Phi-3-मिनी आणि OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - एक रॅपर तयार करणे आणि MLFlow सह Phi-3 वापरणे](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [मॉडेल ऑप्टिमायझेशन - Olive सह ONNX Runtime Web साठी Phi-3-min मॉडेल कसे ऑप्टिमाइज़ करायचे](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [Phi-3 मिनी-4k-instruct-onnx सह WinUI3 अॅप](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 मल्टी मॉडेल AI-चालित नोट्स अॅप सॅम्पल](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [कस्टम Phi-3 मॉडेल्सचे फाइन-ट्युनिंग आणि Prompt flow सह एकत्रीकरण](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Microsoft Foundry मध्ये Prompt flow सह कस्टम Phi-3 मॉडेल्सचे फाइन-ट्युनिंग आणि एकत्रीकरण](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Microsoft च्या जबाबदार AI तत्त्वांवर लक्ष केंद्रित करून Microsoft Foundry मध्ये फाइन-ट्यून केलेल्या Phi-3 / Phi-3.5 मॉडेलचे मूल्यांकन](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-मिनी-इन्स्ट्रक्शन्स भाषिक भविष्यवाणी सॅम्पल (चिनी/इंग्रजी)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU वापरून Phi-3.5-Instruct ONNX सह Prompt flow सोल्यूशन तयार करणे](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Android अॅप तयार करण्यासाठी Microsoft Phi-3.5 tflite वापरणे](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Microsoft.ML.OnnxRuntime वापरून स्थानिक ONNX Phi-3 मॉडेलसह Q&A .NET उदाहरण](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Semantic Kernel आणि Phi-3 सह कन्सोल चॅट .NET अॅप](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI Inference SDK कोड आधारित सॅम्पल्स - - Phi-4 सॅम्पल्स - - [📓] [Phi-4-multimodal वापरून प्रोजेक्ट कोड तयार करा](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 सॅम्पल्स - - [Microsoft Phi-3 फॅमिली वापरून तुमचे स्वतःचे Visual Studio Code GitHub Copilot Chat तयार करा](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub मॉडेल्ससह Phi-3.5 वापरून तुमचा स्वतःचा Visual Studio Code Chat Copilot एजंट तयार करा](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - प्रगत तर्कशास्त्र सॅम्पल्स - - Phi-4 सॅम्पल्स - - [📓] [Phi-4-मिनी-तर्कशास्त्र किंवा Phi-4-तर्कशास्त्र सॅम्पल्स](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Microsoft Olive सह Phi-4-मिनी-तर्कशास्त्राचे फाइन-ट्युनिंग](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Apple MLX सह Phi-4-मिनी-तर्कशास्त्राचे फाइन-ट्युनिंग](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [GitHub मॉडेल्ससह Phi-4-मिनी-तर्कशास्त्र](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Microsoft Foundry मॉडेल्ससह Phi-4-मिनी-तर्कशास्त्र](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [Microsoft Phi-3 कुटुंबासह Visual Studio Code GitHub Copilot चॅट तयार करा](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub मॉडेल्सद्वारे Phi-3.5 सह तुमचा स्वतःचा Visual Studio Code चॅट कॉपिलोٽ एजंट तयार करा](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - प्रगत तर्कशास्त्र नमुने + - Phi-4 नमुने + - [📓] [Phi-4-mini-reasoning किंवा Phi-4-reasoning नमुने](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Microsoft Olive सह Phi-4-mini-reasoning फाइन-ट्यूनिंग](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLX सह Phi-4-mini-reasoning फाइन-ट्यूनिंग](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [GitHub मॉडेल्ससह Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Microsoft Foundry मॉडेल्ससह Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - डेमो - - [Phi-4-मिनी डेमो Hugging Face Spaces वर होस्ट केलेले](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal डेमो Hugging Face Spaces वर होस्ट केलेले](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - व्हिजन सॅम्पल्स - - Phi-4 सॅम्पल्स - - [📓] [Phi-4-multimodal वापरून प्रतिमा वाचा आणि कोड तयार करा](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 सॅम्पल्स - - [📓][Phi-3-व्हिजन-प्रतिमा मजकूरातून मजकूरात रुपांतरण](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-व्हिजन-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-व्हिजन CLIP एम्बेडिंग](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [डेमो: Phi-3 पुनर्वापर](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-व्हिजन - दृश्य भाषा सहाय्यक - Phi3-व्हिजन आणि OpenVINO सह](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 व्हिजन Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 व्हिजन OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 व्हिजन मल्टी-फ्रेम किंवा मल्टी-इमेज सॅम्पल](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Microsoft.ML.OnnxRuntime .NET वापरून Phi-3 व्हिजन स्थानिक ONNX मॉडेल](../../md/04.HOL/dotnet/src/LabsPhi303) - - [मेनू आधारित Phi-3 व्हिजन स्थानिक ONNX मॉडेल Microsoft.ML.OnnxRuntime .NET वापरून](../../md/04.HOL/dotnet/src/LabsPhi304) - - - तर्कशास्त्र-व्हिजन सॅम्पल्स + - [Phi-4-mini डेमो Hugging Face Spaces वर होस्ट केलेले](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal डेमो Hugginge Face Spaces वर होस्ट केलेले](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - दृष्टी नमुने + - Phi-4 नमुने + - [📓] [Phi-4-multimodal वापरून प्रतिमा वाचणे आणि कोड तयार करणे](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 नमुने + - [📓][Phi-3-vision-प्रतिमा मजकूर ते मजकूर](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vision CLIP एम्बेडिंग](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [डेमो: Phi-3 रीसायकलिंग](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - व्हिज्युअल भाषा सहाय्यक - Phi3-Vision आणि OpenVINO सह](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision मल्टी-फ्रेम किंवा मल्टी-इमेज नमुना](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision स्थानिक ONNX मॉडेल Microsoft.ML.OnnxRuntime .NET वापरून](../../md/04.HOL/dotnet/src/LabsPhi303) + - [मेनू-आधारित Phi-3 Vision स्थानिक ONNX मॉडेल Microsoft.ML.OnnxRuntime .NET वापरून](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Reasoning-Vision नमुने - Phi-4-Reasoning-Vision-15B - - [📓] [Phi-4-Reasoning-Vision-15B वापरून jaywalking ओळखणे](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B वापरून jaywalking शोधणे](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [Phi-4-Reasoning-Vision-15B वापरून गणित](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-Reasoning-Vision-15B वापरून UI ओळखणे](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B वापरून UI शोधणे](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - गणित सॅम्पल्स - - Phi-4-मिनी-फ्लॅश-तर्कशास्त्र-इन्स्ट्रक्शन सॅम्पल्स [Phi-4-मिनी-फ्लॅश-तर्कशास्त्र-इन्स्ट्रक्शनसह गणित डेमो](./md/02.Application/09.Math/MathDemo.ipynb) + - गणित नमुने + - Phi-4-Mini-Flash-Reasoning-Instruct नमुने [Phi-4-Mini-Flash-Reasoning-Instruct सह गणित डेमो](./md/02.Application/09.Math/MathDemo.ipynb) - - ऑडिओ सॅम्पल्स - - Phi-4 सॅम्पल्स + - ऑडिओ नमुने + - Phi-4 नमुने - [📓] [Phi-4-multimodal वापरून ऑडिओ ट्रान्सक्रिप्ट्स काढणे](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal ऑडिओ सॅम्पल](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal भाषांतर सॅम्पल](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET कन्सोल अॅप्लिकेशन वापरून ऑडिओ फाइलचे विश्लेषण करण्यासाठी आणि ट्रान्सक्रिप्ट तयार करण्यासाठी Phi-4-multimodal](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - MOE सॅम्पल्स - - Phi-3 / 3.5 सॅम्पल्स - - [📓] [Phi-3.5 मिक्स्चर ऑफ एक्सपर्ट्स मॉडेल्स (MoEs) सोशल मीडिया सॅम्पल](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Retrieval-Augmented Generation (RAG) पाइपलाइन तयार करणे NVIDIA NIM Phi-3 MOE, Azure AI Search, आणि LlamaIndex सह](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-4-multimodal ऑडिओ नमुना](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal भाषांतर नमुना](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET कन्सोल अॅप्लिकेशन Phi-4-multimodal वापरून ऑडिओ फाइलचे विश्लेषण आणि ट्रान्सक्रिप्ट तयार करणे](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE नमुने + - Phi-3 / 3.5 नमुने + - [📓] [Phi-3.5 मिक्सचर ऑफ एक्स्पर्ट्स मॉडेल्स (MoEs) सोशल मीडिया नमुना](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search, आणि LlamaIndex सह रिट्रीवल-ऑगमेंटेड जनरेशन (RAG) पाईपलाइन तयार करणे](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - फंक्शन कॉलिंग सॅम्पल्स - - Phi-4 सॅम्पल्स 🆕 - - [📓] [Phi-4-मिनी सह फंक्शन कॉलिंग वापरणे](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Phi-4-मिनी सह मल्टी-एजंट तयार करण्यासाठी फंक्शन कॉलिंग वापरणे](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - फंक्शन कॉलिंग नमुने + - Phi-4 नमुने 🆕 + - [📓] [Phi-4-mini सह फंक्शन कॉलिंग वापरणे](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Phi-4-mini सह मल्टि-एजंट्स तयार करण्यासाठी फंक्शन कॉलिंग वापरणे](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Ollama सह फंक्शन कॉलिंग वापरणे](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [ONNX सह फंक्शन कॉलिंग वापरणे](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - मल्टीमॉडाल मिक्सिंग सॅम्पल्स - - Phi-4 सॅम्पल्स 🆕 + - मल्टीमॉडल मिक्सिंग नमुने + - Phi-4 नमुने 🆕 - [📓] [Phi-4-multimodal तंत्रज्ञान पत्रकार म्हणून वापरणे](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET कन्सोल अॅप्लिकेशन वापरून Phi-4-multimodal वापरून प्रतिमा विश्लेषण करणे](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Phi फाइन-ट्युनिंग सॅम्पल्स - - [फाइन-ट्युनिंग संकसनार्यांसाठी](./md/03.FineTuning/FineTuning_Scenarios.md) - - [फाइन-ट्युनिंग विरुद्ध RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Phi-3 ला उद्योगातील तज्ञ बनवा फाइन-ट्युनिंग](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [AI Toolkit for VS Code सह Phi-3 चे फाइन-ट्युनिंग](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Azure मशीन लर्निंग सेवा सह Phi-3 चे फाइन-ट्युनिंग](./md/03.FineTuning/Introduce_AzureML.md) - - [Lora सह Phi-3 चे फाइन-ट्युनिंग](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora सह Phi-3 चे फाइन-ट्युनिंग](./md/03.FineTuning/FineTuning_Qlora.md) - - [Microsoft Foundry सह Phi-3 चे फाइन-ट्युनिंग](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Azure ML CLI/SDK सह Phi-3 चे फाइन-ट्युनिंग](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Microsoft Olive सह फाइन-ट्युनिंग](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive Hands-On Lab सह फाइन-ट्युनिंग](./md/03.FineTuning/olive-lab/readme.md) - - [Weights and Bias सह Phi-3-vision चे फाइन-ट्युनिंग](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Apple MLX फ्रेमवर्क सह Phi-3 चे फाइन-ट्युनिंग](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-vision (अधिकृत समर्थन) चे फाइन-ट्युनिंग](./md/03.FineTuning/FineTuning_Vision.md) - - [काइटो AKS सह Phi-3 चे फाइन-ट्यूनिंग, Azure कंटेनर्स (अधिकृत समर्थन)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Phi-3 आणि 3.5 व्हिजनचे फाइन-ट्यूनिंग](https://github.com/2U1/Phi3-Vision-Finetune) - -- हाताळणी लॅब - - [काटकनीतील मॉडेल्सचा शोध: LLMs, SLMs, स्थानिक विकास आणि बरेच काही](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP क्षमता अनलॉक करणे: Microsoft Olive सह फाइन-ट्यूनिंग](https://github.com/azure/Ignite_FineTuning_workshop) - -- शैक्षणिक संशोधन पेपर्स आणि प्रकाशने + - [.NET कन्सोल अॅप्लिकेशन Phi-4-multimodal वापरून प्रतिमा विश्लेषण करणे](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Phi नमुने फाइन-ट्यूनिंग + - [फाइन-ट्यूनिंग परिस्थिती](./md/03.FineTuning/FineTuning_Scenarios.md) + - [फाइन-ट्यूनिंग विरुद्ध RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [फाइन-ट्यूनिंग Phi-3 ला एक उद्योग तज्ञ बनवा](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [AI टूलकिटसाठी VS कोडसह Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure मशीन लर्निंग सेवा सह Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/Introduce_AzureML.md) + - [Lora सह Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_Lora.md) + - [QLora सह Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_Qlora.md) + - [Microsoft Foundry सह Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Azure ML CLI/SDK सह Phi-3 फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Microsoft Olive सह फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Microsoft Olive Hands-On Lab सह फाइन-ट्यूनिंग](./md/03.FineTuning/olive-lab/readme.md) + - [Weights and Bias सह Phi-3-vision फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Apple MLX फ्रेमवर्कसह Phi-3 चे फाइन-ट्यूनिंग](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision चे फाइन-ट्यूनिंग (अधिकृत समर्थन)](./md/03.FineTuning/FineTuning_Vision.md) + - [Kaito AKS, Azure कंटेनर्ससह Phi-3 चे फाइन-ट्यूनिंग (अधिकृत समर्थन)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Phi-3 आणि 3.5 Vision चे फाइन-ट्यूनिंग](https://github.com/2U1/Phi3-Vision-Finetune) + +- हाताळण्याचे प्रयोगशाळा + - [काटेकोर मॉडेल्सचे अन्वेषण: LLMs, SLMs, स्थानिक विकास आणि बरेच काही](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP ची क्षमता उघडणे: Microsoft Olive सह फाइन-ट्यूनिंग](https://github.com/azure/Ignite_FineTuning_workshop) + +- शैक्षणिक संशोधन पेपर आणि प्रकाशने - [Textbooks Are All You Need II: phi-1.5 तांत्रिक अहवाल](https://arxiv.org/abs/2309.05463) - - [Phi-3 तांत्रिक अहवाल: तुमच्या फोनवर स्थानिकरीत्या एक अत्यंत सक्षम भाषा मॉडेल](https://arxiv.org/abs/2404.14219) + - [Phi-3 तांत्रिक अहवाल: तुमच्या फोनवर स्थानिकपणे उच्च क्षमता असलेली भाषा मॉडेल](https://arxiv.org/abs/2404.14219) - [Phi-4 तांत्रिक अहवाल](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini तांत्रिक अहवाल: मिश्रण-ऑफ-LoRAs द्वारे संकुचित परंतु शक्तिशाली मल्टीमोडल भाषा मॉडेल्स](https://arxiv.org/abs/2503.01743) - - [वाहनातील फंक्शन-कॉलिंगसाठी लहान भाषा मॉडेल्सचे ऑप्टिमायझेशन](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) बहुपर्यायी प्रश्न उत्तर देण्यासाठी PHI-3 चे फाइन-ट्यूनिंग: पद्धतशास्त्र, परिणाम आणि आव्हाने](https://arxiv.org/abs/2501.01588) - - [Phi-4-तर्कशास्त्र तांत्रिक अहवाल](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-मिनी-तर्कशास्त्र तांत्रिक अहवाल](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Phi-4-Mini तांत्रिक अहवाल: मिश्रण-ऑफ-LoRAs द्वारे संक्षिप्त परंतु शक्तिशाली मल्टीमॉडल भाषा मॉडेल्स](https://arxiv.org/abs/2503.01743) + - [वाहनांमधील फंक्शन-कॉलिंगसाठी लहान भाषा मॉडेल्सचे ऑप्टिमायझेशन](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) मल्टिपल-चॉईस प्रश्नोत्तरेसाठी PHI-3 चे फाइन-ट्यूनिंग: पद्धतीशास्त्र, निकाल आणि आव्हाने](https://arxiv.org/abs/2501.01588) + - [Phi-4-तर्कशक्ती तांत्रिक अहवाल](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-मिनी-तर्कशक्ती तांत्रिक अहवाल](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Phi मॉडेल्सचा वापर -### Microsoft Foundry वर Phi +### Microsoft Foundry वरील Phi -तुम्ही Microsoft Phi कसा वापरायचा आणि तुमच्या वेगवेगळ्या हार्डवेअर उपकरणांमध्ये E2E सोल्यूशन्स कसे तयार करायचे हे शिकू शकता. Phi स्वतः अनुभवण्यासाठी, मॉडेल्सशी खेळण्यापासून सुरुवात करा आणि तुमच्या परिस्थितीसाठी Phi सानुकूलित करा [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) वापरून. अधिक जाणून घेण्यासाठी [Microsoft Foundry](./md/02.QuickStart/AzureAIFoundry_QuickStart.md) मधील Getting Started पहा. +तुम्ही Microsoft Phi वापरणे आणि तुमच्या वेगवेगळ्या हार्डवेअर डिव्हाइसवर E2E सोल्यूशन्स कसे तयार करायचे हे शिकू शकता. स्वतःसाठी Phi अनुभवण्यासाठी, मॉडेल्सशी खेळायला सुरुवात करा आणि Phi ला तुमच्या परिस्थितीसाठी सानुकूलित करा [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) वापरून. तुम्ही [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) पासून सुरूवात करून अधिक जाणून घेऊ शकता. -**प्लेयग्राउंड** -प्रत्येक मॉडेलसाठी समर्पित प्लेयग्राउंड आहे [Azure AI Playground](https://aka.ms/try-phi3) मध्ये मॉडेल ची चाचणी करण्यासाठी. +**प्लेग्राउंड** +प्रत्येक मॉडेलसाठी मॉडेलची चाचणी करण्यासाठी समर्पित प्लेग्राउंड आहे [Azure AI Playground](https://aka.ms/try-phi3). -### GitHub मॉडेल्सवर Phi +### GitHub मॉडेल्स वरील Phi -तुम्ही Microsoft Phi कसा वापरायचा आणि तुमच्या वेगवेगळ्या हार्डवेअर उपकरणांमध्ये E2E सोल्यूशन्स कसे तयार करायचे हे शिकू शकता. Phi स्वतः अनुभवण्यासाठी, मॉडेलशी खेळण्यापासून सुरुवात करा आणि तुमच्या परिस्थितीसाठी Phi सानुकूलित करा [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) वापरून. अधिक जाणून घेण्यासाठी [GitHub Model Catalog](./md/02.QuickStart/GitHubModel_QuickStart.md) मधील Getting Started पहा. +तुम्ही Microsoft Phi वापरणे आणि तुमच्या वेगवेगळ्या हार्डवेअर डिव्हाइसवर E2E सोल्यूशन्स कसे तयार करायचे हे शिकू शकता. स्वतःसाठी Phi अनुभवण्यासाठी, मॉडेलशी खेळायला सुरुवात करा आणि Phi ला तुमच्या परिस्थितीसाठी सानुकूलित करा [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) वापरून. तुम्ही [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) पासून सुरूवात करून अधिक जाणून घेऊ शकता. -**प्लेयग्राउंड** -प्रत्येक मॉडेलसाठी समर्पित [प्लेयग्राउंड आहे मॉडेलची चाचणी करण्यासाठी](./md/02.QuickStart/GitHubModel_QuickStart.md). +**प्लेग्राउंड** +प्रत्येक मॉडेलसाठी समर्पित [मॉडेल तपासणीसाठी प्लेग्राउंड](/md/02.QuickStart/GitHubModel_QuickStart.md) आहे. -### Hugging Face वर Phi +### Hugging Face वरील Phi -तुम्ही मॉडेल [Hugging Face](https://huggingface.co/microsoft) वरही शोधू शकता. +तुम्ही मॉडेल [Hugging Face](https://huggingface.co/microsoft) वर देखील शोधू शकता. -**प्लेयग्राउंड** - [Hugging Chat प्लेयग्राउंड](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**प्लेग्राउंड** + [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) - ## 🎒 इतर कोर्स + ## 🎒 इतर अभ्यासक्रम -आमच्या टीमने इतर कोर्सेस तयार केले आहेत! पाहा: +आमची टीम इतर अभ्यासक्रम तयार करते! तपासा: ### LangChain @@ -264,54 +267,57 @@ Phi सध्या सर्वात शक्तिशाली आणि क --- -### जनरेटिव AI सिरीज -[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### जनरेटिव AI मालिका +[![जनरेटिव AI सुरु करण्यासाठी](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![जनरेटिव AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![जनरेटिव AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![जनरेटिव AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- -### मूलभूत शिकणे -[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![सायबरसिक्युरिटीसाठी नवीन](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![वेब डेव्हसाठी नवीन](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT साठी नवीन](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR Development साठी नवीन](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +### कोअर शिक्षण +[![कॉम्प्युटर शिक्षण सुरु करण्यासाठी](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![डेटा सायन्स सुरु करण्यासाठी](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI सुरु करण्यासाठी](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![सायबरसुरक्षा सुरु करण्यासाठी](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![वेब विकास सुरु करण्यासाठी](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT सुरु करण्यासाठी](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR विकास सुरु करण्यासाठी](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### कॉपायलेट सिरीज -[![AI जोडलेल्या प्रोग्रामिंगसाठी Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +### Copilot मालिका +[![AI जोड्यांसाठी Copilot कार्यक्रम](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![C#/.NET साठी Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot साहस](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## जबाबदार AI -Microsoft आपल्या ग्राहकांना आमची AI उत्पादने जबाबदारीने वापरण्यात मदत करण्यासाठी, आमचे शिक्षण सामायिक करण्यासाठी आणि Transparency Notes आणि Impact Assessments सारख्या साधनांद्वारे विश्वासावर आधारित भागीदारी तयार करण्यासाठी वचनबद्ध आहे. अनेक हे संसाधने [https://aka.ms/RAI](https://aka.ms/RAI) येथे मिळू शकतात. -जवाबदार AI बाबतीत Microsoft ची दृष्टिकोन आमच्या AI नीतींवर आधारित आहे ज्यात नीतिमत्ता, विश्वासार्हता आणि सुरक्षा, गोपनीयता आणि सुरक्षितता, समावेशकता, पारदर्शकता आणि जबाबदारी यांचा समावेश आहे. +Microsoft आमच्या ग्राहकांना आमचे AI उत्पादने जबाबदारीने वापरण्यास मदत करण्यासाठी, आमचे अनुभव सामायिक करण्यासाठी, आणि पारदर्शकता नोट्स आणि परिणाम मूल्यांकनांसारखे साधने वापरून विश्वासावर आधारित भागीदारी तयार करण्यासाठी कटिबद्ध आहे. या संसाधनांचे अनेक प्रकार तुम्हाला [https://aka.ms/RAI](https://aka.ms/RAI) येथे मिळू शकतात. +Microsoft ची जबाबदार AI धोरण आमच्या AI तत्त्वांवर आधारित आहे ज्यात न्याय, विश्वासार्हता आणि सुरक्षितता, गोपनीयता आणि सुरक्षा, समावेशकता, पारदर्शकता, आणि जबाबदारी या गोष्टींचा समावेश आहे. + +मोठ्या प्रमाणावर नैसर्गिक भाषा, प्रतिमा, आणि भाषण मॉडेल्स - जे या नमुन्यात वापरले आहेत - अस्वीकार्य, असमर्थनीय, किंवा असभ्य वर्तन करू शकतात, ज्यामुळे त्रास होऊ शकतो. कृपया धोके आणि मर्यादा जाणून घेण्यासाठी [Azure OpenAI सेवा पारदर्शकता नोट](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) पहा. + -मोठ्या प्रमाणात नैसर्गिक भाषा, प्रतिमा, आणि भाषण मॉडेल्स - या नमुन्यात वापरल्या गेलेल्या सारख्या - कदाचित अनुचित, अविश्वसनीय किंवा आक्षेपार्ह वर्तन करू शकतात, ज्यामुळे हानी होऊ शकते. कृपया धोक्यां आणि मर्यादा जाणून घेण्यासाठी [Azure OpenAI सेवा Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) तपासा. -या जोखमी कमी करण्यासाठी शिफारस केलेली पद्धत म्हणजे आपल्या आर्किटेक्चरमध्ये एक सुरक्षितता प्रणाली समाविष्ट करणे जी हानिकारक वर्तन ओळखू आणि टाळू शकते. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) स्वतंत्र संरक्षणाच्या स्तराची पूर्तता करते, जी अनुप्रयोगांमध्ये आणि सेवांमध्ये वापरकर्त्यांनी तयार केलेले आणि AI-निर्मित हानिकारक सामग्री ओळखू शकते. Azure AI Content Safety मध्ये टेक्स्ट आणि इमेज API समाविष्ट आहेत, जे तुम्हाला हानिकारक सामग्री ओळखण्याची परवानगी देतात. Microsoft Foundry मध्ये, Content Safety सेवा तुम्हाला वेगवेगळ्या माध्यमांमध्ये हानिकारक सामग्री शोधण्यासाठी नमुना कोड पाहण्याची, शोधण्याची आणि वापरून पाहण्याची संधी देते. खालील [quickstart दस्तऐवज](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) तुम्हाला सेवेवर विनंत्या कशा करायच्या हे मार्गदर्शन करतो. +या धोके कमी करण्याचा शिफारस केलेला दृष्टिकोन म्हणजे आपल्या आर्किटेक्चरमध्ये एक सुरक्षा प्रणाली समाविष्ट करणे ज्यामुळे हानिकारक वर्तन ओळखता येईल आणि टाळता येईल. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) स्वतंत्र संरक्षणाची एक स्तर प्रदान करते, जी अनुप्रयोगांमध्ये आणि सेवा मध्ये हानिकारक वापरकर्त्याद्वारे तयार केलेला आणि AI-निर्मित सामग्री ओळखू शकते. Azure AI Content Safety मध्ये मजकूर आणि प्रतिमा APIs समाविष्ट आहेत जे तुम्हाला हानिकारक सामग्री ओळखण्याची परवानगी देतात. Microsoft Foundry मध्ये, Content Safety सेवा तुम्हाला विविध प्रकारच्या माध्यमांमध्ये हानिकारक सामग्री ओळखण्यासाठी नमुना कोड पाहण्याची, एक्सप्लोर करण्याची आणि वापरून पाहण्याची परवानगी देते. खालील [जलद प्रारंभ दस्तऐवज](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) तुम्हाला सेवेवर विनंत्या करण्यास मार्गदर्शन करतो. -दुसरा विचार करण्याचा मुद्दा म्हणजे एकूण अ‍ॅप्लिकेशन परफॉर्मन्स. मल्टी-मोडल आणि मल्टी-मॉडेल्स अ‍ॅप्लिकेशन्ससाठी, आपण परफॉर्मन्स म्हणजे की प्रणाली तुमच्याला आणि तुमच्या वापरकर्त्यांना अपेक्षित असणारी कामगिरी करते, ज्यात हानिकारक आउटपुट तयार न करणे समाविष्ट आहे, असे मानतो. [परफॉर्मन्स आणि गुणवत्ता आणि जोखमी आणि सुरक्षितता मूल्यांकनकर्ता](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) वापरून आपले संपूर्ण अ‍ॅप्लिकेशनचे कामकाज मोजणे महत्त्वाचे आहे. तुम्हाला [कस्टम मूल्यांकन करणार्‍या साधनांसह](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) तयार करण्याची आणि मोजमाप करण्याची देखील क्षमता आहे. +विचारात घेण्याचा आणखी एक पैलू म्हणजे एकूण अनुप्रयोग कार्यक्षमता. बहु-माध्यम आणि बहु-मॉडेल अनुप्रयोगांसह, आम्ही कार्यक्षमता याचा अर्थ असा घेतो की प्रणाली तुम्ही आणि तुमचे वापरकर्ते अपेक्षित करतात तशी कार्य करते, ज्यात हानिकारक आउटपुट तयार न करणे देखील समाविष्ट आहे. तुमच्या एकूण अनुप्रयोगाच्या कार्यक्षमतेचे मूल्यांकन करणे महत्त्वाचे आहे [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) वापरून. तुम्ही [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) तयार करणे आणि मूल्यांकन करण्याची क्षमता देखील ठेवता. -तुम्ही तुमच्या विकास वातावरणात [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) वापरून तुमचे AI अ‍ॅप्लिकेशन मूल्यमापन करू शकता. दिलेला टेस्ट डेटासेट किंवा टार्गेट सापडल्यावर, तुमचे जनरेटिव्ह AI अ‍ॅप्लिकेशनचे जनरेट केलेले परिणाम अंगभूत मूल्यांकन किंवा तुमच्या पसंतीनुसार कस्टम मूल्यांकन साधनांद्वारे प्रमाणात्मकपणे मोजले जातात. तुमची प्रणाली मूल्यमापन करण्यासाठी Azure AI Evaluation SDK सह सुरू करण्यासाठी, तुम्ही [quickstart मार्गदर्शक](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) फॉलो करू शकता. एकदा मूल्यमापन सत्र चालविल्यावर, तुम्ही [Microsoft Foundry मध्ये निकालांचे दृश्य पाहू शकता](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +तुम्ही तुमच्या AI अनुप्रयोगाचे मूल्यांकन तुमच्या विकास वातावरणात [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) वापरून करू शकता. एखाद्या चाचणी डेटासेट किंवा लक्ष्य दिल्यास, तुमच्या जनरेटिव्ह AI अनुप्रयोगाच्या निर्मितीचे सैद्धांतिक मोजमाप केले जातात अंगभूत मूल्यांकन करणाऱ्यांनी किंवा तुमच्या पसंतीनुसार तयार केलेल्या मूल्यांकन करणाऱ्यांनी. तुमच्या सिस्टमचे मूल्यांकन करण्यासाठी azure ai evaluation sdk सह प्रारंभ करण्यासाठी, तुम्ही [जलद प्रारंभ मार्गदर्शक](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) पालन करू शकता. एकदा तुम्ही मूल्यांकन रॅन कार्यान्वित केल्यावर, तुम्ही [Microsoft Foundry मध्ये निकालांचे दृश्य](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) पाहू शकता. -## ट्रेडमार्क +## ट्रेडमार्क्स -हा प्रकल्प प्रकल्पांचे, उत्पादनांचे किंवा सेवा यांचे ट्रेडमार्क किंवा लोगो समाविष्ट करू शकतो. Microsoft ट्रेडमार्क किंवा लोगोच्या अधिकृत वापरासाठी [Microsoft चे ट्रेडमार्क आणि ब्रँड मार्गदर्शक](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) पालनीय आहेत आणि त्यांचे पालन करणे आवश्यक आहे. Microsoft ट्रेडमार्क किंवा लोगोच्या सुधारित आवृत्त्यांमध्ये वापरामुळे गैरसमज होऊ नये किंवा Microsoft च्या प्रायोजकत्वाची भावना होऊ नये. तृतीय पक्ष ट्रेडमार्क किंवा लोगोच्या वापरासाठी त्या तृतीय पक्षाच्या धोरणांचे पालन करणे आवश्यक आहे. +हा प्रकल्प प्रकल्प, उत्पादने किंवा सेवा यांचे ट्रेडमार्क किंवा लोगो समाविष्ट करू शकतो. Microsoft ट्रेडमार्क किंवा लोगोच्या अधिकृत वापरावर [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) च्या अटी लागू होतात आणि त्या पाळाव्या लागतात. +Microsoft ट्रेडमार्क किंवा लोगोचा या प्रकल्पाच्या सुधारित आवृत्त्यांमध्ये वापर केल्यास गोंधळ निर्माण होऊ नये किंवा Microsoft च्या प्रायोजकत्वाचा भास होऊ नये. तृतीय पक्षाच्या ट्रेडमार्क किंवा लोगोचा वापर त्या तृतीय पक्षाच्या धोरणांनुसार असेल. ## मदत घेणे -तुम्हाला अडचण आल्यास किंवा AI अ‍ॅप तयार करण्याबाबत काही प्रश्न असल्यास, सामील व्हा: +तुम्हाला अडचण आली किंवा AI अनुप्रयोग तयार करण्याबाबत काही प्रश्न असतील तर, सामील व्हा: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -जर तुम्हाला उत्पादन अभिप्राय किंवा बिल्ड करताना त्रुटी आढळल्यास येथे भेट द्या: +उत्पादन अभिप्राय किंवा त्रुटी असल्यास, भेट द्या: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) @@ -319,5 +325,5 @@ Microsoft आपल्या ग्राहकांना आमची AI उ **अस्वीकरण**: -हा दस्तऐवज AI अनुवाद सेवा [Co-op Translator](https://github.com/Azure/co-op-translator) वापरून अनुवादित केला आहे. आम्ही अचूकतेसाठी प्रयत्नशील असलो तरी, कृपया लक्षात घ्या की स्वयंचलित अनुवादांमध्ये चुका किंवा अचूकतेची तूट असू शकते. मूळ दस्तऐवज त्याच्या स्थानिक भाषेत अधिकृत स्रोत मानले पाहिजे. महत्वपूर्ण माहितीसाठी व्यावसायिक मानवी अनुवाद शिफारस केली जाते. या अनुवादाच्या वापरामुळे उद्भवणाऱ्या कोणत्याही गैरसमजुती किंवा चुकीच्या अर्थसमजांसाठी आम्ही जबाबदार नाही. +हा दस्तऐवज AI भाषांतर सेवा [Co-op Translator](https://github.com/Azure/co-op-translator) चा वापर करून अनुवादित केला आहे. जरी आम्ही अचूकतेसाठी प्रयत्न करतो, तरी कृपया लक्षात घ्या की स्वयंचलित भाषांतरांमध्ये त्रुटी किंवा अचूकतेची कमतरता असू शकते. मूळ दस्तऐवज त्याच्या मूळ भाषेत अधिकृत स्रोत मानला पाहिजे. महत्त्वाची माहिती असल्यास, व्यावसायिक मानवी भाषांतराची शिफारस केली जाते. या भाषांतराच्या वापरामुळे उद्भवणाऱ्या कोणत्याही गैरसमज किंवा चुकीच्या अर्थलावणीसाठी आम्ही जबाबदार नाही. \ No newline at end of file diff --git a/translations/mr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/mr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..08f9ed719 --- /dev/null +++ b/translations/mr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 मायक्रोसॉफ्ट phi-4-mini & SQLite FTS5 सह हायब्रिड RAG (झिरो-क्लाउड SLM)\n", + "\n", + "> **लेखक:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **फोकस:** स्मॉल लँग्वेज मॉडेल्स (SLMs), SQLite FTS5 BM25, डेंस एम्बेडिंग्ज, रीसिप्रोकल रँक फ्युजन (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. प्रेरणा: स्थानिक SLMs मधील कीवर्ड रिकॉल समस्या\n", + "फक्त डेंस व्हेक्टर एम्बेडिंग्जवर अवलंबून असलेल्या मानक RAG आर्किटेक्चर्समध्ये बरोबर संख्यात्मक टोकन्स (उदा., `2,340,000 TL`, कॉन्ट्रॅक्ट कोड्स, खाते क्रमांक) शोधण्यात अयशस्वी होते. \n", + "याउलट, विरळ लेक्सिकल शोध (BM25) अर्थसूचक समानार्थक आणि परिभाषित प्रश्नांना चुकतो.\n", + "\n", + "हा कुकबुक एक **उच्च-गती, इन-मेमरी हायब्रिड रिट्रीवल इंजिन** कसा अंमलात आणायचा हे दाखवितो, ज्यामध्ये समाविष्ट आहे:\n", + "1. **डेंस व्हेक्टर** (कोसाईन समानता)\n", + "2. **विरळ लेक्सिकल शोध** (SQLite FTS5 BM25)\n", + "3. **रीसिप्रोकल रँक फ्युजन (RRF, $k=60$)**\n", + "4. मायक्रोसॉफ्ट `phi-4-mini` सह **ग्राउंडेड सिटेशन जनरेशन (`[1]`, `[2]`)**.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. द्विSQLite योजना (घन व्हेक्टर + आभासी FTS5 BM25 तक्ता)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 ३. नमुना ग्रहण आणि अंमलबजावणी तुलना\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. मायक्रोसॉफ्ट phi-4-mini साठी आधारभूत प्रॉम्प्ट फॉर्म्युलेशन\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**अस्वीकरण**:\nहा दस्तऐवज AI भाषांतर सेवा [Co-op Translator](https://github.com/Azure/co-op-translator) चा वापर करून अनुवादित केला आहे. जरी आम्ही अचूकतेसाठी प्रयत्न करतो, तरी कृपया लक्षात घ्या की स्वयंचलित भाषांतरांमध्ये त्रुटी किंवा अचूकतेची कमतरता असू शकते. मूळ दस्तऐवज त्याच्या मूळ भाषेत अधिकृत स्रोत मानला पाहिजे. महत्त्वाची माहिती असल्यास, व्यावसायिक मानवी भाषांतराची शिफारस केली जाते. या भाषांतराच्या वापरामुळे उद्भवणाऱ्या कोणत्याही गैरसमज किंवा चुकीच्या अर्थलावणीसाठी आम्ही जबाबदार नाही.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/ms/.co-op-translator.json b/translations/ms/.co-op-translator.json index 1f68de1a9..a7d884699 100644 --- a/translations/ms/.co-op-translator.json +++ b/translations/ms/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "ms" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:39:53+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:29:00+00:00", "source_file": "README.md", "language_code": "ms" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "ms" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:32:59+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "ms" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:40+00:00", diff --git a/translations/ms/README.md b/translations/ms/README.md index 0eb5552cc..b96dfc840 100644 --- a/translations/ms/README.md +++ b/translations/ms/README.md @@ -1,42 +1,42 @@ -# Buku Masakan Phi: Contoh Praktikal dengan Model Phi Microsoft +# Phi Cookbook: Contoh Amali dengan Model Phi Microsoft -[![Buka dan gunakan sampel dalam GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Buka dalam Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Buka dan gunakan sampel di GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Buka di Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![Penyumbang GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![Isu GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Permintaan Tarikan GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Dialu-alukan](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Permintaan tarik GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Selamat Datang](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![Penonton GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![Fork GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![Forks GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![Bintang GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi adalah siri model AI sumber terbuka yang dibangunkan oleh Microsoft. +Phi ialah satu siri model AI sumber terbuka yang dibangunkan oleh Microsoft. -Phi kini merupakan model bahasa kecil (SLM) yang paling berkuasa dan berkesan dari segi kos, dengan penanda aras yang sangat baik dalam pelbagai bahasa, penaakulan, penjanaan teks/bual, pengkodan, imej, audio dan senario lain. +Phi pada masa ini adalah model bahasa kecil (SLM) yang paling berkuasa dan kos efektif, dengan penanda aras yang sangat baik dalam pelbagai bahasa, penalaran, penjanaan teks/ sembang, pengkodan, imej, audio dan senario lain. -Anda boleh menyebarkan Phi ke awan atau ke peranti tepi, dan anda boleh dengan mudah membina aplikasi AI generatif dengan kuasa pengkomputeran yang terhad. +Anda boleh menggunakan Phi di awan atau di peranti tepi, dan anda boleh dengan mudah membina aplikasi AI generatif dengan kuasa pengkomputeran yang terhad. -Ikuti langkah-langkah ini untuk mula menggunakan sumber ini: -1. **Fork Repositori**: Klik [![Fork GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Klon Repositori**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Sertai Komuniti Discord AI Microsoft dan temui pakar serta pembangun lain**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +Ikuti langkah ini untuk memulakan penggunaan sumber ini: +1. **Fork Repositori**: Klik [![Forks GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Clone Repositori**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Sertai Komuniti Discord Microsoft AI dan berjumpa pakar serta pembangun lain**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![sampul](../../translated_images/ms/cover.eb18d1b9605d754b.webp) +![cover](../../translated_images/ms/cover.eb18d1b9605d754b.webp) ### 🌐 Sokongan Pelbagai Bahasa -#### Disokong melalui Tindakan GitHub (Automatik & Sentiasa Dikemas Kini) +#### Disokong melalui GitHub Action (Automatik & Sentiasa Dikemaskini) -[Arab](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgaria](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Cina (Ringkas)](../zh-CN/README.md) | [Cina (Tradisional, Hong Kong)](../zh-HK/README.md) | [Cina (Tradisional, Macau)](../zh-MO/README.md) | [Cina (Tradisional, Taiwan)](../zh-TW/README.md) | [Croatia](../hr/README.md) | [Ceko](../cs/README.md) | [Denmark](../da/README.md) | [Belanda](../nl/README.md) | [Estonia](../et/README.md) | [Finland](../fi/README.md) | [Perancis](../fr/README.md) | [Jerman](../de/README.md) | [Greek](../el/README.md) | [Ibrani](../he/README.md) | [Hindi](../hi/README.md) | [Hungaria](../hu/README.md) | [Indonesia](../id/README.md) | [Itali](../it/README.md) | [Jepun](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korea](../ko/README.md) | [Lithuania](../lt/README.md) | [Melayu](./README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepal](../ne/README.md) | [Pidgin Nigeria](../pcm/README.md) | [Norway](../no/README.md) | [Parsi (Farsi)](../fa/README.md) | [Poland](../pl/README.md) | [Portugis (Brazil)](../pt-BR/README.md) | [Portugis (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romania](../ro/README.md) | [Rusia](../ru/README.md) | [Serbia (Cyrilic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenia](../sl/README.md) | [Sepanyol](../es/README.md) | [Swahili](../sw/README.md) | [Sweden](../sv/README.md) | [Tagalog (Filipina)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turki](../tr/README.md) | [Ukraine](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnam](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](./README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **Lebih suka Klon Tempatan?** +> **Lebih suka Clone Secara Tempatan?** > -> Repositori ini merangkumi lebih 50 terjemahan bahasa yang secara signifikan meningkatkan saiz muat turun. Untuk klon tanpa terjemahan, gunakan sparse checkout: +> Repositori ini merangkumi lebih 50 terjemahan bahasa yang secara signifikan meningkatkan saiz muat turun. Untuk clone tanpa terjemahan, gunakan sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -58,17 +58,17 @@ Ikuti langkah-langkah ini untuk mula menggunakan sumber ini: ## Jadual Kandungan - Pengenalan - - [Selamat Datang ke Keluarga Phi](./md/01.Introduction/01/01.PhiFamily.md) - - [Menyediakan persekitaran anda](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Selamat datang ke Keluarga Phi](./md/01.Introduction/01/01.PhiFamily.md) + - [Menyiapkan persekitaran anda](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Memahami Teknologi Utama](./md/01.Introduction/01/01.Understandingtech.md) - [Keselamatan AI untuk Model Phi](./md/01.Introduction/01/01.AISafety.md) - [Sokongan Perkakasan Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - [Model Phi & Ketersediaan merentasi platform](./md/01.Introduction/01/01.Edgeandcloud.md) - [Menggunakan Guidance-ai dan Phi](./md/01.Introduction/01/01.Guidance.md) - - [Model Pasar GitHub](https://github.com/marketplace/models) + - [Model Marketplace GitHub](https://github.com/marketplace/models) - [Katalog Model AI Azure](https://ai.azure.com) -- Inference Phi dalam persekitaran berbeza +- Inferens Phi dalam persekitaran berbeza - [Hugging face](./md/01.Introduction/02/01.HF.md) - [Model GitHub](./md/01.Introduction/02/02.GitHubModel.md) - [Katalog Model Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) @@ -77,22 +77,22 @@ Ikuti langkah-langkah ini untuk mula menggunakan sumber ini: - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Lokal](./md/01.Introduction/02/07.FoundryLocal.md) -- Inference Keluarga Phi - - [Inference Phi pada iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Inference Phi pada Android](./md/01.Introduction/03/Android_Inference.md) - - [Inference Phi pada Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Inference Phi pada PC AI](./md/01.Introduction/03/AIPC_Inference.md) - - [Inference Phi dengan Rangka Kerja Apple MLX](./md/01.Introduction/03/MLX_Inference.md) - - [Inference Phi pada Pelayan Tempatan](./md/01.Introduction/03/Local_Server_Inference.md) - - [Inference Phi pada Pelayan Jauh menggunakan AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Inference Phi dengan Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inference Phi--Vision di Tempatan](./md/01.Introduction/03/Vision_Inference.md) - - [Inference Phi dengan Kaito AKS, Azure Containers (sokongan rasmi)](./md/01.Introduction/03/Kaito_Inference.md) +- Inferens Keluarga Phi + - [Inferens Phi dalam iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Inferens Phi dalam Android](./md/01.Introduction/03/Android_Inference.md) + - [Inferens Phi dalam Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Inferens Phi dalam PC AI](./md/01.Introduction/03/AIPC_Inference.md) + - [Inferens Phi dengan Rangka Kerja Apple MLX](./md/01.Introduction/03/MLX_Inference.md) + - [Inferens Phi dalam Server Tempatan](./md/01.Introduction/03/Local_Server_Inference.md) + - [Inferens Phi dalam Server Jauh menggunakan AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Inferens Phi dengan Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Inferens Phi--Vision secara Tempatan](./md/01.Introduction/03/Vision_Inference.md) + - [Inferens Phi dengan Kaito AKS, Kontena Azure (sokongan rasmi)](./md/01.Introduction/03/Kaito_Inference.md) - [Pengukuran Keluarga Phi](./md/01.Introduction/04/QuantifyingPhi.md) - - [Mengukur Phi-3.5 / 4 dengan llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Mengukur Phi-3.5 / 4 dengan sambungan AI Generatif untuk onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Mengukur Phi-3.5 / 4 dengan Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Mengukur Phi-3.5 / 4 dengan Rangka Kerja Apple MLX](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Pengukuran Phi-3.5 / 4 menggunakan llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Pengukuran Phi-3.5 / 4 menggunakan Sambungan AI Generatif untuk onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Pengukuran Phi-3.5 / 4 menggunakan Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Pengukuran Phi-3.5 / 4 menggunakan Rangka Kerja Apple MLX](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Penilaian Phi - [AI Bertanggungjawab](./md/01.Introduction/05/ResponsibleAI.md) @@ -100,89 +100,91 @@ Ikuti langkah-langkah ini untuk mula menggunakan sumber ini: - [Menggunakan Promptflow untuk Penilaian](./md/01.Introduction/05/Promptflow.md) - RAG dengan Azure AI Search - - [Cara menggunakan Phi-4-mini dan Phi-4-multimodal(RAG) dengan Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Cara menggunakan Phi-4-mini dan Phi-4-multimodal (RAG) dengan Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Hibrid Tempatan Zero-Awan RAG dengan SQLite FTS5 dan phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Contoh pembangunan aplikasi Phi - - Aplikasi Teks & Bualan - - Sampel Phi-4 - - [📓] [Bual dengan Model Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Bual dengan Model ONNX Phi-4 tempatan .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Aplikasi Konsol Bual .NET dengan Phi-4 ONNX menggunakan Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Sampel Phi-3 / 3.5 + - Aplikasi Teks & Sembang + - Contoh Phi-4 + - [📓] [Sembang Dengan Model Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Sembang dengan Model Phi-4 ONNX tempatan .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Aplikasi Konsol Sembang .NET dengan Phi-4 ONNX menggunakan Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Contoh - [Chatbot Tempatan dalam pelayar menggunakan Phi3, ONNX Runtime Web dan WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Model Pelbagai - Interaktif Phi-3-mini dan OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [Model Pelbagai - Phi-3-mini Interaktif dan OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Membina pembalut dan menggunakan Phi-3 dengan MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [Pengoptimuman Model - Cara mengoptimumkan model Phi-3-min untuk ONNX Runtime Web dengan Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [Aplikasi WinUI3 dengan Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Contoh Aplikasi Nota Berkuasa AI Multi Model WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Laras halus dan Integrasi model Phi-3 tersuai dengan Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Laras halus dan Integrasi model Phi-3 tersuai dengan Prompt flow di Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Menilai Model Phi-3 / Phi-3.5 yang Dilaras Halus dalam Microsoft Foundry dengan Fokus pada Prinsip AI Bertanggungjawab Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + -[Contoh Aplikasi Nota Diperkasa AI Model Pelbagai WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Laraskan halus dan Integrasi model Phi-3 tersuai dengan Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Laraskan halus dan Integrasi model Phi-3 tersuai dengan Prompt flow dalam Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Nilai Model Phi-3 / Phi-3.5 yang Dilaras Halus dalam Microsoft Foundry dengan Fokus pada Prinsip AI Bertanggungjawab Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Contoh ramalan bahasa Phi-3.5-mini-instruct (Cina/Inggeris)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Menggunakan GPU Windows untuk membuat penyelesaian Prompt flow dengan Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Menggunakan Microsoft Phi-3.5 tflite untuk membuat aplikasi Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Contoh Q&A .NET menggunakan model ONNX Phi-3 tempatan dengan Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Aplikasi chat konsol .NET dengan Semantic Kernel dan Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Contoh Kod SDK Azure AI Inference Berdasarkan Kod - - Contoh Phi-4 - - [📓] [Menjana kod projek menggunakan Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [Menggunakan Windows GPU untuk mencipta penyelesaian Prompt flow dengan Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Menggunakan Microsoft Phi-3.5 tflite untuk mencipta aplikasi Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Contoh Q&A .NET menggunakan model ONNX Phi-3 tempatan menggunakan Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Aplikasi sembang konsol .NET dengan Semantic Kernel dan Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Contoh Berasaskan Kod Azure AI Inference SDK + - Contoh Phi-4 + - [📓] [Jana kod projek menggunakan Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Contoh Phi-3 / 3.5 - - [Bina Visual Studio Code GitHub Copilot Chat anda sendiri dengan Keluarga Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Buat Ejen Chat Copilot Visual Studio Code anda sendiri dengan Phi-3.5 oleh Model GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Bina GitHub Copilot Chat Visual Studio Code anda sendiri dengan Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Cipta Ejen Copilot Sembang Visual Studio Code sendiri dengan Phi-3.5 oleh Model GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Contoh Penalaran Lanjutan - - Contoh Phi-4 - - [📓] [Contoh Phi-4-mini-penalaran atau Phi-4-penalaran](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Laraskan halus Phi-4-mini-penalaran dengan Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Laraskan halus Phi-4-mini-penalaran dengan Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-penalaran dengan Model GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-penalaran dengan Model Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demonstrasi + - Contoh Phi-4 + - [📓] [Contoh Phi-4-mini-reasoning atau Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Laraskan halus Phi-4-mini-reasoning dengan Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Laraskan halus Phi-4-mini-reasoning dengan Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-reasoning dengan Model GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-reasoning dengan Model Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Demo - [Demo Phi-4-mini dihoskan di Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - [Demo Phi-4-multimodal dihoskan di Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Contoh Penglihatan - - Contoh Phi-4 - - [📓] [Gunakan Phi-4-multimodal untuk membaca imej dan menjana kod](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Contoh Visi + - Contoh Phi-4 + - [📓] [Gunakan Phi-4-multimodal untuk membaca imej dan jana kod](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Contoh Phi-3 / 3.5 - - [📓][Phi-3-vision-Imej teks ke teks](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-Teks imej ke teks](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Kitar semula Phi-3](https://github.com/jennifermarsman/PhiRecycling/) + - [DEMO: Kitar Semula Phi-3](https://github.com/jennifermarsman/PhiRecycling/) - [Phi-3-vision - Pembantu bahasa visual - dengan Phi3-Vision dan OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Contoh berbilang bingkai atau berbilang imej Phi-3.5 Vision](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [📓][Phi-3.5 Vision contoh multi-frame atau multi-imej](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - [Model ONNX Tempatan Phi-3 Vision menggunakan Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menu Berasaskan Model ONNX Tempatan Phi-3 Vision menggunakan Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [Model ONNX Tempatan Phi-3 Vision berasaskan Menu menggunakan Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - Contoh Penalaran-Penglihatan - - Phi-4-Penalaran-Penglihatan-15B - - [📓] [Menggunakan Phi-4-Penalaran-Penglihatan-15B untuk mengesan jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Menggunakan Phi-4-Penalaran-Penglihatan-15B untuk matematik](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Menggunakan Phi-4-Penalaran-Penglihatan-15B untuk mengesan UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - Contoh Penalaran-Visi + - Phi-4-Reasoning-Vision-15B + - [📓] [Menggunakan Phi-4-Reasoning-Vision-15B untuk mengesan jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Menggunakan Phi-4-Reasoning-Vision-15B untuk matematik](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Menggunakan Phi-4-Reasoning-Vision-15B untuk mengesan UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Contoh Matematik - - Contoh Phi-4-Mini-Flash-Reasoning-Instruct [Demo Matematik dengan Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Contoh Phi-4-Mini-Flash-Reasoning-Instruct [Demo Matematik dengan Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - Contoh Audio - - Contoh Phi-4 + - Contoh Phi-4 - [📓] [Mengekstrak transkrip audio menggunakan Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Contoh Audio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Contoh Terjemahan Ucapan Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [Aplikasi konsol .NET menggunakan Phi-4-multimodal Audio untuk menganalisis fail audio dan menjana transkrip](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [Aplikasi konsol .NET menggunakan Audio Phi-4-multimodal untuk menganalisis fail audio dan menjana transkrip](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - Contoh MOE - Contoh Phi-3 / 3.5 - - [📓] [Contoh Model Campuran Pakar Phi-3.5 (MoEs) Media Sosial](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Membina Saluran Penjanaan Dipertingkatkan Pengambilan (RAG) dengan NVIDIA NIM Phi-3 MOE, Azure AI Search, dan LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Model Campuran Pakar Phi-3.5 (MoEs) Contoh Media Sosial](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Membina Saluran Retrieval-Augmented Generation (RAG) dengan NVIDIA NIM Phi-3 MOE, Azure AI Search, dan LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - Contoh Panggilan Fungsi - Contoh Phi-4 🆕 - [📓] [Menggunakan Panggilan Fungsi Dengan Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Menggunakan Panggilan Fungsi untuk membuat ejen berbilang Dengan Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Menggunakan Panggilan Fungsi untuk mencipta pelbagai ejen Dengan Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Menggunakan Panggilan Fungsi dengan Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [Menggunakan Panggilan Fungsi dengan ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - Contoh Campuran Multimodal @@ -190,60 +192,61 @@ Ikuti langkah-langkah ini untuk mula menggunakan sumber ini: - [📓] [Menggunakan Phi-4-multimodal sebagai wartawan Teknologi](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [Aplikasi konsol .NET menggunakan Phi-4-multimodal untuk menganalisis imej](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Laras halus Contoh Phi - - [Senario Laras halus](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Laras halus vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Laras halus Biarkan Phi-3 menjadi pakar industri](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Laras halus Phi-3 dengan AI Toolkit untuk VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Laras halus Phi-3 dengan Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Laras halus Phi-3 dengan Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Laras halus Phi-3 dengan QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Laras halus Phi-3 dengan Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Laras halus Phi-3 dengan Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Laras halus dengan Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Makmal Amali Laras halus dengan Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [Laras halus Phi-3-vision dengan Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Laras halus Phi-3 dengan Rangka kerja Apple MLX](./md/03.FineTuning/FineTuning_MLX.md) - - [Laras halus Phi-3-vision (sokongan rasmi)](./md/03.FineTuning/FineTuning_Vision.md) - - [Penalaan Halus Phi-3 dengan Kaito AKS, Azure Containers (Sokongan Rasmi)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Penalaan Halus Phi-3 dan 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) +- Laraskan Halus Sampel Phi + - [Senario Laraskan Halus](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Laraskan Halus vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Laraskan Halus Biarkan Phi-3 menjadi pakar industri](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Laraskan Halus Phi-3 dengan AI Toolkit untuk VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Laraskan Halus Phi-3 dengan Perkhidmatan Pembelajaran Mesin Azure](./md/03.FineTuning/Introduce_AzureML.md) + - [Laraskan Halus Phi-3 dengan Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Laraskan Halus Phi-3 dengan QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Laraskan Halus Phi-3 dengan Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Laraskan Halus Phi-3 dengan Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Laraskan Halus dengan Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Laraskan Halus dengan Makmal Praktikal Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) + - [Laraskan Halus Phi-3-vision dengan Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Penalaan Halus Phi-3 dengan Rangka Kerja Apple MLX](./md/03.FineTuning/FineTuning_MLX.md) + - [Penalaan Halus Phi-3-vision (sokongan rasmi)](./md/03.FineTuning/FineTuning_Vision.md) + - [Penalaan Halus Phi-3 dengan Kaito AKS , Kontena Azure (Sokongan Rasmi)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Penalaan Halus Phi-3 dan Vision 3.5](https://github.com/2U1/Phi3-Vision-Finetune) - Makmal Praktikal - - [Meneroka model termaju: LLMs, SLMs, pembangunan tempatan dan lain-lain](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Meneroka model terkini: LLM, SLM, pembangunan tempatan dan banyak lagi](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - [Membuka Potensi NLP: Penalaan Halus dengan Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - Kertas Penyelidikan Akademik dan Penerbitan - - [Buku Teks Adalah Semua Yang Anda Perlukan II: Laporan teknikal phi-1.5](https://arxiv.org/abs/2309.05463) - - [Laporan Teknikal Phi-3: Model Bahasa Berupaya Tinggi Secara Tempatan pada Telefon Anda](https://arxiv.org/abs/2404.14219) + - [Buku Teks Adalah Semua Yang Anda Perlu II: laporan teknikal phi-1.5](https://arxiv.org/abs/2309.05463) + - [Laporan Teknikal Phi-3: Model Bahasa Berkemampuan Tinggi Secara Tempatan di Telefon Anda](https://arxiv.org/abs/2404.14219) - [Laporan Teknikal Phi-4](https://arxiv.org/abs/2412.08905) - - [Laporan Teknikal Phi-4-Mini: Model Bahasa Multimodal Padat Tetapi Berkuasa melalui Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Laporan Teknikal Phi-4-Mini: Model Bahasa Multimodal Kompak tetapi Berkuasa melalui Campuran LoRA](https://arxiv.org/abs/2503.01743) - [Mengoptimumkan Model Bahasa Kecil untuk Panggilan Fungsi Dalam Kenderaan](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Penalaan Halus PHI-3 untuk Menjawab Soalan Pilihan Berganda: Kaedah, Keputusan, dan Cabaran](https://arxiv.org/abs/2501.01588) - - [Laporan Teknikal Phi-4-penalaran](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Laporan Teknikal Phi-4-mini-penalaran](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [(WhyPHI) Penalaan Halus PHI-3 untuk Penjawaban Soalan Pilihan Berganda: Metodologi, Keputusan, dan Cabaran](https://arxiv.org/abs/2501.01588) + - [Laporan Teknikal Rasional Phi-4](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Laporan Teknikal Rasional Mini Phi-4](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Menggunakan Model Phi -### Phi pada Microsoft Foundry +### Phi di Microsoft Foundry -Anda boleh belajar cara menggunakan Microsoft Phi dan cara membina penyelesaian E2E dalam pelbagai peranti perkakasan anda. Untuk mengalami Phi sendiri, mulakan dengan bermain dengan model dan menyesuaikan Phi untuk senario anda menggunakan [Katalog Model AI Azure Microsoft Foundry](https://aka.ms/phi3-azure-ai) anda boleh belajar lebih lanjut di Memulakan dengan [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Anda boleh mempelajari cara menggunakan Microsoft Phi dan membina penyelesaian E2E pada pelbagai peranti perkakasan anda. Untuk mengalami Phi sendiri, mulakan dengan bermain dengan model dan menyesuaikan Phi untuk senario anda menggunakan [Katalog Model AI Microsoft Foundry Azure](https://aka.ms/phi3-azure-ai) anda boleh belajar lebih lanjut di Memulakan dengan [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Tempat Latihan** -Setiap model mempunyai tempat latihan khusus untuk menguji model [Azure AI Playground](https://aka.ms/try-phi3). +**Tempat Bermain** +Setiap model mempunyai tempat bermain khusus untuk menguji model [Azure AI Playground](https://aka.ms/try-phi3). -### Phi pada Model GitHub +### Phi di Model GitHub -Anda boleh belajar cara menggunakan Microsoft Phi dan cara membina penyelesaian E2E dalam pelbagai peranti perkakasan anda. Untuk mengalami Phi sendiri, mulakan dengan bermain dengan model dan menyesuaikan Phi untuk senario anda menggunakan [Katalog Model GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) anda boleh belajar lebih lanjut di Memulakan dengan [Katalog Model GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) +Anda boleh mempelajari cara menggunakan Microsoft Phi dan membina penyelesaian E2E pada pelbagai peranti perkakasan anda. Untuk mengalami Phi sendiri, mulakan dengan bermain dengan model dan menyesuaikan Phi untuk senario anda menggunakan [Katalog Model GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) anda boleh belajar lebih lanjut di Memulakan dengan [Katalog Model GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Tempat Latihan** -Setiap model mempunyai [tempat latihan khusus untuk menguji model](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Tempat Bermain** +Setiap model mempunyai tempat bermain khusus untuk menguji model [playground](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi di Hugging Face Anda juga boleh mencari model di [Hugging Face](https://huggingface.co/microsoft) -**Tempat Latihan** - [Tempat Latihan Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Tempat Bermain** + [Tempat bermain Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) ## 🎒 Kursus Lain @@ -256,11 +259,11 @@ Pasukan kami menghasilkan kursus lain! Semak: [![LangChain untuk Pemula](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Ejen +### Azure / Edge / MCP / Agen [![AZD untuk Pemula](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI untuk Pemula](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP untuk Pemula](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Ejen AI untuk Pemula](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Agen AI untuk Pemula](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- @@ -273,7 +276,7 @@ Pasukan kami menghasilkan kursus lain! Semak: --- ### Pembelajaran Teras -[![ML untuk Pemula](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Pembelajaran Mesin untuk Pemula](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Sains Data untuk Pemula](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI untuk Pemula](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![Keselamatan Siber untuk Pemula](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) @@ -289,36 +292,38 @@ Pasukan kami menghasilkan kursus lain! Semak: [![Pengembaraan Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## AI Bertanggungjawab +## AI Bertanggungjawab Microsoft komited untuk membantu pelanggan kami menggunakan produk AI kami secara bertanggungjawab, berkongsi pembelajaran kami, dan membina perkongsian berasaskan kepercayaan melalui alat seperti Nota Ketelusan dan Penilaian Impak. Banyak sumber ini boleh didapati di [https://aka.ms/RAI](https://aka.ms/RAI). -Pendekatan Microsoft terhadap AI bertanggungjawab berasaskan prinsip AI kami iaitu keadilan, kebolehpercayaan dan keselamatan, privasi dan keselamatan, keterangkuman, ketelusan, dan akauntabiliti. +Pendekatan Microsoft terhadap AI bertanggungjawab adalah berasaskan prinsip AI kami iaitu keadilan, kebolehpercayaan dan keselamatan, privasi dan keselamatan, keterangkuman, ketelusan, dan akauntabiliti. + +Model bahasa semula jadi, imej, dan ucapan skala besar - seperti yang digunakan dalam contoh ini - berpotensi untuk berkelakuan dengan cara yang tidak adil, tidak boleh dipercayai, atau menyinggung, yang boleh menyebabkan kemudaratan. Sila rujuk [Nota Ketelusan perkhidmatan Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) untuk dimaklumkan mengenai risiko dan had. + -Model bahasa, imej, dan ucapan berskala besar - seperti yang digunakan dalam contoh ini - berpotensi berkelakuan secara tidak adil, tidak boleh dipercayai, atau menyinggung perasaan, yang boleh menyebabkan kerugian. Sila rujuk [nota Ketelusan perkhidmatan Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) untuk maklumat tentang risiko dan had. -Pendekatan yang disyorkan untuk mengurangkan risiko ini adalah dengan memasukkan sistem keselamatan dalam seni bina anda yang boleh mengesan dan menghalang tingkah laku berbahaya. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) menyediakan lapisan perlindungan bebas, mampu mengesan kandungan yang berbahaya yang dihasilkan oleh pengguna dan AI dalam aplikasi dan perkhidmatan. Azure AI Content Safety merangkumi API teks dan imej yang membolehkan anda mengesan bahan yang berbahaya. Dalam Microsoft Foundry, perkhidmatan Content Safety membolehkan anda melihat, meneroka dan mencuba contoh kod untuk mengesan kandungan berbahaya merentas pelbagai modaliti. Dokumentasi [quickstart berikut](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) membimbing anda melalui cara membuat permintaan kepada perkhidmatan tersebut. +Pendekatan yang disyorkan untuk mengurangkan risiko ini adalah dengan memasukkan sistem keselamatan dalam seni bina anda yang dapat mengesan dan menghalang tingkah laku berbahaya. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) menyediakan lapisan perlindungan bebas, mampu mengesan kandungan berbahaya yang dihasilkan oleh pengguna dan AI dalam aplikasi dan perkhidmatan. Azure AI Content Safety merangkumi API teks dan imej yang membolehkan anda mengesan bahan yang berbahaya. Dalam Microsoft Foundry, perkhidmatan Content Safety membolehkan anda melihat, meneroka dan mencuba contoh kod untuk mengesan kandungan berbahaya merentasi pelbagai modaliti. Dokumentasi [quickstart berikut](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) membimbing anda melalui permintaan kepada perkhidmatan. -Satu aspek lain yang perlu diambil kira ialah prestasi keseluruhan aplikasi. Dengan aplikasi multi-modal dan multi-model, kami anggap prestasi bermaksud sistem berfungsi seperti yang anda dan pengguna anda jangkakan, termasuk tidak menjana output yang berbahaya. Adalah penting untuk menilai prestasi keseluruhan aplikasi anda menggunakan [penilai Prestasi dan Kualiti serta Risiko dan Keselamatan](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Anda juga mempunyai keupayaan untuk mencipta dan menilai dengan [penilai khusus](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Satu lagi aspek yang perlu diambil kira adalah prestasi keseluruhan aplikasi. Dengan aplikasi multi-modal dan multi-model, kami menganggap prestasi bermaksud sistem berfungsi seperti yang anda dan pengguna anda jangkakan, termasuk tidak menghasilkan output berbahaya. Adalah penting untuk menilai prestasi aplikasi keseluruhan anda menggunakan [Penilai Prestasi dan Kualiti serta Penilai Risiko dan Keselamatan](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Anda juga mempunyai keupayaan untuk mencipta dan menilai dengan [penilai tersuai](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Anda boleh menilai aplikasi AI anda dalam persekitaran pembangunan anda menggunakan [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Berdasarkan dataset ujian atau sasaran, generasi aplikasi AI generatif anda diukur secara kuantitatif dengan penilai terbina dalam atau penilai khusus pilihan anda. Untuk memulakan dengan azure ai evaluation sdk bagi menilai sistem anda, anda boleh mengikuti [panduan quickstart](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Setelah anda menjalankan sesi penilaian, anda boleh [menvisualisasikan hasil dalam Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Anda boleh menilai aplikasi AI anda dalam persekitaran pembangunan anda menggunakan [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Dengan dataset ujian atau sasaran, generasi aplikasi AI generatif anda diukur secara kuantitatif dengan penilai terbina dalam atau penilai tersuai pilihan anda. Untuk memulakan dengan azure ai evaluation sdk untuk menilai sistem anda, anda boleh mengikuti [panduan permulaan cepat](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Setelah anda menjalankan penilaian, anda boleh [memvisualisasikan keputusan dalam Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## Tanda Dagangan +## Cap Dagangan -Projek ini mungkin mengandungi tanda dagangan atau logo untuk projek, produk, atau perkhidmatan. Penggunaan tanda dagangan atau logo Microsoft yang dibenarkan tertakluk kepada dan mesti mengikut [Garis Panduan Tanda Dagangan & Jenama Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Penggunaan tanda dagangan atau logo Microsoft dalam versi projek yang diubah suai tidak boleh menyebabkan kekeliruan atau menimbulkan tanggapan penajaan oleh Microsoft. Sebarang penggunaan tanda dagangan atau logo pihak ketiga tertakluk kepada polisi pihak ketiga tersebut. +Projek ini mungkin mengandungi cap dagangan atau logo untuk projek, produk, atau perkhidmatan. Penggunaan sah cap dagangan atau logo Microsoft tertakluk kepada dan mesti mengikuti [Garis Panduan Cap Dagangan & Jenama Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Penggunaan cap dagangan atau logo Microsoft dalam versi projek yang diubah tidak boleh menyebabkan kekeliruan atau memberikan gambaran tajaan Microsoft. Sebarang penggunaan cap dagangan atau logo pihak ketiga tertakluk kepada polisi pihak ketiga tersebut. ## Mendapatkan Bantuan -Jika anda menghadapi kesukaran atau mempunyai sebarang pertanyaan tentang membina aplikasi AI, sertai: +Jika anda tersekat atau mempunyai sebarang soalan mengenai pembinaan aplikasi AI, sertai: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Jika anda ada maklum balas produk atau ralat semasa membina, lawati: +Jika anda mempunyai maklum balas produk atau ralat semasa membina lawati: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Penafian**: -Dokumen ini telah diterjemahkan menggunakan perkhidmatan terjemahan AI [Co-op Translator](https://github.com/Azure/co-op-translator). Walaupun kami berusaha untuk ketepatan, sila ambil perhatian bahawa terjemahan automatik mungkin mengandungi kesilapan atau ketidaktepatan. Dokumen asal dalam bahasa asalnya perlu dianggap sebagai sumber yang sahih. Untuk maklumat kritikal, terjemahan profesional oleh manusia adalah disyorkan. Kami tidak bertanggungjawab atas sebarang salah faham atau salah tafsir yang timbul daripada penggunaan terjemahan ini. +**Penafian**: +Dokumen ini telah diterjemahkan menggunakan perkhidmatan terjemahan AI [Co-op Translator](https://github.com/Azure/co-op-translator). Walaupun kami berusaha untuk ketepatan, sila ambil maklum bahawa terjemahan automatik mungkin mengandungi kesilapan atau ketidaktepatan. Dokumen asal dalam bahasa asalnya harus dianggap sebagai sumber yang sahih. Untuk maklumat penting, terjemahan oleh manusia profesional adalah disyorkan. Kami tidak bertanggungjawab terhadap sebarang salah faham atau salah tafsir yang timbul daripada penggunaan terjemahan ini. \ No newline at end of file diff --git a/translations/ms/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/ms/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..9373ce54b --- /dev/null +++ b/translations/ms/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrid RAG dengan Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Pengarang:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Fokus:** Model Bahasa Kecil (SLMs), SQLite FTS5 BM25, Embeddings Padat, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivasi: Dilema Ingatan Kata Kunci dalam SLM Tempatan\n", + "Reka bentuk RAG standard yang bergantung sepenuhnya pada embeddings vektor padat sering gagal untuk mengambil token berangka tepat (contoh, `2,340,000 TL`, kod kontrak, nombor akaun). \n", + "Sebaliknya, carian leksikal jarang (BM25) terlepas sinonim semantik dan soalan parafrasa.\n", + "\n", + "Buku panduan ini menunjukkan cara melaksanakan **enjin Dapatan Hibrid berkelajuan tinggi, dalam ingatan** yang menggabungkan:\n", + "1. **Vektor Padat** (Kesamaan Kosinus)\n", + "2. **Carian Leksikal Jarang** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Penjanaan Rujukan Berasas (`[1]`, `[2]`)** dengan Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Skema SQLite Berganda (Vektor Padat + Jadual Maya FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Penanda Aras Penyerdotan & Pelaksanaan Sampel\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Penyediaan Arahan Berdasarkan Fakta untuk Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Penafian**:\nDokumen ini telah diterjemahkan menggunakan perkhidmatan terjemahan AI [Co-op Translator](https://github.com/Azure/co-op-translator). Walaupun kami berusaha untuk ketepatan, sila ambil maklum bahawa terjemahan automatik mungkin mengandungi kesilapan atau ketidaktepatan. Dokumen asal dalam bahasa asalnya harus dianggap sebagai sumber yang sahih. Untuk maklumat penting, terjemahan oleh manusia profesional adalah disyorkan. Kami tidak bertanggungjawab terhadap sebarang salah faham atau salah tafsir yang timbul daripada penggunaan terjemahan ini.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/my/.co-op-translator.json b/translations/my/.co-op-translator.json index 60595019c..68d08a2ca 100644 --- a/translations/my/.co-op-translator.json +++ b/translations/my/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "my" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:21:20+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T16:41:36+00:00", "source_file": "README.md", "language_code": "my" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "my" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T16:50:08+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "my" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:32:28+00:00", diff --git a/translations/my/README.md b/translations/my/README.md index a8ae63e6f..ea030fca7 100644 --- a/translations/my/README.md +++ b/translations/my/README.md @@ -1,42 +1,42 @@ -# Phi Cookbook: Microsoft ၏ Phi မော်ဒယ်များနှင့် လက်တွေ့နမူနာများ +# Phi စားပွဲစာအုပ်: Microsoft ရဲ့ Phi မော်ဒယ်များနှင့် လက်တွေ့ အထောက်အထားများ -[![GitHub Codespaces တွင် နမူနာများကိုဖွင့်၍ အသုံးပြုပါ](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Dev Containers တွင် ဖြင့်ပါ](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![GitHub Codespaces တွင် စမ်းသပ်မှုများ ဖွင့်၍ အသုံးပြုရန်](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Dev Containers တွင် ဖွင့်ရန်](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub ဆောင်ရွက်သူများ](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ပြဿနာများ](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub အပ်ဒိတ်တောင်းခံချက်များ](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub အပံ့ပိုးသူများ](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub အကြောင်းအရာများ](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub မဲပေးဖော်ပြချက်များ](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![PRs ကြိုဆိုပါသည်](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![GitHub ကြည့်ရှုသူများ](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub သမားပေါ်များ](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub ဗစ်](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub ကြယ်များ](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi သည် Microsoft မှဖန်တီးထားသော အမှတ်တံဆိပ်ဖြစ်ပြီး ဖွင့်လှစ်အရင်းမြစ် AI မော်ဒယ်စီးရီးတစ်ခုဖြစ်သည်။ +Phi သည် Microsoft မှ ဖန်တီးထားသော ပြည့်ဝသော မြှင့်တင်မှု AI မော်ဒယ် စီးရီးဖြစ်သည်။ -Phi သည် လက်ရှိတွင်အကြီးဆုံးစွမ်းအားဖြင့် သက်သာဆုံးသော သေးငယ်သောဘာသာစကားမော်ဒယ် (SLM) ဖြစ်ပြီး၊ ဘာသာစကားအမျိုးမျိုး၊ အတွေးအခေါ်၊ စာသား/စကားပြောထုတ်လုပ်ခြင်း၊ ကုဒ်ရေးခြင်း၊ ပုံများ၊ အသံနှင့် အခြား မ်ားစွာသောဖြစ်ရပ်များတွင် အကောင်းဆုံး စမ်းသပ်ချက်များ ရရှိထားသည်။ +Phi သည် လက်ရှိတွင် အများဆုံးစွမ်းဆောင်ရည်ရှိပြီး ကုန်ကျစရိတ်တွက် ပြီးသော သေးငယ်သော ဘာသာစကားမော်ဒယ်(SLM) ဖြစ်ပြီး၊ ဘာသာစကားများစွာ၊ မှတ်ချက်ပေးခြင်း၊ စာသား/စကားပြော များထုတ်လုပ်ခြင်း၊ ကုဒ်ရေးခြင်း၊ ရုပ်ပုံ၊ အသံနှင့် အခြားဖြစ်ရပ်များတွင် အလွန်ကောင်းမွန်သော စမ်းသပ်မှုများရှိပါသည်။ -Phi ကို Cloud သို့မဟုတ် ဒြပ်စင်ကိရိယာများတွင် တပ်ဆင်နိုင်ပြီး ကွန်ပျူတာစွမ်းအား ကန့်သတ်ချက် ရှိသော်လည်း လွယ်ကူစွာ စိတ်ကူးမှန် AI အပလီကေးရှင်းများ တည်ဆောက်နိုင်သည်။ +Phi ကို ကောင်းကင် မိုဃ်းဆီသို့ သို့မဟုတ် အနားတွင် ကြိုးမဲ့စက်ပစ္စည်းများတွင် တပ်ဆင်နိုင်ပြီး၊ ကွန်ပျူတာ စွမ်းအားကန့်သတ်ချက်နည်းပါးသည့် အခြေအနေတွင် generative AI အက်ပလီကေးရှင်းများ လွယ်ကူစွာ တည်ဆောက်နိုင်ပါသည်။ -ဒီအရင်းအမြစ်များကို အသုံးပြု၍ စတင်ရန် အဆင့်များကို လိုက်နာပါ။ -1. **Repository ကို Fork လုပ်ပါ**: Click [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +ဤ စနစ်များကို အသုံးပြုရန် အစပြုခြင်းအတွက် အောက်ပါအဆင့်များကိုလိုက်နာပါ - +1. **Repository ကို Fork လုပ်ပါ**: နှိပ်ပါ [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Repository ကို Clone လုပ်ပါ**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Microsoft AI Discord လူမှုအသိုင်းအဝိုင်းတွင် ပါဝင်၍ ကျွမ်းကျင်သူများနှင့် တက်ကြွဖော်ရွေများနှင့် တွေ့ဆုံပါ**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Microsoft AI Discord နဲ့ ထပ်တူဆောင်ရွက်သူများနှင့် တွေ့ဆုံဖို့ လက်ခံပါ**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/my/cover.eb18d1b9605d754b.webp) -### 🌐 ဘာသာစကားစုံပံ့ပိုးမှု +### 🌐 ဘာသာစကားများစွာ ထောက်ပံ့မှု -#### GitHub Action မှတဆင့် ပံ့ပိုးထား (အလိုအလျောက် နှင့် အမြဲတမ်း အပ်ဒိတ်ဖြစ်) +#### GitHub Action ဖြင့် ထောက်ပံ့သည် (အလိုအလျောက်နှင့် အမြဲနောက်ဆုံးပေါ်ဖြစ်နေသည်) [Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](./README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **ဒေသခံစကာင်လုပ်ဖို့ ပိုနှစ်သက်ပါသလား?** +> **တစ်ကိုယ်တည်း ဒေါင်းလုပ်လုပ်လိုပါသလား?** > -> ဤ repository မှာ ဘာသာစကား ၅၀ ကျော် အနက် များစွာသော ဘာသာပြန်ချက်များပါရှိသည်။ ၎င်းကြောင့် ဒေါင်းလုဒ်အရွယ်အစား ကြီးပွားပါသည်။ ဘာသာပြန်ချက်များမပါဘဲ clone လုပ်ချင်ပါက sparse checkout ကို အသုံးပြုပါ။ +> ဤ repository တွင် ဘာသာစကား ၅၀ ကျော်ကို ဘာသာပြန်ထားသောကြောင့် ဒေါင်းလုပ်အရွယ်အစား မြင့်မားပါသည်။ ဘာသာပြန်မှုများမပါဘဲ clone လုပ်ရန် sparse checkout ကို အသုံးပြုနိုင်ပါသည်။ > > **Bash / macOS / Linux:** > ```bash @@ -52,173 +52,176 @@ Phi ကို Cloud သို့မဟုတ် ဒြပ်စင်ကိရ > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> ၎င်းသည် သင်တန်းကို အရှိန်မြန်စွာ ပြီးမြောက်စေရန် လိုအပ်သည့် အရာအားလုံးကို ပေးသည်။ +> ဒါကြောင့် သင့်အတွက် လိုအပ်သမျှအရာအားလုံးကို ပိုမိုလျင်မြန်စွာ ဒေါင်းလုပ်လုပ်နိုင်မည်ဖြစ်သည်။ -## အကြောင်းအရာဇယား +## အကြောင်းအရာ စာရင်း -- နိဒါန်း - - [Phi မိသားစုမှ ကြိုဆိုပါသည်](./md/01.Introduction/01/01.PhiFamily.md) - - [သင့်ပတ်ဝန်းကျင် စီစဉ်ခြင်း](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [အဓိကနည်းပညာများကို နားလည်မှု](./md/01.Introduction/01/01.Understandingtech.md) - - [Phi မော်ဒယ်များအတွက် AI လုံခြုံမှု](./md/01.Introduction/01/01.AISafety.md) - - [Phi ဟာ့ဒ်ဝဲ ပံ့ပိုးမှု](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi မော်ဒယ်များနှင့် ပလက်ဖောင်းအလိုက်ရရှိနိုင်မှု](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Guidance-ai နှင့် Phi အသုံးပြုခြင်း](./md/01.Introduction/01/01.Guidance.md) +- အစပြုချက် + - [Phi မိသားစုမှ ကြိုဆိုပါတယ်](./md/01.Introduction/01/01.PhiFamily.md) + - [သင်၏ ပတ်ဝန်းကျင်ကို အဆင်သင့် ပြုလုပ်ခြင်း](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [အဓိက နည်းပညာများကို နားလည်ခြင်း](./md/01.Introduction/01/01.Understandingtech.md) + - [Phi မော်ဒယ်များအတွက် AI ဘေးကင်းလုံခြုံမှု](./md/01.Introduction/01/01.AISafety.md) + - [Phi ပစ္စည်းကိရိယာ ထောက်ပံ့မှု](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Phi မော်ဒယ်များနှင့် ပလက်ဖောင်းများတွင်ရရှိနိုင်မှု](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Guidance-aiနှင့် Phi ကို အသုံးပြုခြင်း](./md/01.Introduction/01/01.Guidance.md) - [GitHub Marketplace မော်ဒယ်များ](https://github.com/marketplace/models) - - [Azure AI မော်ဒယ်ကက်တားလော့(စ်)](https://ai.azure.com) + - [Azure AI Model ကတ်တာလော့(စာမျက်နှာ)](https://ai.azure.com) -- မတူညီသော ပတ်ဝန်းကျင်များတွင် Phi ကို အကြံပြုခြင်း +- စိတ်ကူးကြံဆပြီး Phi ကို မတူညီသော ပတ်ဝန်းကျင်များတွင် အသုံးပြုခြင်း - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub မော်ဒယ်များ](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry မော်ဒယ်ကက်တားလော့](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Microsoft Foundry Model ကတ်တာလော့](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Phi မိသားစု အကြံပြုခြင်း - - [iOS တွင် Phi အကြံပြုခြင်း](./md/01.Introduction/03/iOS_Inference.md) - - [Android တွင် Phi အကြံပြုခြင်း](./md/01.Introduction/03/Android_Inference.md) - - [Jetson တွင် Phi အကြံပြုခြင်း](./md/01.Introduction/03/Jetson_Inference.md) - - [AI PC တွင် Phi အကြံပြုခြင်း](./md/01.Introduction/03/AIPC_Inference.md) - - [Apple MLX Framework နှင့် Phi အကြံပြုခြင်း](./md/01.Introduction/03/MLX_Inference.md) - - [ဒေသခံ ဆာဗာတွင် Phi အကြံပြုခြင်း](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI Toolkit ဖြင့် Remote ဆာဗာတွင် Phi အကြံပြုခြင်း](./md/01.Introduction/03/Remote_Interence.md) - - [Rust ဖြင့် Phi အကြံပြုခြင်း](./md/01.Introduction/03/Rust_Inference.md) - - [ဒေသခံတွင် Phi--Vision အကြံပြုခြင်း](./md/01.Introduction/03/Vision_Inference.md) - - [Kaito AKS, Azure Containers (တရားဝင်ပံ့ပိုးမှု) ဖြင့် Phi အကြံပြုခြင်း](./md/01.Introduction/03/Kaito_Inference.md) -- [Phi မိသားစုကို Quantify လုပ်ခြင်း](./md/01.Introduction/04/QuantifyingPhi.md) - - [llama.cpp ဖြင့် Phi-3.5 / 4 ကို Quantize လုပ်ခြင်း](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime အတွက် Generative AI ပေါင်းစပ်ခြင်းဖြင့် Phi-3.5 / 4 ကို Quantize လုပ်ခြင်း](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Intel OpenVINO ဖြင့် Phi-3.5 / 4 ကို Quantize လုပ်ခြင်း](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Apple MLX Framework ဖြင့် Phi-3.5 / 4 ကို Quantize လုပ်ခြင်း](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Phi အကဲဖြတ်ခြင်း - - [တာဝန်ရှိသော AI](./md/01.Introduction/05/ResponsibleAI.md) +- Phi မိသားစုကို အသုံးပြုခြင်း + - [iOS တွင် Phi အသုံးပြုခြင်း](./md/01.Introduction/03/iOS_Inference.md) + - [Android တွင် Phi အသုံးပြုခြင်း](./md/01.Introduction/03/Android_Inference.md) + - [Jetson တွင် Phi အသုံးပြုခြင်း](./md/01.Introduction/03/Jetson_Inference.md) + - [AI PC တွင် Phi အသုံးပြုခြင်း](./md/01.Introduction/03/AIPC_Inference.md) + - [Apple MLX Framework ဖြင့် Phi အသုံးပြုခြင်း](./md/01.Introduction/03/MLX_Inference.md) + - [ဒေသိယဆာဗာတွင် Phi အသုံးပြုခြင်း](./md/01.Introduction/03/Local_Server_Inference.md) + - [AI Toolkit အသုံးပြု၍ အဝေးဆာဗာတွင် Phi အသုံးပြုခြင်း](./md/01.Introduction/03/Remote_Interence.md) + - [Rust ဖြင့် Phi အသုံးပြုခြင်း](./md/01.Introduction/03/Rust_Inference.md) + - [ဒေသိယတွင် Phi--Vision အသုံးပြုခြင်း](./md/01.Introduction/03/Vision_Inference.md) + - [Kaito AKS, Azure Containers (တရားဝင်ထောက်ခံမှု) ဖြင့် Phi အသုံးပြုခြင်း](./md/01.Introduction/03/Kaito_Inference.md) +- [Phi မိသားစု အရေအတွက်တွက်ချက်ခြင်း](./md/01.Introduction/04/QuantifyingPhi.md) + - [llama.cpp ဖြင့် Phi-3.5 / 4 အရေအတွက် ခန့်မှန်းခြင်း](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntime အတွက် Generative AI အာရုံအသုံးပြု၍ Phi-3.5 / 4 အရေအတွက် ခန့်မှန်းခြင်း](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Intel OpenVINO အသုံးပြု၍ Phi-3.5 / 4 အရေအတွက် ခန့်မှန်းခြင်း](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Apple MLX Framework ဖြင့် Phi-3.5 / 4 အရေအတွက် ခန့်မှန်းခြင်း](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- Phi ကို အကဲဖြတ်ခြင်း + - [အသစ်တင် AI ကို တာဝန်ရှိစွာ သုံးစွဲရန်](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry ဖြင့် အကဲဖြတ်ခြင်း](./md/01.Introduction/05/AIFoundry.md) - [Promptflow အသုံးပြု၍ အကဲဖြတ်ခြင်း](./md/01.Introduction/05/Promptflow.md) -- Azure AI ရှာဖွေမှုနှင့် RAG - - [Phi-4-mini နှင့် Phi-4-multimodal (RAG) ကို Azure AI ရှာဖွေမှုနှင့် သုံးနည်း](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- Azure AI Search ဖြင့် RAG + - [Phi-4-mini နှင့် Phi-4-multimodal(RAG) ကို Azure AI Search နှင့် တွဲသုံးနည်း](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [SQLite FTS5 နှင့် phi-4-mini အသုံးပြုသော Zero-Cloud ဒေသိယ Hybrid RAG](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi အပလီကေးရှင်း ဖွံ့ဖြိုးမှု နမူနာများ - - စာသားနှင့် စကားပြော အပလီကေးရှင်းများ +- Phi အက်ပလီကေးရှင်း ဖန်တီးခြင်းနမူနာများ + - စာသားနှင့် စကားပြော အက်ပလီကေးရှင်းများ - Phi-4 နမူနာများ - - [📓] [Phi-4-mini ONNX မော်ဒယ်ဖြင့် စကားပြော](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 ဒေသခံ ONNX မော်ဒယ်နှင့် စကားပြော .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Semantic Kernel အသုံးပြု၍ Phi-4 ONNX နှင့် .NET console app တွင် စကားပြော](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - [📓] [Phi-4-mini ONNX မော်ဒယ်နှင့် စကားပြော](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Phi-4 ဒေသိယ ONNX မော်ဒယ်နှင့် စကားပြော .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Phi-4 ONNX ဖြင့် Semantic Kernel အသုံးပြုသော .NET console app စကားပြော](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 နမူနာများ + - [Phi3, ONNX Runtime Web နှင့် WebGPU အသုံးပြုသော ဘရောက်ဇာတွင် နေရာတိုင်းတွင်ရှိသော စကားပြောစက်](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [OpenVino စကားပြော](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [အမျိုးမျိုးသော မော်ဒယ်များ - ဆက်သွယ်မှုရှိသော Phi-3-mini နှင့် OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Phi-3 နှင့် MLFlow ကိုအသုံးပြု၍ wrapper တည်ဆောက်ခြင်း](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [မော်ဒယ် အဆင်ပြေမှု - Olive ဖြင့် ONNX Runtime Web အတွက် Phi-3-min မော်ဒယ် အဆင်ပြေစေရန် မည်သို့လုပ်မည်](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 အက်ပ် Phi-3 mini-4k-instruct-onnx ဖြင့်](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 အမျိုးမျိုး မော်ဒယ် AI ထောက်ပံ့မှုပတ်သက်သော မှတ်စု အက်ပ် နမူနာ](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Prompt flow နှင့် စိတ်ကြိုက် Phi-3 မော်ဒယ်များကို fine-tune နှင့် ပေါင်းစပ်ခြင်း](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Microsoft Foundry တွင် Prompt flow ဖြင့် စိတ်ကြိုက် Phi-3 မော်ဒယ်များကို fine-tune နှင့် ပေါင်းစပ်ခြင်း](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Microsoft ရဲ့ တာဝန်ယူသော AI သဘောတူညီချက်များအပေါ် အာရုံစိုက်၍ Microsoft Foundry တွင် Fine-tuned Phi-3 / Phi-3.5 မော်ဒယ်ကို သုံးသပ်ခြင်း](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct ဘာသာစကားခန့်မှန်းမှု နမူနာ (တရုတ်/အင်္ဂလိပ်)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG စကားပြောစက်](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Windows GPU အသုံးပြု၍ Phi-3.5-Instruct ONNX ဖြင့် Prompt flow ဖြေရှင်းချက် ဖန်တီးခြင်း](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Microsoft Phi-3.5 tflite ကို အသုံးပြု၍ Android အက်ပ် ဖန်တီးခြင်း](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Microsoft.ML.OnnxRuntime ကို အသုံးပြု၍ ဒေသတွင်း ONNX Phi-3 မော်ဒယ်ဖြင့် Q&A .NET နမူနာ](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Semantic Kernel နှင့် Phi-3 ဖြင့် console chat .NET အက်ပ်](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI Inference SDK ကုဒ်အခြေပြု နမူနာများ + - Phi-4 နမူနာများ + - [📓] [Phi-4-multimodal ကို အသုံးပြု၍ ပရောဂျက်ကုဒ် ထုတ်လုပ်ခြင်း](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Phi-3 / 3.5 နမူနာများ + - [Microsoft Phi-3 မိသားစုနှင့် ကိုယ်ပိုင် Visual Studio Code GitHub Copilot စကားပြော ဖန်တီးခြင်း](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub မော်ဒယ်များဖြင့် Phi-3.5 ဖြင့် ကိုယ်ပိုင် Visual Studio Code စကားပြော Copilot Agent ဖန်တီးခြင်း](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - အဆင့်မြင့် အကြံပြုမှု နမူနာများ + - Phi-4 နမူနာများ + - [📓] [Phi-4-mini-reasoning သို့မဟုတ် Phi-4-reasoning နမူနာများ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Microsoft Olive ဖြင့် Phi-4-mini-reasoning ကို fine-tune လုပ်ခြင်း](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLX ဖြင့် Phi-4-mini-reasoning ကို fine-tune လုပ်ခြင်း](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [GitHub မော်ဒယ်များဖြင့် Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Microsoft Foundry မော်ဒယ်များဖြင့် Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - ပြသနာများ + - [Phi-4-mini ပြသနာများကို Hugging Face Spaces တွင် ဟိုစ့်လုပ်ခြင်း](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal ပြသနာများကို Hugginge Face Spaces တွင် ဟိုစ့်လုပ်ခြင်း](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - တွေ့မြင်နိုင်ခြင်း နမူနာများ + - Phi-4 နမူနာများ + - [📓] [Phi-4-multimodal ကို အသုံးပြု၍ ပုံများ ဖတ်ရှုခြင်းနှင့် ကုဒ် ထုတ်လုပ်ခြင်း](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 နမူနာများ - - [Phi3, ONNX Runtime Web နှင့် WebGPU အသုံးပြုပြီး ဒေါ်ဘရေဇာတွင် ဒေသခံ Chatbot](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - Interactive Phi-3-mini and OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Building a wrapper and using Phi-3 with MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Model Optimization - How to optimize Phi-3-min model for ONNX Runtime Web with Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 App with Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Multi Model AI Powered Notes App Sample](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Fine-tune and Integrate custom Phi-3 models with Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Fine-tune and Integrate custom Phi-3 models with Prompt flow in Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Evaluate the Fine-tuned Phi-3 / Phi-3.5 Model in Microsoft Foundry Focusing on Microsoft's Responsible AI Principles](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct language prediction sample (Chinese/English)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Using Windows GPU to create Prompt flow solution with Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Using Microsoft Phi-3.5 tflite to create Android app](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET Example using local ONNX Phi-3 model using the Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Console chat .NET app with Semantic Kernel and Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI Inference SDK Code Based Samples - - Phi-4 Samples - - [📓] [Generate project code using Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 Samples - - [Build your own Visual Studio Code GitHub Copilot Chat with Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Create your own Visual Studio Code Chat Copilot Agent with Phi-3.5 by GitHub Models](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - Advanced Reasoning Samples - - Phi-4 Samples - - [📓] [Phi-4-mini-reasoning or Phi-4-reasoning Samples](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Fine-tuning Phi-4-mini-reasoning with Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Fine-tuning Phi-4-mini-reasoning with Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning with GitHub Models](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning with Microsoft Foundry Models](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demos - - [Phi-4-mini demos hosted on Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demos hosted on Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Vision Samples - - Phi-4 Samples - - [📓] [Use Phi-4-multimodal to read images and generate code](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 Samples - - [📓][Phi-3-vision-Image text to text](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-ပုံ မှ စာ ရေးသားခြင်းသို့ စာသား ပြောင်းခြင်း](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 Recycling](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Visual language assistant - with Phi3-Vision and OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [📓][Phi-3-vision CLIP စားလုံးချက်ထည့်ခြင်း](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [ပြသနာ: Phi-3 ပြန်လည်အသုံးပြုခြင်း](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - တွေ့မြင်နိုင်သော ဘာသာစကားအကူအညီ - Phi3-Vision နှင့် OpenVINO ဖြင့်](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision multi-frame or multi-image sample](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision Local ONNX Model using the Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menu based Phi-3 Vision Local ONNX Model using the Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Reasoning-Vision Samples - - Phi-4-Reasoning-Vision-15B - - [📓] [Using Phi-4-Reasoning-Vision-15B to detect jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Using Phi-4-Reasoning-Vision-15B to math](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Using Phi-4-Reasoning-Vision-15B to detect UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - Math Samples - - Phi-4-Mini-Flash-Reasoning-Instruct Samples [Math Demo with Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - - Audio Samples - - Phi-4 Samples - - [📓] [Extracting audio transcripts using Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal Audio Sample](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal Speech Translation Sample](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET console application using Phi-4-multimodal Audio to analyze an audio file and generate transcript](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - MOE Samples - - Phi-3 / 3.5 Samples - - [📓] [Phi-3.5 Mixture of Experts Models (MoEs) Social Media Sample](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Building a Retrieval-Augmented Generation (RAG) Pipeline with NVIDIA NIM Phi-3 MOE, Azure AI Search, and LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓][Phi-3.5 Vision များစွာသော frame သို့မဟုတ် များစွာသော ပုံ နမူနာ](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Microsoft.ML.OnnxRuntime .NET ဖြင့် Phi-3 Vision ဒေသတွင်း ONNX မော်ဒယ်](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Menu အခြေပြု Phi-3 Vision ဒေသတွင်း ONNX မော်ဒယ် Microsoft.ML.OnnxRuntime .NET ဖြင့်](../../md/04.HOL/dotnet/src/LabsPhi304) + + - အကြံပြုမှု-တွေ့မြင်နိုင်ခြင်း နမူနာများ + - Phi-4-Reasoning-Vision-15B + - [📓] [Phi-4-Reasoning-Vision-15B အသုံးပြု၍ လမ်းလွှဲခြင်း မှတ်ထားခြင်း](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B အား သင်္ချာတွက်ချက်မှုအတွက် အသုံးပြုခြင်း](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B အသုံးပြု၍ UI မှတ်ယူခြင်း](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - သင်္ချာ နမူနာများ + - Phi-4-Mini-Flash-Reasoning-Instruct နမူနာများ [Phi-4-Mini-Flash-Reasoning-Instruct နမူနာနှင့် သင်္ချာပြသမှု](./md/02.Application/09.Math/MathDemo.ipynb) + + - အသံ နမူနာများ + - Phi-4 နမူနာများ + - [📓] [Phi-4-multimodal ဖြင့် အသံစာသား ထုတ်ယူခြင်း](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal အသံ နမူနာ](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal စကားပြန် နမူနာ](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET console app ဖြင့် Phi-4-multimodal အသံကို သုံး၍ အသံဖိုင်ခွဲခြမ်းစိတ်ဖြာခြင်းနှင့် စာသားထုတ်ခြင်း](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE နမူနာများ + - Phi-3 / 3.5 နမူနာများ + - [📓] [Phi-3.5 Mixture of Experts Models (MoEs) လူမှုမီဒီယာ နမူနာ](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search နှင့် LlamaIndex ဖြင့် Retrieval-Augmented Generation (RAG) Pipeline တည်ဆောက်ခြင်း](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Function Calling Samples - - Phi-4 Samples 🆕 - - [📓] [Using Function Calling With Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Using Function Calling to create multi-agents With Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Using Function Calling with Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Using Function Calling with ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Multimodal Mixing Samples - - Phi-4 Samples 🆕 - - [📓] [Using Phi-4-multimodal as a Technology journalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET console application using Phi-4-multimodal to analyze images](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Fine-tuning Phi Samples - - [Fine-tuning Scenarios](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Fine-tuning vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Fine-tuning Let Phi-3 become an industry expert](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Fine-tuning Phi-3 with AI Toolkit for VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Fine-tuning Phi-3 with Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Fine-tuning Phi-3 with Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Fine-tuning Phi-3 with QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Fine-tuning Phi-3 with Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Fine-tuning Phi-3 with Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Fine-tuning with Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Fine-tuning with Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - - [Fine-tuning Phi-3-vision with Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Fine-tuning Phi-3 with Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Fine-tuning Phi-3-vision (official support)](./md/03.FineTuning/FineTuning_Vision.md) - - [Kaito AKS နှင့် Phi-3 ကို Fine-Tuning ပြုလုပ်ခြင်း၊ Azure Containers(အတည်ပြုမှု ရရှိထား)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Phi-3 နှင့် 3.5 Vision Fine-Tuning](https://github.com/2U1/Phi3-Vision-Finetune) - -- လက်တွေ့လေ့ကျင့်ခန်း - - [နောက်ဆုံးပေါ်မော်ဒယ်များကို လေ့လာခြင်း: LLMs, SLMs, ဒေသတွင်း ဖွံ့ဖြိုးတိုးတက်မှုနှင့် အခြားများ](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP အာဏာကို ဖွင့်လှစ်ခြင်း: Microsoft Olive ဖြင့် Fine-Tuning](https://github.com/azure/Ignite_FineTuning_workshop) - -- အကောလိပ်သုတေသနစာတမ်းများနှင့် ထုတ်ဝေမှုများ - - [Textbooks Are All You Need II: phi-1.5 နည်းပညာအစီရင်ခံစာ](https://arxiv.org/abs/2309.05463) - - [Phi-3 နည်းပညာအစီရင်ခံစာ: သင့်ဖုန်းပေါ်တွင် ဒေသဆိုင်ရာ အွန်လိုင်း စကားဝိုင်းမော်ဒယ်](https://arxiv.org/abs/2404.14219) + - Function Calling နမူနာများ + - Phi-4 နမူနာများ 🆕 + - [📓] [Phi-4-mini နှင့် Function Calling အသုံးပြုခြင်း](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Phi-4-mini နှင့် Function Calling အသုံးပြု၍ multi-agents ဖန်တီးခြင်း](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Ollama နှင့် Function Calling အသုံးပြုခြင်း](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [ONNX နှင့် Function Calling အသုံးပြုခြင်း](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Multimodal Mixing နမူနာများ + - Phi-4 နမူနာများ 🆕 + - [📓] [Phi-4-multimodal ကို နည်းပညာ သတင်းထောက် အဖြစ် အသုံးပြုခြင်း](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET console app ဖြင့် Phi-4-multimodal အသုံးပြုကာ ပုံများ ခွဲခြမ်းစိတ်ဖြာခြင်း](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Phi ကို Fine-tuning လုပ်ခြင်း + - [Fine-tuning အခြေအနေများ](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Fine-tuning နှင့် RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Phi-3 ကို အလုပ်အကိုင်များတွင် ကျွမ်းကျင်သူ ဖြစ်လာစေနိုင်ရန် Fine-tuning လုပ်ခြင်း](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [AI Toolkit for VS Code ဖြင့် Phi-3 ကို Fine-tuning လုပ်ခြင်း](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure Machine Learning Service ဖြင့် Phi-3 ကို Fine-tuning လုပ်ခြင်း](./md/03.FineTuning/Introduce_AzureML.md) + - [Lora ဖြင့် Phi-3 ကို Fine-tuning လုပ်ခြင်း](./md/03.FineTuning/FineTuning_Lora.md) + - [QLora ဖြင့် Phi-3 ကို Fine-tuning လုပ်ခြင်း](./md/03.FineTuning/FineTuning_Qlora.md) + - [Microsoft Foundry နှင့် Phi-3 ကို Fine-tuning လုပ်ခြင်း](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Azure ML CLI/SDK ဖြင့် Phi-3 ကို Fine-tuning လုပ်ခြင်း](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Microsoft Olive ဖြင့် Fine-tuning လုပ်ခြင်း](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Microsoft Olive Hands-On Lab ဖြင့် Fine-tuning လုပ်ခြင်း](./md/03.FineTuning/olive-lab/readme.md) + - [Weights and Bias ဖြင့် Phi-3-vision ကို Fine-tuning လုပ်ခြင်း](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Apple MLX Framework ဖြင့် Phi-3 ကို ပြုပြင်ပြောင်းလဲခြင်း](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision ကို ပြုပြင်ပြောင်းလဲခြင်း (တရားဝင် ထောက်ခံမှု)](./md/03.FineTuning/FineTuning_Vision.md) + - [Kaito AKS နှင့် Azure Containers ဖြင့် Phi-3 ကို ပြုပြင်ပြောင်းလဲခြင်း (တရားဝင် ထောက်ခံမှု)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Phi-3 နှင့် 3.5 Vision ကို ပြုပြင်ပြောင်းလဲခြင်း](https://github.com/2U1/Phi3-Vision-Finetune) + +- လက်တွေ့သင်ကြားခန်းမ + - [နောက်ဆုံးပေါ်ပုံစံများကို လေ့လာခြင်း: LLMs, SLMs, ဒေသတွင်းဖွံ့ဖြိုးတိုးတက်မှုနှင့် အခြားများ](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP အလားအလာဖြင့်ဖွင့်မောင်းခြင်း: Microsoft Olive ဖြင့် ပြုပြင်ပြောင်းလဲခြင်း](https://github.com/azure/Ignite_FineTuning_workshop) + +- ပညာရေးလေ့လာမှုစာတမ်းများနှင့် မိတ်ဆက်စာအုပ်များ + - [စာအုပ်များသာလိုအပ်သည် II: phi-1.5 နည်းပညာအစီရင်ခံစာ](https://arxiv.org/abs/2309.05463) + - [Phi-3 နည်းပညာအစီရင်ခံစာ: မိုဘိုင်းဖုန်းပေါ်တွင် ထိရောက်သော ဘာသာစကားမော်ဒယ်](https://arxiv.org/abs/2404.14219) - [Phi-4 နည်းပညာအစီရင်ခံစာ](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini နည်းပညာအစီရင်ခံစာ: LoRA များပေါင်းစပ်မှုမှတဆင့် ချိုချဥ်သော်လည်း ထူးချွန်သော Multimodal စကားဝိုင်းမော်ဒယ်များ](https://arxiv.org/abs/2503.01743) - - [ကားအတွင်း လုပ်ဆောင်ချက်ခေါ်ဆိုမှုအတွက် အငယ်စားစကားဝိုင်းမော်ဒယ်များ ကိုအကောင်းဆုံးလုပ်ဆောင်ခြင်း](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) ခုနှစ်ရွေးချယ်မေးခွန်းဖြေဆိုရန် PHI-3 Fine-Tuning: နည်းဗျူဟာ၊ ရလဒ်များနှင့် စိန်ခေါ်မှုများ](https://arxiv.org/abs/2501.01588) + - [Phi-4-Mini နည်းပညာအစီရင်ခံစာ: Mixture-of-LoRAs မှတဆင့် ကျစ်လစ်ပြီး အင်အားကြီးသော မိုက်တယ်မော်ဒယ်များ](https://arxiv.org/abs/2503.01743) + - [ကားအတွင်း လုပ်ဆောင်ချက်ခေါ်ဆိုမှုအတွက် အသေးစား ဘာသာစကားမော်ဒယ်များ အကောင်းဆုံးပြုလုပ်ခြင်း](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) PHI-3 ကို မေးခွန်းအမျိုးအစား အများအပြားအတွက် ပြုပြင်ပြောင်းလဲခြင်း: နည်းလမ်း, ရလဒ်များနှင့် စိန်ခေါ်မှုများ](https://arxiv.org/abs/2501.01588) - [Phi-4-reasoning နည်းပညာအစီရင်ခံစာ](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Phi-4-mini-reasoning နည်းပညာအစီရင်ခံစာ](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) @@ -226,98 +229,101 @@ Phi ကို Cloud သို့မဟုတ် ဒြပ်စင်ကိရ ### Microsoft Foundry တွင် Phi -Microsoft Phi ကို မည်သည့် hardware ကိရိယာပေါ်တွင်မဆို အသုံးပြုပုံနှင့် E2E ဖြေရှင်းချက်များ ဆောက်လုပ်နည်းကို သင်ယူနိုင်သည်။ Phi ကို ကိုယ်တိုင်အတွေ့အကြုံရရှိရန်အတွက် မော်ဒယ်များနှင့် ကစားပြီး သင့်အခြေအနေများအတွက် Phi ကို ချိန်ညှိခြင်းကို [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) မှတဆင့် စတင်ကြည့်ရှုနိုင်ပြီး Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) တွင် ထပ်မံလေ့လာနိုင်သည်။ +Microsoft Phi ကို အသုံးပြုနည်းနှင့် မတူညီသော ဟာ့ဒ်ဝဲ စက်ပစ္စည်းများတွင် E2E ဖြေရှင်းချက်များ ဖန်တီးနည်းကို သင်ယူနိုင်သည်။ သင်၏ ရှုထောင့်များအတွက် Phi ကို စိတ်ကြိုက်ပြင်ဆင်ပြီး မော်ဒယ်များနှင့် ကြိုးစားမောင်းနှင်ပုံကို အတွေ့အကြုံရရှိရန် [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) ကို ဘယ်လိုစတင်အသုံးပြုမလဲ၊ [Microsoft Foundry](../../../../../../../md/02.QuickStart/AzureAIFoundry_QuickStart.md ) တွင် အသေးစိတ်သိရှိနိုင်သည်။ -**ကစားရန်နေရာ** -မော်ဒယ်တိုင်းတွင် စမ်းသပ်နိုင်သောစမ်းသပ်ခန်းတစ်ခုရှိပြီး [Azure AI Playground](https://aka.ms/try-phi3) တွင် လုပ်ဆောင်နိုင်ပါသည်။ +**ကစားကွင်း** +မော်ဒယ်တိုင်းမှာ မော်ဒယ်ကို စမ်းသပ်ရန် သီးသန့်ကစားကွင်းရှိသည် [Azure AI Playground](https://aka.ms/try-phi3)။ ### GitHub မော်ဒယ်များတွင် Phi -Microsoft Phi ကို မည်သည့် hardware ကိရိယာပေါ်တွင်မဆို အသုံးပြုပုံနှင့် E2E ဖြေရှင်းချက်များ ဆောက်လုပ်နည်းကို သင်ယူနိုင်သည်။ Phi ကို ကိုယ်တိုင်အတွေ့အကြုံရရှိရန်အတွက် မော်ဒယ်နှင့်ကစားပြီး သင့်အခြေအနေများအတွက် Phi ကို ချိန်ညှိနိုင်ရန် [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ကိုအသုံးပြုနိုင်ပြီး Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) တွင် ထပ်မံလေ့လာနိုင်သည်။ +Microsoft Phi ကို အသုံးပြုနည်းနှင့် မတူညီသော ဟာ့ဒ်ဝဲ စက်ပစ္စည်းများတွင် E2E ဖြေရှင်းချက်များ ဖန်တီးနည်းကို သင်ယူနိုင်သည်။ သင်၏ ရှုထောင့်များအတွက် Phi ကို စိတ်ကြိုက်ပြင်ဆင်ပြီး မော်ဒယ်နှင့် အတွေ့အကြုံရရှိရန် [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ကို အသုံးပြုပါ၊ [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) တွင် စတင်သင်ယူနိုင်သည်။ -**ကစားရန်နေရာ** -မော်ဒယ်တိုင်းအား [စမ်းသပ်ရန် ကစားရန်နေရာ](/md/02.QuickStart/GitHubModel_QuickStart.md) တစ်ခု ဖော်ပြထားသည်။ +**ကစားကွင်း** +မော်ဒယ်တိုင်းမှာ မော်ဒယ်ကို စမ်းသပ်ရန် သီးသန့် [ကစားကွင်းရှိသည်](/md/02.QuickStart/GitHubModel_QuickStart.md)။ ### Hugging Face တွင် Phi -မော်ဒယ်အား [Hugging Face](https://huggingface.co/microsoft) တွင်လည်း ရှာဖွေတွေ့ရှိနိုင်သည်။ +မော်ဒယ်ကို [Hugging Face](https://huggingface.co/microsoft) တွင်လည်း တွေ့ရှိနိုင်သည်။ -**ကစားရန်နေရာ** - [Hugging Chat ကစားရန်နေရာ](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**ကစားကွင်း** + [Hugging Chat ကစားကွင်း](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) - ## 🎒 အခြားသင်တန်းများ + ## 🎒 အခြား သင်တန်းများ -ကျွန်ုပ်တို့၏အဖွဲ့က အခြားသင်တန်းများကိုထုတ်လုပ်သည်! ကြည့်ရှုပါ: +ကျွန်ုပ်တို့ အဖွဲ့သည် အခြား သင်တန်းများကို ထုတ်လုပ်ပါသည်! ကြည့်ရှုပါ: ### LangChain -[![LangChain4j အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain for Beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / အေးဂျင့်များ -[![AZD အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agents +[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Generative AI အစဉ် -[![Generative AI အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +### Generative AI စီးရီး +[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) [![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) [![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- -### အခြေခံသင်ထုတ်မှုများ -[![ML အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web Dev အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR Development အသစ်စတင်သူများအတွက်](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +### အခြေခံသင်ယူမှု +[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Copilot အစဉ် -[![Copilot AI သက်ဆိုင်ရာ အတူတူအစီအစဉ်များအတွက်](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot C#/.NET အတွက်](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot စွန့်စားခန်း](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +### Copilot စီးရီး +[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## တာဝန်ရှိသော AI -Microsoft သည် သုံးစွဲသူများအား ကျွန်ုပ်တို့၏ AI ထုတ်ကုန်များကို တာဝန်ရှိကြီးကြပ်မှုဖြင့်အသုံးပြုနိုင်ရန် ကူညီရန်၊ ကျွန်ုပ်တို့၏ သင်ယူမှုများကို မျှဝေရန်နှင့် Transparency Notes နှင့် Impact Assessments စသည့် ကိရိယာများမှတဆင့် ယုံကြည်မှုအခြေပြု ပူးပေါင်းဆောင်ရွက်မှုများ တည်ဆောက်ပေးရန် ကတိပြုထားသည်။ ဤကိရိယာအများအပြားကို [https://aka.ms/RAI](https://aka.ms/RAI) တွင် တွေ့နိုင်သည်။ -Microsoft ၏ တာဝန်ရှိသော AI သဘောထားသည် တရားမျှတမှု၊ ယုံကြည်စိတ်ချရမှုနှင့် ဘေးကင်းမှု၊ ပုဂ္ဂိုလ်ရေးနှင့် လုံခြုံမှု၊ အပါအဝင် ဖြစ်စဉ်၊ ဖွင့်ဝင်းမှုနှင့် တာဝန်ယူမှုတို့အပေါ်အခြေခံသည်။ +Microsoft သည် ကျွန်ုပ်တို့၏ AI ထုတ်ကုန်များကို တာဝန်ရှိစွာ အသုံးပြုနိုင်ရန် ကူညီပေးခြင်း၊ သင်ယူချက်များ မျှဝေပေးခြင်းနှင့် Transparency Notes နှင့် Impact Assessments ကဲ့သို့သော ကိရိယာများမှတဆင့် ယုံကြည်မှုအခြေခံ မိတ်ဖက်ဆက်ဆံရေးများ တည်ဆောက်ခြင်းတွင် အလေးထားဆောင်ရွက်လျက်ရှိသည်။ ဤအရင်းအမြစ်များပိုများသည် [https://aka.ms/RAI](https://aka.ms/RAI) တွင် တွေ့နိုင်သည်။ +Microsoft ၏ တာဝန်ရှိသော AI မှုဟာ တန်ဖိုးတရားမျှတခြင်း၊ ယုံကြည်စိတ်ချရမှုနှင့် ဘေးကင်းလုံခြုံမှု၊ ကိုယ်ရေးရာဇဝင်နှင့် လုံခြုံရေး၊ ပါဝင်ဆောင်ရွက်မှု၊ ဖန်တီးပြန်ဆိုခြင်းနှင့် တာဝန်ယူမှုတို့ အခြေခံသော AI လုပ်ထုံးလုပ်နည်းများပေါ်တွင် တည်ရှိသည်။ + +ဤနမူနာတွင် သုံးထားသော ဆီးမားအရွယ် ဘာသာစကား၊ ပုံနှင့် အသံပုံစံများသည် မမျှတသော၊ ယုံကြည်စိတ်ချရမှုမရှိသော သို့မဟုတ် တုံ့ပြန်မှု မရွေးမချော် ဖြစ်နိုင်ခြေရှိပြီး ထိခိုက်မှုများ ဖြစ်ပေါ်စေနိုင်သည်။ အန္တရာယ်များနှင့် ကန့်သတ်ချက်များကို သိရှိရန် [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ကို အကြံပြုပါသည်။ + -ဤဥပမာတွင် အသုံးပြုထားသည့် ကြီးမားသော အကြောင်းအရာသဘာဝဘာသာစကား၊ ဓာတ်ပုံ၊ မိန့်ခွန်း မော်ဒယ်များသည် မတရားသော၊ ယုံကြည်စိတ်ချရမှုမရှိသော်လည်း ဒဏ်ရာဖြစ်စေနိုင်သော အပြုအမူများ ပြုလုပ်နိုင်သည်။ အန္တရာယ်များနှင့် ကန့်သတ်ချက်များကို သိရှိရန် [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ကို ကြည့်ရှုပါ။ -ဤရိုက်ခတ်မှုများကိုလျော့ပါးစေရေးအတွက် အကြံပြုထားသောနည်းလမ်းမှာ အန္တရာယ်ရှိသောအပြုအမူများကို ရှာဖွေတားမြစ်နိုင်သော ဘေးကင်းရေးစနစ်တစ်ခုကို သင်၏ ဖွဲ့စည်းမှုအတွင်း ထည့်သွင်းခြင်းဖြစ်သည်။ [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) သည် လွတ်လပ်၍ဘာသာရပ်ပိုင် မဟုတ်သော ကာကွယ်မှုအလွှာတစ်ခုဖြစ်ပြီး၊ အပလီကေးရှင်းများနှင့် ဝန်ဆောင်မှုများအတွက် အသုံးပြုသူထံမှထုတ်လုပ်သည့် မကောင်းသော အကြောင်းအရာများနှင့် AI ထုတ်လုပ်သော မကောင်းသော အကြောင်းအရာများကို ရှာဖွေစစ်ဆေးနိုင်သည်။ Azure AI Content Safety တွင် မကောင်းသော အကြောင်းအရာများကို ရှာဖွေရန် အကူအညီပေးသည့် စာသားနှင့် ပုံ API များ ပါဝင်သည်။ Microsoft Foundry အတွင်းရှိ Content Safety ဝန်ဆောင်မှုက မကောင်းသောအကြောင်းအရာများကို မတူညီသော ပုံစံများအားဖြင့် ရှာဖွေရန်၊ စူးစမ်းကြည့်ရှုရန်နှင့် နမူနာကုဒ်များကို စမ်းသပ်ကြည့်နိုင်ရန် ခွင့်ပြုသည်။ အောက်ပါ [အမြန်စတင်လမ်းညွှန်စာမျက်နှာ](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) သည် ဝန်ဆောင်မှုသို့ တောင်းဆိုချက်များပေးပို့ရာတွင် ဦးဆောင်လမ်းညွှန်ပေးပါသည်။ +ဤအန္တရာယ်များကို လျော့နည်းစေသည့် အကြံပြုနည်းမှာ သင့်အဆောက်အအုံတွင် ထိခိုက်နိုင်သော အပြုအမူများကို စိစစ်၍ ကာကွယ်နိုင်သော လုံခြုံရေးစနစ်တစ်ခု ထည့်သွင်းခြင်းဖြစ်သည်။ [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) သည် လွတ်လပ်ပြီး သီးခြားသော ကာကွယ်မှုအလွှာတစ်ခုကို ပံ့ပိုးပေးပြီး၊ လျှပ်စစ်နှင့် AI ထုတ်လုပ်သော ထိခိုက်မှုရှိနိုင်သော အကြောင်းအရာများကို အပလီကေးရှင်းများနှင့် ဝန်ဆောင်မှုများတွင် စိစစ်နိုင်စွမ်း ရှိသည်။ Azure AI Content Safety တွင် ထိခိုက်မှုရှိသော အကြောင်းအရာများကို စိစစ်ပေးနိုင်သော စာသားနှင့် ပုံ API များပါဝင်သည်။ Microsoft Foundry တွင် Content Safety ဝန်ဆောင်မှုသည် သင့်ကို ထိခိုက်မှုရှိသော အကြောင်းအရာများကို လေ့လာ၊ ရှာဖွေ၊ နမူနာ ကုဒ်များနှင့် စမ်းသပ်ရန် ခွင့်ပြုသည်။ အောက်ပါ [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) သည် ဝန်ဆောင်မှုသို့ အမိန့်ပေးမှုများပြုလုပ်ရန် ဦးတည် ပြသသည်။ -စဉ်းစားရန်ရှိသော တခြားအချက်မှာ အပလီကေးရှင်းတစ်ခုလုံး၏ စွမ်းဆောင်ရည်ဖြစ်သည်။ မျိုးစုံနည်းနှင့် မျိုးစုံမော်ဒယ်ပါဝင်သည့် အပလီကေးရှင်းများအတွက် ကျွန်ုပ်တို့အနေဖြင့် စွမ်းဆောင်ရည်မှာ သင်နှင့် သင်၏ အသုံးပြုသူများမျှော်လင့်သောအတိုင်း စနစ်သည် ဆောင်ရွက်ပေးနိုင်ခြင်း၊ မကောင်းသော ထုတ်လွှင့်မှုများမဖြစ်ပေါ်စေရန်ဖြစ်သည်ဟုယူဆပါသည်။ သင့်အပလီကေးရှင်းတစ်ခုလုံး၏ စွမ်းဆောင်ရည်ကို [စွမ်းဆောင်ရည်နှင့် အရည်အသွေး၊ ဖြစ်နိုင်သော အန္တရာယ်နှင့် ဘေးကင်းရေးအကဲဖြတ်ကိရိယာများ](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)ဖြင့် တိုင်းတာသုံးသပ်ခြင်းမှာ အရေးကြီးသည်။ သင်၌ [သင်၏ကိုယ်ပိုင်အကဲဖြတ်ကိရိယာများ](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) ဖြင့် ဖန်တီးပြီး အကဲဖြတ်နိုင်သည်။ +ကိုင်တွယ်စရာ အခြားအချက်မှာ အပလီကေးရှင်း၏ စုစုပေါင်း လုပ်ဆောင်ချက် ဖြစ်သည်။ မျိုးစုံ မော်ဒယ်များနှင့် မျိုးစုံ မော်ဒယ်များပါသော အပလီကေးရှင်းများတွင် လူနှင့် သင့်အသုံးပြုသူများ မျှော်မှန်းသလို စနစ်အလုပ်လုပ်ခြင်းအား ကို လုပ်ဆောင်ချက်ဖြစ်သည်ဟု ရှုမှတ်ပါသည်၊ ထိခိုက်စေသော ထွက်ရှိမှု မဖြစ်ပေါ်စေရန် ပါဝင်သည်။ စုစုပေါင်း အပလီကေးရှင်း၏ လုပ်ဆောင်ချက်ကို [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) အသုံးပြု၍ သုံးသပ်ရန် အရေးကြီးသည်။ [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) နှင့်တစ်ပြိုင်နက် ဖန်တီးအကဲဖြတ်နိုင်ရန် လည်း ရရှိပါသည်။ -သင့် AI အပလီကေးရှင်းကို ဖွံ့ဖြိုးမှုပတ်ဝန်းကျင်၌ [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) အသုံးပြု၍ အကဲဖြတ်နိုင်သည်။ စမ်းသပ်ဒေတာစုတစ်ခု သို့မဟုတ် ရည်မှန်းချက်တစ်ခုကို အသုံးပြုပြီး သင့်ရဲ့ ထုတ်လုပ်မှုမျိုးစုံ generative AI မှ ထုတ်လွှင့်မှုများကို ဖွဲ့စည်းထားသော အကဲဖြတ်ကိရိယာများ သို့မဟုတ် သင်ရွေးချယ်နိုင်သော ကိုယ်ပိုင်အကဲဖြတ်ကိရိယာများဖြင့် အရေအတွက်ဖြင့်တိုင်းတာပေးသည်။ သင့်စနစ်ကို အကဲဖြတ်ရန် azure ai evaluation sdk ဖြင့် စတင်လိုပါက [အမြန်စတင်လမ်းညွှန်စာမျက်နှာ](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) ကို လိုက်နာနိုင်သည်။ အကဲဖြတ်အလုပ်ဆောင်ပြီးနောက်၊ [Microsoft Foundry တွင် ရလဒ်များကို မြင်ကွင်းဖော်ပြနိုင်သည်](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)။ +သင်၏ AI အပလီကေးရှင်းကို ဖွံ့ဖြိုးတိုးတက်ရေးပတ်ဝန်းကျင်တွင် [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) အသုံးပြု၍ အကဲဖြတ်နိုင်သည်။ စမ်းသပ်ဒေတာစနစ် သို့မဟုတ် အာရုံစိုက်ထားသော ရည်မှန်းချက်တစ်ခုဖြင့်၊ သင်၏ စွမ်းဆောင်နိုင်သော AI application ၏ မူလတန်ဖိုးများကို ဆန်းစစ်အကဲဖြတ်မှုတွင် ပါဝင်သော စနစ်တွေ သို့မဟုတ် သင့်ရွေးချယ်သော custom evaluators အသုံးပြု၍ မျှတပြတ်သားစွာ တိုင်းတာသည်။ သင့်စနစ်ကို အကဲဖြတ်ရန် azure ai evaluation sdk ကို စတင်အသုံးပြုရန် အတွက် [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) ကို လိုက်နာနိုင်သည်။ အကဲဖြတ် ချေးပေးမှုပတ်ပတ်လည် လုပ်ဆောင်ပြီးနောက် [Microsoft Foundry တွင် ရလဒ်များကို မြင်ကွင်းဖော်ပြ](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) နိုင်သည်။ -## Trademarks +## လက်မှတ်များ -ဤပရောဂျက်တွင် ပရောဂျက်များ၊ ထုတ်ကုန်များ သို့မဟုတ် ဝန်ဆောင်မှုများအတွက် trademark များ သို့မဟုတ် အမှတ်တံဆိပ်များ ပါဝင်နိုင်သည်။ Microsoft အမှတ်တံဆိပ်များ သို့မဟုတ် အမှတ်တံဆိပ်များကို ကျင့်သုံးခြင်းသည် [Microsoft ၏ Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) များနှင့် ကိုက်ညီရမည်ဖြစ်သည်။ ဤပရောဂျက်၏ ပြင်ဆင်ပြီး ဗားရှင်းများတွင် Microsoft ၏ trademark များ သို့မဟုတ် အမှတ်တံဆိပ်များကို အသုံးပြုမှုမှာ မရောထွေးမှု ဖြစ်စေမှာမဟုတ်ရ၊ Microsoft ၏ အာမခံမှုရှိကြောင်း အဓိပ္ပါယ်မပေးရပါ။ တတိယပါတီ trademark များ သို့မဟုတ် အမှတ်တံဆိပ်များကို အသုံးပြုခြင်းသည် ယင်း တတိယပါတီ၏ မူဝါဒများအတိုင်း ဖြစ်ရမည်ဖြစ်သည်။ +ဤပရောဂျက်တွင် ပရောဂျက်များ၊ ထုတ်ကုန်များ သို့မဟုတ် ဝန်ဆောင်မှုများအတွက် လက်မှတ်များ သို့မဟုတ် အမှတ်တံဆိပ်များ ပါဝင်နိုင်သည်။ Microsoft လက်မှတ်များ သို့မဟုတ် အမှတ်တံဆိပ်များကို တရားဝင် အသုံးပြုခြင်းမှာ [Microsoft ၏ လက်မှတ်နှင့် အမှတ်တံဆိပ် လမ်းညွှန်ချက်များ](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) ကို လိုက်နာရမည်။ +Microsoft လက်မှတ် သို့မဟုတ် အမှတ်တံဆိပ်များကို ဤပရောဂျက်၏ မူလပုံစံမှ ပြင်ဆင်ထားသည့် ဗားရှင်းများတွင် အသုံးပြုရာ၌ Microsoft ၏ ထောက်ခံမှု ဖြစ်နေသည်ဟု ယူဆပုံမဖန်တီးရ။ တတိယပါတီ လက်မှတ် သို့မဟုတ် အမှတ်တံဆိပ်များကို အသုံးပြုရာတွင် အဆိုပါ တတိယပါတီ၏ မူဝါဒများကို လိုက်နာရမည်။ -## Getting Help +## ကူညီမှု ရယူခြင်း -AI အပလီကေးရှင်းများ ဖန်တီးရာတွင် အခက်အခဲရှိပါက သို့မဟုတ် မေးခွန်းများရှိပါက အောက်ပါနေရာသို့ ဝင်ရောက်ပါ- +AI အက်ပ်များ တည်ဆောက်ရာတွင် ပြဿနာတက်ပါက သို့မဟုတ် မေးခွန်းရှိပါက ပူးပေါင်းပါ။ [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -ထုတ်ကုန်တုံ့ပြန်ချက်များ သို့မဟုတ် ဖွဲ့စည်းနေစဉ် အမှားတွေရှိပါက အောက်ပါနေရာသို့ သွားပါ- +ထုတ်ကုန်တုံ့ပြန်မှု သို့မဟုတ် အဆင်မပြေမှုများရှိပါက သွားရောက်ဆက်သွယ်ပါ။ [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**ကန့်သတ်ချက်** -ဤစာတမ်းကို AI ဘာသာပြန်ဝန်ဆောင်မှုဖြစ်သော [Co-op Translator](https://github.com/Azure/co-op-translator) အသုံးပြု၍ ဘာသာပြန်ထားပါသည်။ တိကျမှုအတွက် ကြိုးပမ်းဆောင်ရွက်သည်မှ ဖြစ်ပေမယ့် အလိုအလျောက် ဘာသာပြန်ချက်များတွင် အမှားများ သို့မဟုတ် မှားယွင်းမှုများ ရှိနိုင်ကြောင်း ကျေးဇူးပြု၍ သတိထားပါ။ မူလစာတမ်းသည် မိမိဘာသာစကားဖြင့် ရှိသည့် ထိုစာရင်းကို အတည်ပြုရင်းမြစ်အဖြစ် သတ်မှတ်သင့်ပါသည်။ အရေးကြီးသော သတင်းအချက်အလက်များအတွက် အတတ်ပညာရှင် လူသားဘာသာပြန်မှုကို အကြံပြုပါသည်။ ဤဘာသာပြန်မှုကို အသုံးပြုမှုကြောင့် ဖြစ်ပေါ်လာသော မဖတ်ခွင့်ရမှု သို့မဟုတ် အမှားနားလည်မှုများအတွက် ကျွန်ုပ်တို့ တာဝန်မယူပါ။ +**ပြောကြားချက်** +ဤစာတမ်းကို AI ဘာသာပြန်ဝန်ဆောင်မှု [Co-op Translator](https://github.com/Azure/co-op-translator) အသုံးပြု၍ ဘာသာပြန်ထားပါသည်။ ကျွန်ုပ်တို့သည် တိကျမှန်ကန်မှုအတွက် ကြိုးပမ်းနေသော်လည်း၊ စက်ကိရိယာဘာသာပြန်ခြင်းများတွင် အမှားများ သို့မဟုတ် မှားယွင်းချက်များ ပါဝင်နိုင်ကြောင်း သတိပြုပါရန် လိုအပ်ပါသည်။ မူလစာတမ်းကို မူရင်းဘာသာဖြင့်သာ ယုံကြည်စိတ်ချရသော အချက်အလက်အဖြစ် သတ်မှတ်သင့်သည်။ အရေးကြီးသည့် သတင်းအချက်အလက်များအတွက် ပရော်ဖက်ရှင်နယ် လူသားဘာသာပြန်သူဝန်ဆောင်မှုကို အကြံပြုပါသည်။ ဤဘာသာပြန်ချက်ကို အသုံးပြုခြင်းမှ ဖြစ်ပေါ်လာသော နားလည်မှုကွာခြားမှုများ သို့မဟုတ် မမှန်ကန်သော အသုံးပြုမှုများအတွက် ကျွန်ုပ်တို့ တာဝန်မခံပါ။ \ No newline at end of file diff --git a/translations/my/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/my/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..ed4fed136 --- /dev/null +++ b/translations/my/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Microsoft phi-4-mini နှင့် SQLite FTS5 ဖြင့် ဆန်းစစ်ပေါင်းစပ် RAG (Zero-Cloud SLM)\n", + "\n", + "> **ရေးသားသူ:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **အာရုံစိုက်မှု:** အစိတ်အပိုင်းသေးသော ဘာသာစကား မော်ဒယ်များ (SLMs), SQLite FTS5 BM25, မကျဉ်းမှုဖွဲ့စည်းမှုများ၊ Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. တိုက်ပြောမှု: ဒေသခံ SLM များ၏ စကားလုံး အသိပညာပြဿနာ\n", + "တည်ငြိမ်သည့် RAG မျိုးစိတ်များသည် သန့်ရှင်းသော ဗက်တာတစ်ခုလုံးမှတ်တမ်းအပေါ် အခြေခံမှုရှိပြီး၊ မှန်ကန်သော အရေအချင်း စာလုံးများ (ဥပမာ - `2,340,000 TL`, စာချုပ် ကုဒ်များ၊ အကောင့်နံပါတ်များ) ကို ရယူနိုင်ခြင်း မရှိကြောင်း တွေ့ရသည်။ \n", + "ဒုတိယအားဖြင့်၊ ချွတ်တောက်သော စာကြောင်းရှာဖွေရေး (BM25) သည် အဓိပ္ပါယ်တူ စကားလုံးများနှင့် စကားပြန်မေးခွန်းများကို လွဲမှားလွန်းသည်။\n", + "\n", + "ဤ စားပွဲစာအုပ်တွင် မြန်နှုန်းမြင့်၊ မှတ်ဥာဏ်အတွင်း ဆန်းစစ်ပေါင်းစပ် စနစ်တစ်ခု ပြုလုပ်ခြင်းကို ပြသသည်၊ စုပေါင်းမှုမှာ-\n", + "1. **ကျွမ်းကျင် ဗက်တာများ** (Cosine Similarity)\n", + "2. **ချွတ်တောက်သော စာကြောင်းရှာဖွေရေး** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Microsoft `phi-4-mini` ဖြင့် ရေးသားသော အခြေခံမှုနှင့် ကြည့်ရှုခြင်း (`[1]`, `[2]`)**\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ ၂။ နှစ်ဖက် SQLite စနစ် (အထူဗက်တာများ + အလွှာ FTS5 BM25 ကွက်တန်း)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. စမ်းသပ်မှု ထည့်သွင်းခြင်းနှင့် ဆောင်ရွက်မှု စမ်းသပ်ခြင်း\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Microsoft phi-4-mini အတွက် အခြေခံထားသော Prompt ဖော်ပြခြင်း\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**ပြောကြားချက်**\nဤစာတမ်းကို AI ဘာသာပြန်ဝန်ဆောင်မှု [Co-op Translator](https://github.com/Azure/co-op-translator) အသုံးပြု၍ ဘာသာပြန်ထားပါသည်။ ကျွန်ုပ်တို့သည် တိကျမှန်ကန်မှုအတွက် ကြိုးပမ်းနေသော်လည်း၊ စက်ကိရိယာဘာသာပြန်ခြင်းများတွင် အမှားများ သို့မဟုတ် မှားယွင်းချက်များ ပါဝင်နိုင်ကြောင်း သတိပြုပါရန် လိုအပ်ပါသည်။ မူလစာတမ်းကို မူရင်းဘာသာဖြင့်သာ ယုံကြည်စိတ်ချရသော အချက်အလက်အဖြစ် သတ်မှတ်သင့်သည်။ အရေးကြီးသည့် သတင်းအချက်အလက်များအတွက် ပရော်ဖက်ရှင်နယ် လူသားဘာသာပြန်သူဝန်ဆောင်မှုကို အကြံပြုပါသည်။ ဤဘာသာပြန်ချက်ကို အသုံးပြုခြင်းမှ ဖြစ်ပေါ်လာသော နားလည်မှုကွာခြားမှုများ သို့မဟုတ် မမှန်ကန်သော အသုံးပြုမှုများအတွက် ကျွန်ုပ်တို့ တာဝန်မခံပါ။\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/ne/.co-op-translator.json b/translations/ne/.co-op-translator.json index 468eac2bc..aab354f80 100644 --- a/translations/ne/.co-op-translator.json +++ b/translations/ne/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "ne" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:00:38+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:42:59+00:00", "source_file": "README.md", "language_code": "ne" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "ne" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:43:53+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "ne" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:27+00:00", diff --git a/translations/ne/README.md b/translations/ne/README.md index a1d0acb13..e1620d7ae 100644 --- a/translations/ne/README.md +++ b/translations/ne/README.md @@ -1,42 +1,42 @@ -# Phi कुकबुक: Microsoft का Phi मोडेलहरूसँग हातेमालो गर्ने उदाहरणहरू +# फाइ कुकबुक: माइक्रोसफ्टको फाइ मोडेलहरूसँग हातले काम गर्ने उदाहरणहरू [![GitHub Codespaces मा नमूनाहरू खोल्नुहोस् र प्रयोग गर्नुहोस्](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Dev Containers मा खोल्नुहोस्](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub योगदानकर्ताहरू](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub समस्याहरू](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub पुल अनुरोधहरू](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PR हरू स्वागतयोग्य](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub पुल-रिक्वेस्टहरू](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PR स्वागतयोग्य](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub हेराइकर्ताहरू](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub अवलोकन गर्नेहरू](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub फोर्कहरू](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ताराहरू](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub स्टारहरू](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi Microsoft द्वारा विकास गरिएको एक शृङ्खला खुला स्रोत AI मोडेलहरू हो। +फाइ माइक्रोसफ्टद्वारा विकास गरिएको खुला स्रोत एआई मोडेलहरूको एक शृंखला हो। -Phi हालसम्म सबैभन्दा शक्तिशाली र लागत-प्रभावी सानो भाषा मोडेल (SLM) हो, जसले बहुभाषिक, तर्क, पाठ/च्याट उत्पादन, कोडिङ, छवि, अडियो र अन्य परिदृश्यहरूमा अत्यन्त राम्रो बेंचमार्कहरू प्रदर्शन गर्छ। +फाइ हाल सानो भाषा मोडेल (SLM)मा सबैभन्दा शक्तिशाली र लागत-प्रभावकारी हो, धेरै-भाषा, तर्क, पाठ/च्याट सिर्जना, कोडिङ, छवि, अडियो र अन्य परिदृश्यहरूमा राम्रो मापन परिणामहरू सहित। -तपाईं Phi लाई क्लाउडमा वा एज उपकरणहरूमा तैनाथ गर्न सक्नुहुन्छ र सीमित कम्प्युटिङ शक्ति प्रयोग गरेर सजिलैसँग उत्पादनात्मक AI अनुप्रयोगहरू निर्माण गर्न सक्नुहुन्छ। +तपाइँ फाइलाई क्लाउड वा एज उपकरणहरूमा तैनाथ गर्न सक्नुहुन्छ, र सीमित कम्प्युटिङ पावरसहित सजिलैसँग जनरेटिभ एआई अनुप्रयोगहरू निर्माण गर्न सक्नुहुन्छ। -यी स्रोतहरू प्रयोग गर्न सुरू गर्न निम्न चरणहरू पालना गर्नुहोस्: -1. **रिपोजिटरी फोर्क गर्नुहोस्**: क्लिक गर्नुहोस् [![GitHub फोर्कहरू](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +यी स्रोतहरू प्रयोग गर्न सुरू गर्न तलका चरणहरू पालना गर्नुहोस्: +1. **रिपोजिटरी फोर्क गर्नुहोस्**: क्लिक गर्नुहोस् [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **रिपोजिटरी क्लोन गर्नुहोस्**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Microsoft AI Discord समुदायमा सहभागी हुनुहोस् र विशेषज्ञ र साथी विकासकर्ताहरूसँग भेट्नुहोस्**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**माइक्रोसफ्ट एआई डिस्कोर्ड समुदायमा सामेल हुनुहोस् र विशेषज्ञहरू र अन्य विकासकर्ताहरूलाई भेट्नुहोस्**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/ne/cover.eb18d1b9605d754b.webp) -### 🌐 बहुभाषिक समर्थन +### 🌐 बहु-भाषा समर्थन -#### GitHub Action मार्फत समर्थन गरिएको (स्वचालित र सधैं अद्यावधिक) +#### GitHub Action द्वारा समर्थित (स्वचालित र सधैं अद्यावधिक) -[अरबी](../ar/README.md) | [बङ्गाली](../bn/README.md) | [बल्गेरियन](../bg/README.md) | [बर्मीस (म्यानमार)](../my/README.md) | [चिनी (सरलीकृत)](../zh-CN/README.md) | [चिनी (परम्परागत, हङकङ)](../zh-HK/README.md) | [चिनी (परम्परागत, मकाऊ)](../zh-MO/README.md) | [चिनी (परम्परागत, ताइवान)](../zh-TW/README.md) | [क्रोएशियन](../hr/README.md) | [चेक](../cs/README.md) | [डेनिश](../da/README.md) | [डच](../nl/README.md) | [एस्टोनियन](../et/README.md) | [फिनिश](../fi/README.md) | [फ्रेन्च](../fr/README.md) | [जर्मन](../de/README.md) | [ग्रीक](../el/README.md) | [हिब्रू](../he/README.md) | [हिन्दी](../hi/README.md) | [हंगेरीअन](../hu/README.md) | [इन्डोनेशियाली](../id/README.md) | [इटालियन](../it/README.md) | [जापानी](../ja/README.md) | [कन्नड](../kn/README.md) | [खमेर](../km/README.md) | [कोरियन](../ko/README.md) | [लिथुआनियन](../lt/README.md) | [मलाय](../ms/README.md) | [मलयालम](../ml/README.md) | [मराठी](../mr/README.md) | [नेपाली](./README.md) | [नाइजेरियन पिड्गिन](../pcm/README.md) | [नोर्वेजियन](../no/README.md) | [फारसी (पर्सियन)](../fa/README.md) | [पोलिश](../pl/README.md) | [पोर्चुगाली (ब्राजील)](../pt-BR/README.md) | [पोर्चुगाली (पोर्चुगल)](../pt-PT/README.md) | [पंजाबी (गुरमुखी)](../pa/README.md) | [रोमानियन](../ro/README.md) | [रूसी](../ru/README.md) | [सर्बियन (सिरिलिक)](../sr/README.md) | [स्लोभाक](../sk/README.md) | [स्लोभेनियाली](../sl/README.md) | [स्पेनी](../es/README.md) | [स्वाहिली](../sw/README.md) | [स्वीडिश](../sv/README.md) | [ट्यागालग (फिलिपिनो)](../tl/README.md) | [तमिल](../ta/README.md) | [तेलुगु](../te/README.md) | [थाई](../th/README.md) | [तुर्किश](../tr/README.md) | [युक्रेनी](../uk/README.md) | [उर्दू](../ur/README.md) | [भियतनामी](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](./README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **स्थानीय रूपमा क्लोन गर्न प्राथमिकता दिनुहुन्छ?** +> **स्थानीय रूपमा क्लोन गर्न रुचाउनुहुन्छ?** > -> यो रिपोजिटरीमा ५० भन्दा बढी भाषा अनुवादहरू समावेश छन् जसले डाउनलोड आकारलाई उल्लेखनीय रूपमा बढाउँछ। अनुवादहरू बिना क्लोन गर्नको लागि, sparse checkout प्रयोग गर्नुहोस्: +> यो रिपोजिटरीमा ५०+ भाषा अनुवादहरू समावेश छन् जसले डाउनलोड साइजलाई धेरै बृद्धि गर्छ। अनुवादहरू बिना क्लोन गर्न स्पार्स चेकआउट प्रयोग गर्नुहोस्: > > **Bash / macOS / Linux:** > ```bash @@ -52,268 +52,272 @@ Phi हालसम्म सबैभन्दा शक्तिशाली > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> यसले तपाईंलाई कोर्स पूरा गर्न आवश्यक सबै कुरा धेरै छिटो डाउनलोडको साथ प्रदान गर्दछ। +> यसले तपाईंलाई कोर्स पूरा गर्न आवश्यक सबै कुरा छिटो डाउनलोडको साथ दिन्छ। ## विषय सूची - परिचय - - [Phi परिवारमा स्वागत](./md/01.Introduction/01/01.PhiFamily.md) - - [तपाईंको वातावरण सेटअप गर्ने](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [प्रमुख प्रविधिहरूको बुझाइ](./md/01.Introduction/01/01.Understandingtech.md) - - [Phi मोडेलहरूको AI सुरक्षा](./md/01.Introduction/01/01.AISafety.md) - - [Phi हार्डवेयर समर्थन](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi मोडेलहरू र प्लेटफर्महरूमा उपलब्धता](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Guidance-ai र Phi को प्रयोग](./md/01.Introduction/01/01.Guidance.md) + - [फाइ परिवारमा स्वागत छ](./md/01.Introduction/01/01.PhiFamily.md) + - [तपाईंको वातावरण सेटअप गर्दै](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [महत्त्वपूर्ण प्रविधिहरू बुझ्दै](./md/01.Introduction/01/01.Understandingtech.md) + - [फाइ मोडेलहरूको लागि एआई सुरक्षा](./md/01.Introduction/01/01.AISafety.md) + - [फाइ हार्डवेयर समर्थन](./md/01.Introduction/01/01.Hardwaresupport.md) + - [प्लेटफर्महरूमा फाइ मोडेलहरू र उपलब्धता](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Guidance-ai र फाइ उपयोग गर्दै](./md/01.Introduction/01/01.Guidance.md) - [GitHub मार्केटप्लेस मोडेलहरू](https://github.com/marketplace/models) - - [Azure AI मोडेल क्याटलक](https://ai.azure.com) + - [Azure AI मोडेल क्याटलग](https://ai.azure.com) -- विभिन्न वातावरणमा Phi को इन्फरेन्स +- फरक वातावरणमा फाइ इन्फरेन्स - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub मोडेलहरू](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry मोडेल क्याटलक](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Microsoft Foundry मोडेल क्याटलग](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [AI टुलकिट VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - -- Phi परिवारको इन्फरेन्स - - [iOS मा Phi को इन्फरेन्स](./md/01.Introduction/03/iOS_Inference.md) - - [Android मा Phi को इन्फरेन्स](./md/01.Introduction/03/Android_Inference.md) - - [Jetson मा Phi को इन्फरेन्स](./md/01.Introduction/03/Jetson_Inference.md) - - [AI PC मा Phi को इन्फरेन्स](./md/01.Introduction/03/AIPC_Inference.md) - - [Apple MLX फ्रेमवर्कसँग Phi को इन्फरेन्स](./md/01.Introduction/03/MLX_Inference.md) - - [स्थानीय सर्भरमा Phi को इन्फरेन्स](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI Toolkit प्रयोग गरी रिमोट सर्भरमा Phi को इन्फरेन्स](./md/01.Introduction/03/Remote_Interence.md) - - [Rust सँग Phi को इन्फरेन्स](./md/01.Introduction/03/Rust_Inference.md) - - [स्थानीयमा Phi—Vision को इन्फरेन्स](./md/01.Introduction/03/Vision_Inference.md) - - [Kaito AKS, Azure Containers (आधिकारिक समर्थन) सँग Phi को इन्फरेन्स](./md/01.Introduction/03/Kaito_Inference.md) -- [Phi परिवारको मात्राकरण](./md/01.Introduction/04/QuantifyingPhi.md) - - [llama.cpp प्रयोग गरी Phi-3.5 / 4 मात्राकरण](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime का लागि जेनेरेटिभ AI विस्तारहरू प्रयोग गरी Phi-3.5 / 4 मात्राकरण](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Intel OpenVINO प्रयोग गरी Phi-3.5 / 4 मात्राकरण](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Apple MLX फ्रेमवर्क प्रयोग गरी Phi-3.5 / 4 मात्राकरण](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Phi को मूल्याङ्कन - - [जिम्मेवार AI](./md/01.Introduction/05/ResponsibleAI.md) - - [मूल्याङ्कनका लागि Microsoft Foundry](./md/01.Introduction/05/AIFoundry.md) - - [Promptflow प्रयोग गरी मूल्याङ्कन](./md/01.Introduction/05/Promptflow.md) + - [Foundry लोकल](./md/01.Introduction/02/07.FoundryLocal.md) + +- फाइ परिवार इन्फरेन्स + - [iOS मा फाइ इन्फरेन्स](./md/01.Introduction/03/iOS_Inference.md) + - [Android मा फाइ इन्फरेन्स](./md/01.Introduction/03/Android_Inference.md) + - [Jetson मा फाइ इन्फरेन्स](./md/01.Introduction/03/Jetson_Inference.md) + - [AI PC मा फाइ इन्फरेन्स](./md/01.Introduction/03/AIPC_Inference.md) + - [Apple MLX फ्रेमवर्क प्रयोग गरी फाइ इन्फरेन्स](./md/01.Introduction/03/MLX_Inference.md) + - [लोकल सर्भरमा फाइ इन्फरेन्स](./md/01.Introduction/03/Local_Server_Inference.md) + - [AI टुलकिट प्रयोग गरी रिमोट सर्भरमा फाइ इन्फरेन्स](./md/01.Introduction/03/Remote_Interence.md) + - [Rust संग फाइ इन्फरेन्स](./md/01.Introduction/03/Rust_Inference.md) + - [लोकलमा फाइ--भिजन इन्फरेन्स](./md/01.Introduction/03/Vision_Inference.md) + - [Kaito AKS, Azure Containers (अधिकृत समर्थन) संग फाइ इन्फरेन्स](./md/01.Introduction/03/Kaito_Inference.md) +- [फाइ परिवार क्वान्टिफाइङ](./md/01.Introduction/04/QuantifyingPhi.md) + - [llama.cpp प्रयोग गरी Phi-3.5 / 4 क्वान्टाइजिङ](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntime को लागि जनरेटिभ AI विस्तारहरू प्रयोग गरी Phi-3.5 / 4 क्वान्टाइजिङ](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Intel OpenVINO प्रयोग गरी Phi-3.5 / 4 क्वान्टाइजिङ](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Apple MLX Framework प्रयोग गरी Phi-3.5 / 4 क्वान्टाइजिङ](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- फाइको मूल्यांकन + - [उत्तरदायी एआई](./md/01.Introduction/05/ResponsibleAI.md) + - [माइक्रोसफ्ट Foundry को मूल्यांकनका लागि](./md/01.Introduction/05/AIFoundry.md) + - [मूल्यांकनका लागि Promptflow को प्रयोग](./md/01.Introduction/05/Promptflow.md) - Azure AI Search सँग RAG - - [Phi-4-mini र Phi-4-multimodal (RAG) लाई Azure AI Search सँग कसरी प्रयोग गर्ने](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Phi-4-mini र Phi-4-multimodal (RAG) Azure AI Search सँग कसरी प्रयोग गर्ने](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud Local Hybrid RAG SQLite FTS5 र phi-4-mini सँग](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi अनुप्रयोग विकास नमूना +- फाइ अनुप्रयोग विकास नमूनाहरू - पाठ र च्याट अनुप्रयोगहरू - - Phi-4 नमूना - - [📓] [Phi-4-mini ONNX मोडेलसँग च्याट गर्नुहोस्](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 स्थानीय ONNX मोडेल .NET सँग च्याट गर्नुहोस्](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Semantic Kernel प्रयोग गरी Phi-4 ONNX सँग .NET कन्सोल एप्लिकेसनमा च्याट](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 नमूना - - [ब्राउजरमा Phi3, ONNX Runtime Web र WebGPU प्रयोग गरी स्थानीय च्याटबोट](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - Phi-4 नमूनाहरू + - [📓] [Phi-4-mini ONNX मोडेलसँग च्याट](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Phi-4 लोकल ONNX मोडेलसँग च्याट .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Semantic Kernel प्रयोग गरी Phi-4 ONNX सहित च्याट .NET कन्सोल अनुप्रयोग](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 नमुना + - [Phi3, ONNX Runtime Web र WebGPU प्रयोग गरेर ब्राउजरमा स्थानीय च्याटबोट](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino च्याट](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [बहु मोडेल - अन्तरक्रियात्मक Phi-3-mini र OpenAI व्हिस्पर](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - र्यापर निर्माण र Phi-3 लाई MLFlow सँग प्रयोग गर्ने](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [मोडेल अनुकूलन - Olive सँग ONNX Runtime वेबका लागि Phi-3-min मोडेल कसरी अनुकूल गर्ने](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [Phi-3 mini-4k-instruct-onnx सहित WinUI3 ऐप](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 बहु मोडेल एआई पावर्ड नोट्स ऐप नमूना](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [अनुकूलन र एकीकृत गर्नुहोस् अनुकूलित Phi-3 मोडेलहरू Prompt flow सँग](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [माइक्रोसफ्ट फाउन्ड्रीमा Prompt flow सँग अनुकूलित Phi-3 मोडेलहरू अनुकूलन र एकीकृत गर्नुहोस्](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [माइक्रोसफ्टको जिम्मेवार एआई सिद्धान्तहरूमा केन्द्रित माइक्रोसफ्ट फाउन्ड्रीमा अनुकूलित Phi-3 / Phi-3.5 मोडेलको मुल्याङ्कन गर्नुहोस्](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct भाषा भविष्यवाणी नमूना (चिनी/अंग्रेजी)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [धेरै मोडल - अन्तरक्रियात्मक Phi-3-mini र OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - रैपर बनाउन र MLFlow सँग Phi-3 प्रयोग गर्न](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [मोडेल अनुकूलन - Olive सँग ONNX Runtime Web को लागि Phi-3-min मोडल कसरी अनुकूल गर्ने](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 एप Phi-3 mini-4k-instruct-onnx सँग](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 धेरै मोडल AI सञ्चालित नोट्स एप नमुना](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [विशेष अनुकूलन र Prompt flow सँग आफ्नै Phi-3 मोडलहरू एकीकृत गर्नुहोस्](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Microsoft Foundry मा Prompt flow सँग Phi-3 मोडल विशेष अनुकूलन र एकीकरण](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Microsoft को जिम्मेवार AI सिद्धान्तहरूमा केन्द्रित Microsoft Foundry मा विशेष अनुकूलित Phi-3 / Phi-3.5 मोडल मूल्याङ्कन](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct भाषा पूर्वानुमान नमुना (चिनियाँ/अंग्रेजी)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG च्याटबोट](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [विन्डोज GPU प्रयोग गरेर Phi-3.5-Instruct ONNX सँग Prompt flow समाधान सिर्जना गर्ने](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [माइक्रोसफ्ट Phi-3.5 tflite प्रयोग गरेर एन्ड्रोइड ऐप सिर्जना गर्ने](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET उदाहरण स्थानीय ONNX Phi-3 मोडेल उपयोग गरेर Microsoft.ML.OnnxRuntime सँग](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Semantic Kernel र Phi-3 सँग कन्सोल च्याट .NET ऐप](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI Inference SDK कोड आधारित नमूनाहरू - - Phi-4 नमूनाहरू - - [📓] [Phi-4-multimodal प्रयोग गरी परियोजना कोड उत्पादन गर्ने](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 नमूनाहरू - - [Microsoft Phi-3 परिवारको साथ आफ्नो Visual Studio Code GitHub Copilot Chat निर्माण गर्नुहोस्](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub मोडेलहरूसँग Phi-3.5 प्रयोग गरेर Visual Studio Code च्याट कोपिलट एजेन्ट सिर्जना गर्ने](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - उन्नत निर्णय नमूनाहरू - - Phi-4 नमूनाहरू - - [📓] [Phi-4-mini-reasoning वा Phi-4-reasoning नमूनाहरू](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Microsoft Olive सँग Phi-4-mini-reasoning अनुकूलन](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Apple MLX सँग Phi-4-mini-reasoning अनुकूलन](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [GitHub मोडेलहरूसँग Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Microsoft फाउन्ड्री मोडेलहरूसँग Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - डेमोहरू - - [Hugging Face Spaces मा होस्ट गरिएको Phi-4-mini डेमोहरू](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Hugging Face Spaces मा होस्ट गरिएको Phi-4-multimodal डेमोहरू](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - भिजन नमूनाहरू - - Phi-4 नमूनाहरू - - [📓] [Phi-4-multimodal प्रयोग गरी छविहरू पढ्ने र कोड उत्पादन गर्ने](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 नमूनाहरू - - [📓][Phi-3-vision-छवि टेक्स्टबाट टेक्स्टमा](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Windows GPU प्रयोग गरेर Phi-3.5-Instruct ONNX सँग Prompt flow समाधान सिर्जना गर्ने](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Microsoft Phi-3.5 tflite प्रयोग गरेर Android एप सिर्जना गर्ने](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [स्थानीय ONNX Phi-3 मोडल प्रयोग गरेर Microsoft.ML.OnnxRuntime सँग Q&A .NET उदाहरण](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Semantic Kernel र Phi-3 सहित कन्सोल च्याट .NET एप](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI Inferencing SDK कोड आधारित नमुना + - Phi-4 नमुना + - [📓] [Phi-4-multimodal प्रयोग गरेर प्रोजेक्ट कोड निर्माण](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Phi-3 / 3.5 नमुना + - [Microsoft Phi-3 परिवारसँग आफ्नै Visual Studio Code GitHub Copilot च्याट निर्माण गर्नुहोस्](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub मोडेलहरूद्वारा Phi-3.5 सँग आफ्नै Visual Studio Code च्याट कोपिलट एजेन्ट सिर्जना गर्नुहोस्](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - उन्नत तर्क नमुना + - Phi-4 नमुना + - [📓] [Phi-4-mini-तर्क वा Phi-4-तर्क नमुना](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Microsoft Olive सँग Phi-4-mini-तर्क विशेष अनुकूलन](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLX सँग Phi-4-mini-तर्क विशेष अनुकूलन](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [GitHub मोडेलहरूसँग Phi-4-mini-तर्क](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Microsoft Foundry मोडेलहरू साथ Phi-4-mini-तर्क](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - डेमो + - [Phi-4-mini डेमोहरू Hugging Face Spaces मा होस्ट गरिएको](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal डेमोहरू Hugging Face Spaces मा होस्ट गरिएको](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - दृश्य नमुना + - Phi-4 नमुना + - [📓] [Phi-4-multimodal प्रयोग गरेर छविहरू पढ्नुहोस् र कोड उत्पादन गर्नुहोस्](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 नमुना + - [📓][Phi-3-vision-छवि पाठ देखि पाठ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP एम्बेडिङ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [डेमो: Phi-3 रिसाइकलिङ](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - भिजुअल भाषा सहायक - Phi3-Vision र OpenVINO सँग](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [DEMO: Phi-3 रीसाइकलिंग](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - दृश्य भाषा सहायक - Phi3-Vision र OpenVINO सँग](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision बहु-फ्रेम वा बहु-छवि नमूना](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Microsoft.ML.OnnxRuntime .NET प्रयोग गरी Phi-3 Vision स्थानीय ONNX मोडेल](../../md/04.HOL/dotnet/src/LabsPhi303) - - [मेनु आधारित Phi-3 Vision स्थानीय ONNX मोडेल Microsoft.ML.OnnxRuntime .NET प्रयोग गरी](../../md/04.HOL/dotnet/src/LabsPhi304) - - - निर्णय-भिजन नमूनाहरू - - Phi-4-Reasoning-Vision-15B - - [📓] [Jaywalking पत्ता लगाउन Phi-4-Reasoning-Vision-15B प्रयोग गर्दै](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [गणितका लागि Phi-4-Reasoning-Vision-15B प्रयोग गर्दै](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [UI पत्ता लगाउन Phi-4-Reasoning-Vision-15B प्रयोग गर्दै](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - गणित नमूनाहरू - - Phi-4-Mini-Flash-Reasoning-Instruct नमूनाहरू [Phi-4-Mini-Flash-Reasoning-Instruct सहित गणित डेमो](./md/02.Application/09.Math/MathDemo.ipynb) - - - अडियो नमूनाहरू - - Phi-4 नमूनाहरू - - [📓] [Phi-4-multimodal प्रयोग गरी अडियो ट्रान्सक्रिप्ट निकाल्ने](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal अडियो नमूना](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal भाषण अनुवाद नमूना](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET कन्सोल अनुप्रयोग Phi-4-multimodal अडियो प्रयोग गरेर अडियो फाइल विश्लेषण र ट्रान्सक्रिप्ट उत्पादन गर्न](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - MOE नमूनाहरू - - Phi-3 / 3.5 नमूनाहरू - - [📓] [Phi-3.5 विशेषज्ञहरूको मिश्रण (MoEs) सामाजिक मिडिया नमूना](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search, र LlamaIndex सँग RAG पाइपलाइन निर्माण](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓][Phi-3.5 Vision बहु-फ्रेम वा बहु-छवि नमुना](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Microsoft.ML.OnnxRuntime .NET प्रयोग गरेर Phi-3 Vision स्थानीय ONNX मोडल](../../md/04.HOL/dotnet/src/LabsPhi303) + - [मेनु आधारित Phi-3 Vision स्थानीय ONNX मोडल Microsoft.ML.OnnxRuntime .NET सँग](../../md/04.HOL/dotnet/src/LabsPhi304) + + - तर्क-भिजन नमुना + - Phi-4-तर्क-भिजन-15B + - [📓] [Phi-4-तर्क-भिजन-15B प्रयोग गरेर jaywalking पत्ता लगाउने](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-तर्क-भिजन-15B प्रयोग गरेर गणित](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Phi-4-तर्क-भिजन-15B प्रयोग गरेर UI पत्ता लगाउने](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - गणित नमुना + - Phi-4-मिनी-फ्ल्यास-तर्क-निर्देशन नमुना [Phi-4-मिनी-फ्ल्यास-तर्क-निर्देशन संग गणित डेमो](./md/02.Application/09.Math/MathDemo.ipynb) + + - अडियो नमुना + - Phi-4 नमुना + - [📓] [Phi-4-multimodal प्रयोग गरेर अडियो ट्रान्सक्रिप्ट निकाल्ने](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal अडियो नमुना](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal भाषण अनुवाद नमुना](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET कन्सोल अनुप्रयोग Phi-4-multimodal प्रयोग गरेर अडियो फाइल विश्लेषण र ट्रान्सक्रिप्ट उत्पादन](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE नमुना + - Phi-3 / 3.5 नमुना + - [📓] [Phi-3.5 विशेषज्ञहरूको मिश्रण मोडेलहरू (MoEs) सामाजिक मिडिया नमुना](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI खोज, र LlamaIndex सँग Retrieval-Augmented Generation (RAG) पाइपलाइन निर्माण](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - फंक्शन कलिङ नमूनाहरू - - Phi-4 नमूनाहरू 🆕 - - [📓] [Phi-4-mini सँग फंक्शन कलिङ प्रयोग गर्दै](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Phi-4-mini सँग मल्टि-एजेन्टहरूको सिर्जनाका लागि फंक्शन कलिङ प्रयोग गर्दै](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Ollama सँग फंक्शन कलिङ प्रयोग गर्दै](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [ONNX सँग फंक्शन कलिङ प्रयोग गर्दै](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - बहु-मोडल मिसिंग नमूनाहरू - - Phi-4 नमूनाहरू 🆕 - - [📓] [प्रविधि पत्रकारको रूपमा Phi-4-multimodal प्रयोग गर्दै](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET कन्सोल अनुप्रयोग Phi-4-multimodal प्रयोग गरी छविहरू विश्लेषण गर्न](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Phi अनुकूलन नमूनाहरू - - [अनुकूलन परिदृश्यहरू](./md/03.FineTuning/FineTuning_Scenarios.md) - - [अनुकूलन विरुद्ध RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Phi-3 लाई उद्योग विशेषज्ञ बन्न दिनुहोस्](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [AI Toolkit for VS Code सँग Phi-3 अनुकूलन](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Azure Machine Learning सेवा सँग Phi-3 अनुकूलन](./md/03.FineTuning/Introduce_AzureML.md) - - [Lora सँग Phi-3 अनुकूलन](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora सँग Phi-3 अनुकूलन](./md/03.FineTuning/FineTuning_Qlora.md) - - [Microsoft Foundry सँग Phi-3 अनुकूलन](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Azure ML CLI/SDK सँग Phi-3 अनुकूलन](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Microsoft Olive सँग अनुकूलन](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive Hands-On Lab सँग अनुकूलन](./md/03.FineTuning/olive-lab/readme.md) - - [Weights and Bias सँग Phi-3-vision अनुकूलन](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Apple MLX Framework सँग Phi-3 अनुकूलन](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-vision अनुकूलन (आधिकारिक समर्थन)](./md/03.FineTuning/FineTuning_Vision.md) - - [काइटो AKS सहित Phi-3 लाई फाइन-ट्यूनिङ, एजुर कन्टेनरहरू (आधिकारिक समर्थन)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Phi-3 र 3.5 भिजनलाई फाइन-ट्यूनिङ](https://github.com/2U1/Phi3-Vision-Finetune) - -- प्रायोगात्मक प्रयोगशाला - - [अत्याधुनिक मोडेलहरू अन्वेषण: LLMs, SLMs, स्थानीय विकास र थप](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP को सम्भावना उद्घाटन: Microsoft Olive सँग फाइन-ट्यूनिङ](https://github.com/azure/Ignite_FineTuning_workshop) - -- शैक्षिक अनुसन्धान पत्रपत्रिकाहरू र प्रकाशनहरू - - [पाठ्यपुस्तकहरू नै तपाईंलाई चाहिन्छ II: phi-1.5 प्राविधिक प्रतिवेदन](https://arxiv.org/abs/2309.05463) - - [Phi-3 प्राविधिक प्रतिवेदन: तपाईंको फोनमा स्थानीय रूपमा अत्यन्त सक्षम भाषा मोडेल](https://arxiv.org/abs/2404.14219) - - [Phi-4 प्राविधिक प्रतिवेदन](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini प्राविधिक प्रतिवेदन: मिश्रण-ऑफ-लोरा मार्फत सङ्कुचित तर शक्तिशाली बहुमाध्यमी भाषा मोडेलहरू](https://arxiv.org/abs/2503.01743) - - [सवारीमा फङ्शन-कलिङका लागि साना भाषा मोडेलहरू अनुकूलन गर्दै](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) बहु-चयन प्रश्न उत्तरको लागि PHI-3 फाइन-ट्यूनिङ: कार्यप्रणाली, परिणाम र चुनौतीहरू](https://arxiv.org/abs/2501.01588) - - [Phi-4-तर्क प्राविधिक प्रतिवेदन](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-मिनी-तर्क प्राविधिक प्रतिवेदन](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) - -## Phi मोडेलहरू प्रयोग + - कार्य कलिङ नमुना + - Phi-4 नमुना 🆕 + - [📓] [Phi-4-mini प्रयोग गरेर कार्य कलिङ](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Phi-4-mini सँग बहु-एजेन्ट सिर्जना गर्न कार्य कलिङ प्रयोग](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Ollama सँग कार्य कलिङ प्रयोग](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [ONNX सँग कार्य कलिङ प्रयोग](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - बहु-मोडल मिक्सिङ नमुना + - Phi-4 नमुना 🆕 + - [📓] [प्रविधि पत्रकारको रूपमा Phi-4-multimodal प्रयोग](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET कन्सोल अनुप्रयोग Phi-4-multimodal प्रयोग गरेर छविहरू विश्लेषण गर्न](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Phi नमुनाहरूमा विशेष अनुकूलन + - [विशेष अनुकूलन परिदृश्यहरू](./md/03.FineTuning/FineTuning_Scenarios.md) + - [विशेष अनुकूलन र RAG तुलना](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Phi-3 लाई उद्योग विशेषज्ञ बन्नुहोस् विशेष अनुकूलन](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [VS Code का लागि AI टूलकिटको साथ Phi-3 विशेष अनुकूलन](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure मशीन लर्निंग सेवा सँग Phi-3 विशेष अनुकूलन](./md/03.FineTuning/Introduce_AzureML.md) + - [Lora सँग Phi-3 विशेष अनुकूलन](./md/03.FineTuning/FineTuning_Lora.md) + - [QLora सँग Phi-3 विशेष अनुकूलन](./md/03.FineTuning/FineTuning_Qlora.md) + - [Microsoft Foundry सँग Phi-3 विशेष अनुकूलन](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Azure ML CLI/SDK सँग Phi-3 विशेष अनुकूलन](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Microsoft Olive सँग विशेष अनुकूलन](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Microsoft Olive Hands-On Lab सँग विशेष अनुकूलन](./md/03.FineTuning/olive-lab/readme.md) + - [Weights and Bias प्रयोग गरेर Phi-3-vision विशेष अनुकूलन](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [एप्पल MLX फ्रेमवर्कसँग Phi-3 को फाइन-ट्यूनिङ](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision को फाइन-ट्यूनिङ (औपचारिक समर्थन)](./md/03.FineTuning/FineTuning_Vision.md) + - [Kaito AKS, Azure कन्टेनरहरू (औपचारिक समर्थन) सँग Phi-3 को फाइन-ट्यूनिङ](./md/03.FineTuning/FineTuning_Kaito.md) + - [Phi-3 र 3.5 Vision को फाइन-ट्यूनिङ](https://github.com/2U1/Phi3-Vision-Finetune) + +- हैण्ड्स अन ल्याब + - [अत्याधुनिक मोडेलहरू अन्वेषण गर्दै: LLMs, SLMs, स्थानीय विकास र थप](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP क्षमता अनलक गर्दै: Microsoft Olive सँग फाइन-ट्यूनिङ](https://github.com/azure/Ignite_FineTuning_workshop) + +- शैक्षिक अनुसन्धान पत्रहरू र प्रकाशनहरू + - [Textbooks Are All You Need II: phi-1.5 प्राविधिक रिपोर्ट](https://arxiv.org/abs/2309.05463) + - [Phi-3 प्राविधिक रिपोर्ट: तपाईंको फोनमा स्थानीय रूपमा अत्यधिक सक्षम भाषा मोडेल](https://arxiv.org/abs/2404.14219) + - [Phi-4 प्राविधिक रिपोर्ट](https://arxiv.org/abs/2412.08905) + - [Phi-4-Mini प्राविधिक रिपोर्ट: मिक्स्चर-ऑफ-LoRAs मार्फत कम्प्याक्ट तर शक्तिशाली बहुमाध्यम भाषा मोडेलहरू](https://arxiv.org/abs/2503.01743) + - [सवारीसाधन भित्र कार्य-कलिङका लागि साना भाषा मोडेलहरूको अनुकूलन](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) बहुविकल्पीय प्रश्नोत्तरको लागि PHI-3 को फाइन-ट्यूनिङ: विधि, परिणामहरू, र चुनौतीहरू](https://arxiv.org/abs/2501.01588) + - [Phi-4-कारण प्राविधिक रिपोर्ट](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-मिनी-कारण प्राविधिक रिपोर्ट](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + +## Phi मोडेलहरू प्रयोग गर्दै ### Microsoft Foundry मा Phi -तपाईं Microsoft Phi कसरी प्रयोग गर्ने र विभिन्न हार्डवेयर उपकरणहरूमा E2E समाधानहरू कसरी निर्माण गर्ने सिक्न सक्नुहुन्छ। Phi आफैं अनुभव गर्न, मोडेलहरू सँग खेल्न सुरु गर्नुहोस् र तपाईंका परिदृश्यहरूको लागि Phi अनुकूलन गर्नुहोस् [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) मार्फत। तपाईं [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) मा Getting Started मा थप जान्न सक्नुहुन्छ। +तपाईं Microsoft Phi कसरी प्रयोग गर्ने र तपाईका विभिन्न हार्डवेयर उपकरणहरूमा ई2ई समाधान कसरी बनाउने भन्ने सिक्न सक्नुहुन्छ। आफैं Phi अनुभव गर्न, मोडेलहरूसँग खेल्न सुरु गर्नुहोस् र तपाईका परिदृश्यहरूको लागि Phi अनुकूलन गर्नुहोस् [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) प्रयोग गरेर। तपाईले [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) सँग सुरु गर्ने विषयमा थप जान्न सक्नुहुन्छ। -**प्लेटग्राउन्ड** -प्रत्येक मोडेलसँग मोडेल परीक्षण गर्न समर्पित प्लेटग्राउन्ड हुन्छ [Azure AI Playground](https://aka.ms/try-phi3) मा। +**प्लेलगाउन्ड** +प्रत्येक मोडेलसँग परीक्षण गर्नको लागि समर्पित प्लेलगाउन्ड छ [Azure AI Playground](https://aka.ms/try-phi3)। ### GitHub मोडेलहरूमा Phi -तपाईं Microsoft Phi कसरी प्रयोग गर्ने र विभिन्न हार्डवेयर उपकरणहरूमा E2E समाधानहरू कसरी निर्माण गर्ने सिक्न सक्नुहुन्छ। Phi आफैं अनुभव गर्न, मोडेलसँग खेल्न सुरु गर्नुहोस् र तपाईंका परिदृश्यहरूको लागि Phi अनुकूलन गर्नुहोस् [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) मार्फत। तपाईं [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) मा Getting Started मा थप जान्न सक्नुहुन्छ। +तपाईं Microsoft Phi कसरी प्रयोग गर्ने र तपाईका विभिन्न हार्डवेयर उपकरणहरूमा ई2ई समाधान कसरी बनाउने भन्ने सिक्न सक्नुहुन्छ। आफैं Phi अनुभव गर्न, मोडेलसँग खेल्न सुरु गर्नुहोस् र तपाईका परिदृश्यहरूको लागि Phi अनुकूलन गर्नुहोस् [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) प्रयोग गरेर। तपाईले [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) सँग सुरु गर्ने विषयमा थप जान्न सक्नुहुन्छ। -**प्लेटग्राउन्ड** -प्रत्येक मोडेलसँग समर्पित [प्लेटग्राउन्ड छ मोडेल परीक्षणको लागि](/md/02.QuickStart/GitHubModel_QuickStart.md)। +**प्लेलगाउन्ड** +प्रत्येक मोडेलसँग परीक्षण गर्नको लागि समर्पित [प्लेलगाउन्ड छ](/md/02.QuickStart/GitHubModel_QuickStart.md)। ### Hugging Face मा Phi -तपाईं मोडेल [Hugging Face](https://huggingface.co/microsoft) मा पनि फेला पार्न सक्नुहुन्छ। +तपाईं मोडेललाई पनि [Hugging Face](https://huggingface.co/microsoft) मा फेला पार्न सक्नुहुन्छ। -**प्लेटग्राउन्ड** -[Hugging Chat प्लेटग्राउन्ड](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**प्लेलगाउन्ड** + [Hugging Chat प्लेलगाउन्ड](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) ## 🎒 अन्य पाठ्यक्रमहरू -हाम्रो टोलीले अन्य पाठ्यक्रमहरू उत्पादित गर्दछ! जाँच गर्नुहोस्: +हाम्रो टोलीले अन्य पाठ्यक्रमहरू उत्पादन गर्दछ! हेर्नुहोस्: -### LangChain -[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain for Beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +### LangChain +[![सुरुआतीहरूका लागि LangChain4j](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![सुरुआतीहरूका लागि LangChain.js](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![सुरुआतीहरूका लागि LangChain](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agents -[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / एजेन्टहरू +[![सुरुआतीहरूका लागि AZD](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![सुरुआतीहरूका लागि Edge AI](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![सुरुआतीहरूका लागि MCP](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![सुरुआतीहरूका लागि AI एजेन्टहरू](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### जनरेटिभ AI शृंखला -[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### सृजनात्मक AI श्रृंखला +[![सुरुआतीहरूका लागि सृजनात्मक AI](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![सृजनात्मक AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![सृजनात्मक AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![सृजनात्मक AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- -### कोर सिकाइ -[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +### मुख्य सिकाइ +[![सुरुआतीहरूका लागि ML](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![सुरुआतीहरूका लागि डेटा विज्ञान](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![सुरुआतीहरूका लागि AI](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![सुरुआतीहरूका लागि साइबरसुरक्षा](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![सुरुआतीहरूका लागि वेब विकास](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![सुरुआतीहरूका लागि IoT](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![सुरुआतीहरूका लागि XR विकास](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Copilot शृंखला -[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +### Copilot श्रृंखला +[![AI सँग जोडी प्रोग्रामिङका लागि Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![C#/.NET का लागि Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot साहसिक](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## जिम्मेवार AI +## दायित्वपूर्बक AI -Microsoft ले हाम्रो ग्राहकहरूलाई AI उत्पादनहरू जिम्मेवार तरिकाले प्रयोग गर्न मद्दत गर्न प्रतिबद्ध छ, हाम्रो सिकाइहरू साझा गर्दै, र पारदर्शिता नोट्स तथा प्रभाव मूल्याङ्कन जस्ता उपकरणहरू मार्फत विश्वासमा आधारित साझेदारीहरू निर्माण गर्दै। यी मध्ये धेरै स्रोतहरू [https://aka.ms/RAI](https://aka.ms/RAI) मा भेट्न सकिन्छ। -Microsoft को जिम्मेवार AI दृष्टिकोण हाम्रो AI सिद्धान्तहरूमा आधारित छ: निष्पक्षता, विश्वसनीयता र सुरक्षा, गोपनीयता र सुरक्षा, समावेशिता, पारदर्शिता, र जवाफदेहिता। +Microsoft हाम्रा ग्राहकहरूलाई हाम्रो AI उत्पादनहरू दायित्वपूर्वक प्रयोग गर्न मद्दत गर्न, हाम्रो सिकाइहरू साझेदारी गर्न, र पारदर्शिता नोटहरू र प्रभाव मूल्यांकन जस्ता उपकरणहरू मार्फत विश्वासमा आधारित साझेदारीहरू निर्माण गर्न प्रतिबद्ध छ। यी मध्ये धेरै स्रोतहरू [https://aka.ms/RAI](https://aka.ms/RAI) मा फेला पार्न सकिन्छ। +Microsoft को दायित्वपूर्ण AI दृष्टिकोण हाम्रो निष्पक्षता, विश्वसनीयता र सुरक्षा, गोपनीयता र सुरक्षा, समावेशिता, पारदर्शिता, र जवाफदेहिताका AI सिद्धान्तहरूमा आधारित छ। -यो नमुना मा प्रयोग गरिएका ठूला आकारका प्राकृतिक भाषा, छवि, र भाषण मोडेलहरू संभावित रूपमा अनुचित, अविश्वसनीय, वा अपमानजनक तरिकाले व्यवहार गर्न सक्छन्, जसले नोक्सान पुर्याउन सक्छ। कृपया जोखिमहरू र सीमाहरूबारे जानकारी पाउन [Azure OpenAI सेवा पारदर्शिता नोट](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) हेर्नुहोस्। -यी जोखिमहरूलाई कम गर्ने सिफारिस गरिएका तरिका भनेको तपाइँको वास्तुकलामा एउटा सुरक्षा प्रणाली समावेश गर्नु हो जसले हानिकारक व्यवहार पत्ता लगाउन र रोक्न सक्षम हुन्छ। [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) एक स्वतन्त्र सुरक्षा तह प्रदान गर्दछ, जसले अनुप्रयोगहरू र सेवाहरूमा हानिकारक प्रयोगकर्ता-उत्पन्न र AI-उत्पन्न सामग्री पत्ता लगाउन सक्छ। Azure AI Content Safety मा पाठ र छवि API हरू समावेश छन् जसले तपाइँलाई हानिकारक सामग्री पत्ता लगाउन अनुमति दिन्छ। Microsoft Foundry भित्र, Content Safety सेवा तपाईंलाई विभिन्न मोडालिटीहरूमा हानिकारक सामग्री पत्ता लगाउन नमूना कोडहरू हेर्न, अन्वेषण गर्न र प्रयास गर्न अनुमति दिन्छ। तलको [छिटो सुरु गर्ने कागजात](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ले तपाईंलाई सेवामा अनुरोधहरू बनाउने प्रक्रिया देखाउँछ। +ठूलो स्तरका प्राकृतिक भाषा, छवि, र भाषण मोडेलहरू - जस्तै यस नमुनामा प्रयोग गरिएका - सम्भावित रूपमा निष्पक्ष छैनन्, अविश्वसनीय वा आपत्तिजनक व्यवहार गर्न सक्छन्, जसले नोक्सान पुर्‍याउन सक्छ। कृपया जोखिम र सीमाहरूको बारेमा जानकारी पाउन [Azure OpenAI सेवा पारदर्शिता नोट](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) परामर्श गर्नुहोस्। -अर्को पक्ष विचार गर्नुपर्ने कुरा भनेको समग्र अनुप्रयोग प्रदर्शन हो। बहु-मोडाल र बहु-मोडेल अनुप्रयोगहरूसँग, हामी प्रदर्शनले अर्थ राख्छ कि प्रणालीले तपाईं र तपाईंको प्रयोगकर्ताले अपेक्षा गरेजस्तै काम गर्दछ, जसमा हानिकारक नतिजा उत्पादन नगर्नु पनि समावेश छ। तपाईंले आफ्नो समग्र अनुप्रयोगको प्रदर्शनलाई [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) को उपयोग गरी मूल्यांकन गर्नु महत्त्वपूर्ण छ। तपाईंलाई [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) सिर्जना गर्न र मूल्यांकन गर्न पनि क्षमता छ। -तपाईं आफ्नो विकास वातावरणमा [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) को प्रयोग गरेर आफ्नो AI अनुप्रयोग मूल्यांकन गर्न सक्नुहुन्छ। परीक्षण डेटासेट वा लक्ष्य दिइएको अवस्थामा, तपाईंको जेनेरेटिभ AI अनुप्रयोगका उत्पादनहरू built-in evaluators वा आफ्नो रोजाइका custom evaluators द्वारा मात्रात्मक रूपमा मापन गरिन्छ। Azure AI Evaluation SDK प्रयोग गरेर आफ्नो प्रणाली मूल्यांकन गर्न सुरु गर्न, तपाईं [छिटो सुरु गर्ने मार्गदर्शन](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) अनुसरण गर्न सक्नुहुन्छ। मूल्यांकन चलाएपछि, तपाईं [Microsoft Foundry मा परिणामहरू दृश्यात्मक रूपमा देख्न](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) सक्नुहुन्छ। +यी जोखिमहरू कम गर्न सिफारिस गरिएको तरिका भनेको तपाईंको आर्किटेक्चरमा एउटा सुरक्षा प्रणाली समावेश गर्नु हो जसले हानिकारक व्यवहार पत्ता लगाउन र रोक्न सक्छ। [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) एउटा स्वतन्त्र सुरक्षा तह प्रदान गर्छ, जुन अनुप्रयोग र सेवाहरूमा हानिकारक प्रयोगकर्ता-जनित र AI-जनित सामग्री पत्ता लगाउन सक्षम छ। Azure AI Content Safety मा टेक्स्ट र इमेज APIहरू समावेश छन् जसले तपाईंलाई हानिकारक सामग्री पत्ता लगाउन अनुमति दिन्छ। Microsoft Foundry भित्र, Content Safety सेवा तपाईंलाई फरक-फरक प्रकारका हानिकारक सामग्री पत्ता लगाउन नमूना कोड हेर्न, अन्वेषण गर्न, र प्रयास गर्न अनुमति दिन्छ। निम्न [शीघ्र सुरु दस्तावेज](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ले तपाईंलाई सेवामा अनुरोधहरू गर्न मार्गदर्शन गर्छ। -## ट्रेडमार्कहरू +अर्को कुरामा ध्यान दिनुपर्ने हो भने समग्र अनुप्रयोगको प्रदर्शन हो। मल्टि-मोडल र मल्टि-मोडल अनुप्रयोगहरूसँग, हामीले प्रदर्शनलाई यस्तो अर्थ दिन्छौं कि प्रणाली तपाईं र तपाईंका प्रयोगकर्ताहरूले अपेक्षा गरेझैं काम गर्छ, जसमा हानिकारक नतिजा नजन्माउनु पनि पर्दछ। तपाईंले [प्रदर्शन र गुणस्तर र जोखिम र सुरक्षा मूल्यांकनकर्ता](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) प्रयोग गरेर तपाईंको समग्र अनुप्रयोगको प्रदर्शन मूल्यांकन गर्नु महत्वपूर्ण छ। तपाईंले [अनुकूली मूल्यांकनकर्ताहरू](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) सिर्जना र मूल्यांकन गर्ने क्षमता पनि राख्नुहुन्छ। -यो परियोजनामा परियोजना, उत्पादन वा सेवाहरूका लागि ट्रेडमार्क वा लोगोजहरू समावेश हुन सक्छन्। Microsoft ट्रेडमार्क वा लोगोजहरूको अनुमति प्राप्त प्रयोग [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) अनुसार हुनुपर्दछ र त्यसको पालना अनिवार्य छ। +तपाईं आफ्नो विकास वातावरणमा AI अनुप्रयोग मूल्यांकन गर्न सक्नुहुन्छ [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) प्रयोग गरेर। परीक्षण डाटासेट वा लक्ष्य दिइएको अवस्थामा, तपाईंको जेनेरेटिभ AI अनुप्रयोगका उत्पन्नहरू निर्माण गरिएको मूल्यांकनकर्ता वा तपाईंको रोजाइका अनुकूली मूल्यांकनकर्ताहरूको साथ मात्रात्मक रूपमा मापन गरिन्छ। तपाईंको प्रणाली मूल्यांकन गर्न azure ai evaluation sdk सँग सुरु गर्न [शीघ्र सुरु मार्गदर्शन](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) पालन गर्न सक्नुहुन्छ। एकपटक मूल्यांकन चलाएपछि, तपाईं [Microsoft Foundry मा परिणामहरू देख्न](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) सक्नुहुन्छ। + +## ट्रेडमार्कहरू -Microsoft ट्रेडमार्क वा लोगोजहरुको कुनै संशोधित संस्करणहरूमा प्रयोगले भ्रम सिर्जना गर्नु हुँदैन वा Microsoft प्रायोजन भएको अर्थ नलगाउनु पर्छ। तेस्रो-पक्ष ट्रेडमार्क वा लोगोजहरू प्रयोग गर्दा तेस्रो-पक्षको नीतिहरू लागू हुन्छन्। +यो परियोजनामा परियोजना, उत्पादन, वा सेवाका लागि ट्रेडमार्क वा लोगोहरू समावेश हुन सक्छ। Microsoft ट्रेडमार्क वा लोगोहरूको अधिकृत प्रयोग [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) अनुसार हुनुपर्छ र तिनको पालना गर्नुपर्दछ। +यस परियोजनाको परिमार्जित संस्करणमा Microsoft ट्रेडमार्क वा लोगोहरूको प्रयोगले भ्रम उत्पन्न गर्नु हुँदैन वा Microsoft प्रायोजनको संकेत गर्नु हुँदैन। तेस्रो-पक्ष ट्रेडमार्क वा लोगोहरूको कुनै पनि प्रयोग तिनका नीति अनुसार हुनुपर्दछ। -## सहयोग प्राप्त गर्ने +## सहयोग प्राप्त गर्ने तरिका -यदि तपाईं अलमलमा परे वा AI अनुप्रयोगहरू निर्माण गर्न सम्बन्धित कुनै प्रश्न भएमा, सहभागी हुनुहोस्: +तपाईं अड्किनुभए वा AI अनुप्रयोग निर्माणबारे कुनै प्रश्न भएमा, सामेल हुनुहोस्: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -यदि तपाईंसँग उत्पादन प्रतिक्रिया वा निर्माण गर्दा त्रुटिहरू छन् भने भ्रमण गर्नुहोस्: +यदि तपाईंलाई उत्पादन प्रतिक्रिया वा त्रुटिहरू छन् भने: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) @@ -321,5 +325,5 @@ Microsoft ट्रेडमार्क वा लोगोजहरुको **अस्वीकरण**: -यो दस्तावेज़ AI अनुवाद सेवा [Co-op Translator](https://github.com/Azure/co-op-translator) प्रयोग गरी अनुवाद गरिएको हो। हामी शुद्धताको लागि प्रयास गर्छौं, तर कृपया बुझ्नुहोस् कि स्वचालित अनुवादमा त्रुटिहरू वा गलतथापना हुनसक्छ। मूल दस्तावेज़लाई यसको मूल भाषामा आधिकारिक स्रोत मानिनुपर्छ। महत्वपूर्ण जानकारीका लागि, पेशेवर मानवीय अनुवाद सिफारिस गरिन्छ। यस अनुवादको प्रयोगबाट उत्पन्न हुने कुनै पनि गलत बुझाइ वा गलत व्याख्याको लागि हामी जिम्मेवार छैनौं। +यो दस्तावेज़ AI अनुवाद सेवा [Co-op Translator](https://github.com/Azure/co-op-translator) प्रयोग गरेर अनुवाद गरिएको हो। हामी सही हुन प्रयास गर्छौं, तर कृपया जानकार हुनुस् कि स्वचालित अनुवादमा त्रुटिहरू वा अशुद्धताहरू हुन सक्छन्। मूल दस्तावेज़ यसको मूल भाषामा आधिकारिक स्रोत मानिनुपर्छ। महत्वपूर्ण जानकारीका लागि व्यावसायिक मानव अनुवाद सिफारिस गरिन्छ। यस अनुवादको प्रयोगबाट उत्पन्न कुनै पनि गलत बुझाइ वा त्रुटिको लागि हामी जिम्मेवार छैनौं। \ No newline at end of file diff --git a/translations/ne/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/ne/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..70eb9d67b --- /dev/null +++ b/translations/ne/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 माइक्रोसफ्ट phi-4-mini र SQLite FTS5 सँगको Hybrid RAG (शून्य-क्लाउड SLM)\n", + "\n", + "> **लेखक:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **केन्द्रित:** साना भाषा मोडेलहरू (SLMs), SQLite FTS5 BM25, Dense Embeddings, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 १. प्रेरणा: स्थानीय SLMs मा कुञ्जीशब्द सम्झना समस्या\n", + "केवल dense वेक्टर ईम्बेडिङहरूमा निर्भर गर्ने मानक RAG संरचनाहरूले प्रायः ठ्याक्कै सङ्ख्यात्मक टोकनहरू (जस्तै, `2,340,000 TL`, सम्झौता कोडहरू, खाता नम्बरहरू) फेला पार्न असफल हुन्छन्। \n", + "त्यसको विपरीत, sparse lexical खोज (BM25) ले अर्थमैत्री पर्यायवाची र पेराफ्रेज गरिएको प्रश्नहरू छुटाउँछ।\n", + "\n", + "यो कुकबुकले कसरी एक **सर्वोच्च-गतिको, मेमोरीमा आधारित Hybrid Retrieval इन्जिन** कार्यान्वयन गर्ने देखाउँछ जसले मिलाउँछ:\n", + "१. **Dense Vectors** (Cosine Similarity)\n", + "२. **Sparse Lexical Search** (SQLite FTS5 BM25)\n", + "३. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "४. **Grounded Citation Generation (`[1]`, `[2]`)** माइक्रोसफ्ट `phi-4-mini` सँग।\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ २. दोहोरो SQLite स्कीमा (घना भेक्टरहरू + भर्चुअल FTS5 BM२५ तालिका)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 ३। नमुना इनजेस्टन र कार्यान्वयन बेंचमार्क\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 ४. Microsoft phi-4-mini का लागि ग्राउन्डेड प्रॉम्प्ट निर्माण\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**अस्वीकरण**:\nयो दस्तावेज़ AI अनुवाद सेवा [Co-op Translator](https://github.com/Azure/co-op-translator) प्रयोग गरेर अनुवाद गरिएको हो। हामी सही हुन प्रयास गर्छौं, तर कृपया जानकार हुनुस् कि स्वचालित अनुवादमा त्रुटिहरू वा अशुद्धताहरू हुन सक्छन्। मूल दस्तावेज़ यसको मूल भाषामा आधिकारिक स्रोत मानिनुपर्छ। महत्वपूर्ण जानकारीका लागि व्यावसायिक मानव अनुवाद सिफारिस गरिन्छ। यस अनुवादको प्रयोगबाट उत्पन्न कुनै पनि गलत बुझाइ वा त्रुटिको लागि हामी जिम्मेवार छैनौं।\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/nl/.co-op-translator.json b/translations/nl/.co-op-translator.json index 29aecb897..2ff232c90 100644 --- a/translations/nl/.co-op-translator.json +++ b/translations/nl/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "nl" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:04:55+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:03:22+00:00", "source_file": "README.md", "language_code": "nl" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "nl" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:08:17+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "nl" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:23+00:00", diff --git a/translations/nl/README.md b/translations/nl/README.md index c85adc558..fc5458a19 100644 --- a/translations/nl/README.md +++ b/translations/nl/README.md @@ -1,42 +1,42 @@ -# Phi Kookboek: Praktische Voorbeelden met Microsofts Phi-modellen +# Phi Kookboek: Praktische Voorbeelden met Microsoft's Phi Modellen [![Open en gebruik de voorbeelden in GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Open in Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub bijdragers](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub issues](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub pull-requests](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs welkom](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub problemen](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub pull-verzoeken](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Welkom](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub waarnemers](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub volgers](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub sterren](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi is een serie open source AI-modellen ontwikkeld door Microsoft. +Phi is een reeks open source AI-modellen ontwikkeld door Microsoft. -Phi is momenteel het krachtigste en meest kosteneffectieve kleine taalmodel (SLM), met zeer goede benchmarks in meertaligheid, redeneren, tekst/chatgeneratie, coderen, afbeeldingen, audio en andere scenario's. +Phi is momenteel het krachtigste en meest kosteneffectieve kleine taalmodel (SLM), met zeer goede benchmarks in meertaligheid, redeneren, tekst-/chatgeneratie, coderen, afbeeldingen, audio en andere scenario's. -Je kunt Phi inzetten in de cloud of op edge-apparaten, en je kunt gemakkelijk generatieve AI-toepassingen bouwen met beperkte rekenkracht. +Je kunt Phi implementeren in de cloud of op edge-apparaten, en je kunt gemakkelijk generatieve AI-toepassingen bouwen met beperkte rekenkracht. Volg deze stappen om te beginnen met het gebruik van deze bronnen: -1. **Fork de repository**: Klik [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Clone de repository**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Word lid van de Microsoft AI Discord-gemeenschap en ontmoet experts en mede-ontwikkelaars**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +1. **Fork de Repository**: Klik [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Clone de Repository**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Word lid van de Microsoft AI Discord Community en ontmoet experts en mededevelopers**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/nl/cover.eb18d1b9605d754b.webp) ### 🌐 Meertalige Ondersteuning -#### Ondersteund via GitHub Action (Automatisch & Altijd Actueel) +#### Ondersteund via GitHub Action (Geautomatiseerd & Altijd Up-to-Date) -[Arabisch](../ar/README.md) | [Bengaals](../bn/README.md) | [Bulgaars](../bg/README.md) | [Birmaans (Myanmar)](../my/README.md) | [Chinees (Vereenvoudigd)](../zh-CN/README.md) | [Chinees (Traditioneel, Hong Kong)](../zh-HK/README.md) | [Chinees (Traditioneel, Macau)](../zh-MO/README.md) | [Chinees (Traditioneel, Taiwan)](../zh-TW/README.md) | [Kroatisch](../hr/README.md) | [Tsjechisch](../cs/README.md) | [Deens](../da/README.md) | [Nederlands](./README.md) | [Ests](../et/README.md) | [Fins](../fi/README.md) | [Frans](../fr/README.md) | [Duits](../de/README.md) | [Grieks](../el/README.md) | [Hebreeuws](../he/README.md) | [Hindi](../hi/README.md) | [Hongaars](../hu/README.md) | [Indonesisch](../id/README.md) | [Italiaans](../it/README.md) | [Japans](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreaans](../ko/README.md) | [Litouws](../lt/README.md) | [Maleis](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalees](../ne/README.md) | [Nigeriaans Pidgin](../pcm/README.md) | [Noors](../no/README.md) | [Perzisch (Farsi)](../fa/README.md) | [Pools](../pl/README.md) | [Portugees (Brazilië)](../pt-BR/README.md) | [Portugees (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Roemeens](../ro/README.md) | [Russisch](../ru/README.md) | [Servisch (Cyrillisch)](../sr/README.md) | [Slowaaks](../sk/README.md) | [Sloveens](../sl/README.md) | [Spaans](../es/README.md) | [Swahili](../sw/README.md) | [Zweeds](../sv/README.md) | [Tagalog (Filipijns)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thais](../th/README.md) | [Turks](../tr/README.md) | [Oekraïens](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamees](../vi/README.md) +[Arabisch](../ar/README.md) | [Bengaals](../bn/README.md) | [Bulgaars](../bg/README.md) | [Birmaans (Myanmar)](../my/README.md) | [Chinees (Vereenvoudigd)](../zh-CN/README.md) | [Chinees (Traditioneel, Hong Kong)](../zh-HK/README.md) | [Chinees (Traditioneel, Macau)](../zh-MO/README.md) | [Chinees (Traditioneel, Taiwan)](../zh-TW/README.md) | [Kroatisch](../hr/README.md) | [Tsjechisch](../cs/README.md) | [Deens](../da/README.md) | [Nederlands](./README.md) | [Estisch](../et/README.md) | [Fins](../fi/README.md) | [Frans](../fr/README.md) | [Duits](../de/README.md) | [Grieks](../el/README.md) | [Hebreeuws](../he/README.md) | [Hindi](../hi/README.md) | [Hongaars](../hu/README.md) | [Indonesisch](../id/README.md) | [Italiaans](../it/README.md) | [Japans](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreaans](../ko/README.md) | [Litouws](../lt/README.md) | [Maleis](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalees](../ne/README.md) | [Nigeriaans Pidgin](../pcm/README.md) | [Noors](../no/README.md) | [Perzisch (Farsi)](../fa/README.md) | [Pools](../pl/README.md) | [Portugees (Brazilië)](../pt-BR/README.md) | [Portugees (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Roemeens](../ro/README.md) | [Russisch](../ru/README.md) | [Servisch (Cyrillisch)](../sr/README.md) | [Slowaaks](../sk/README.md) | [Sloveens](../sl/README.md) | [Spaans](../es/README.md) | [Swahili](../sw/README.md) | [Zweeds](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turks](../tr/README.md) | [Oekraïens](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamees](../vi/README.md) > **Liever lokaal clonen?** > -> Deze repository bevat 50+ vertalingen, wat de downloadgrootte aanzienlijk vergroot. Om zonder vertalingen te clonen, gebruik sparse checkout: +> Deze repository bevat meer dan 50 taalvertalingen wat de downloadgrootte aanzienlijk verhoogt. Om te clonen zonder vertalingen, gebruik sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -57,142 +57,144 @@ Volg deze stappen om te beginnen met het gebruik van deze bronnen: ## Inhoudsopgave -- Introductie - - [Welkom bij de Phi-familie](./md/01.Introduction/01/01.PhiFamily.md) +- Inleiding + - [Welkom bij de Phi Familie](./md/01.Introduction/01/01.PhiFamily.md) - [Je omgeving opzetten](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Belangrijke technologieën begrijpen](./md/01.Introduction/01/01.Understandingtech.md) - - [AI-veiligheid voor Phi-modellen](./md/01.Introduction/01/01.AISafety.md) + - [Belangrijke Technologieën Begrijpen](./md/01.Introduction/01/01.Understandingtech.md) + - [AI Veiligheid voor Phi Modellen](./md/01.Introduction/01/01.AISafety.md) - [Phi Hardware Ondersteuning](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi-modellen & Beschikbaarheid op verschillende platformen](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Phi Modellen & Beschikbaarheid op platformen](./md/01.Introduction/01/01.Edgeandcloud.md) - [Gebruik van Guidance-ai en Phi](./md/01.Introduction/01/01.Guidance.md) - [GitHub Marketplace Modellen](https://github.com/marketplace/models) - [Azure AI Model Catalogus](https://ai.azure.com) -- Inference Phi in verschillende omgevingen +- Inferentie Phi in verschillende omgevingen - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub Modellen](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry Model Catalogus](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - -- Inference Phi Familie - - [Inference Phi op iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Inference Phi op Android](./md/01.Introduction/03/Android_Inference.md) - - [Inference Phi op Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Inference Phi op AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Inference Phi met Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Inference Phi op lokale server](./md/01.Introduction/03/Local_Server_Inference.md) - - [Inference Phi op remote server met AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Inference Phi met Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inference Phi--Vision lokaal](./md/01.Introduction/03/Vision_Inference.md) - - [Inference Phi met Kaito AKS, Azure Containers (officiële ondersteuning)](./md/01.Introduction/03/Kaito_Inference.md) -- [Quantificeren Phi Familie](./md/01.Introduction/04/QuantifyingPhi.md) - - [Quantificeren Phi-3.5 / 4 met llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Quantificeren Phi-3.5 / 4 met Generative AI extensies voor onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Quantificeren Phi-3.5 / 4 met Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Quantificeren Phi-3.5 / 4 met Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Foundry Lokaal](./md/01.Introduction/02/07.FoundryLocal.md) + +- Inferentie Phi Familie + - [Inferentie Phi op iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Inferentie Phi op Android](./md/01.Introduction/03/Android_Inference.md) + - [Inferentie Phi op Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Inferentie Phi op AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Inferentie Phi met Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Inferentie Phi op Lokale Server](./md/01.Introduction/03/Local_Server_Inference.md) + - [Inferentie Phi op Externe Server met AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Inferentie Phi met Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Inferentie Phi--Vision lokaal](./md/01.Introduction/03/Vision_Inference.md) + - [Inferentie Phi met Kaito AKS, Azure Containers (officiële ondersteuning)](./md/01.Introduction/03/Kaito_Inference.md) +- [Quantificatie Phi Familie](./md/01.Introduction/04/QuantifyingPhi.md) + - [Quantificatie Phi-3.5 / 4 met llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Quantificatie Phi-3.5 / 4 met Generative AI extensies voor onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Quantificatie Phi-3.5 / 4 met Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Quantificatie Phi-3.5 / 4 met Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Evaluatie Phi - - [Verantwoorde AI](./md/01.Introduction/05/ResponsibleAI.md) + - [Verantwoordelijke AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry voor Evaluatie](./md/01.Introduction/05/AIFoundry.md) - [Gebruik van Promptflow voor Evaluatie](./md/01.Introduction/05/Promptflow.md) - RAG met Azure AI Search - - [Hoe Phi-4-mini en Phi-4-multimodal(RAG) te gebruiken met Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Hoe Phi-4-mini en Phi-4-multimodaal(RAG) te gebruiken met Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud Lokale Hybride RAG met SQLite FTS5 en phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Voorbeelden van Phi applicatieontwikkeling - - Tekst & Chat Applicaties - - Phi-4 Voorbeelden + - Tekst- & Chattoepassingen + - Phi-4 Voorbeelden - [📓] [Chatten met Phi-4-mini ONNX Model](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Chat met Phi-4 lokaal ONNX Model .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - [Chat .NET Console App met Phi-4 ONNX met Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 Voorbeelden - - [Lokale chatbot in de browser met Phi3, ONNX Runtime Web en WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Lokale Chatbot in de browser met Phi3, ONNX Runtime Web en WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [Multi Model - Interactieve Phi-3-mini en OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Een wrapper bouwen en Phi-3 gebruiken met MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [Modeloptimalisatie - Hoe het Phi-3-min model te optimaliseren voor ONNX Runtime Web met Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3-app met Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + - [WinUI3 App met Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) -[WinUI3 Multi Model AI Powered Notes App Voorbeeld](https://github.com/microsoft/ai-powered-notes-winui3-sample) - [Fijn afstemmen en integreren van aangepaste Phi-3 modellen met Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - [Fijn afstemmen en integreren van aangepaste Phi-3 modellen met Prompt flow in Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Beoordeel het fijn afgestemde Phi-3 / Phi-3.5 model in Microsoft Foundry met focus op Microsoft's principes voor Verantwoordelijke AI](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [Evalueer het fijn afgestemde Phi-3 / Phi-3.5 model in Microsoft Foundry met focus op Microsoft's Verantwoordelijke AI Principes](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Phi-3.5-mini-instruct taalvoorspellingsvoorbeeld (Chinees/Engels)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU gebruiken om een Prompt flow-oplossing te maken met Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Windows GPU gebruiken om Prompt flow-oplossing te creëren met Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Microsoft Phi-3.5 tflite gebruiken om Android-app te maken](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET voorbeeld met lokaal ONNX Phi-3 model met Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Console chat .NET-app met Semantic Kernel en Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Q&A .NET Voorbeeld met lokaal ONNX Phi-3 model met Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Console chat .NET app met Semantic Kernel en Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK Codegebaseerde Voorbeelden - - Phi-4 Voorbeelden - - [📓] [Genereer projectcode met Phi-4-multimodaal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Azure AI Inference SDK Code Gebaseerde Voorbeelden + - Phi-4 Voorbeelden + - [📓] [Genereer projectcode met Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 Voorbeelden - [Bouw je eigen Visual Studio Code GitHub Copilot Chat met Microsoft Phi-3 Familie](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - [Maak je eigen Visual Studio Code Chat Copilot Agent met Phi-3.5 door GitHub Modellen](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Geavanceerde Redeneringsvoorbeelden - - Phi-4 Voorbeelden + - Phi-4 Voorbeelden - [📓] [Phi-4-mini-redenering of Phi-4-redenering Voorbeelden](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - [📓] [Fijn afstemmen Phi-4-mini-redenering met Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Fijn afstemmen Phi-4-mini-redenering met Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-redenering met GitHub Modellen](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-redenering met Microsoft Foundry Modellen](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demo's - - [Phi-4-mini demos gehost op Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodale demos gehost op Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Vision Voorbeelden - - Phi-4 Voorbeelden - - [📓] [Gebruik Phi-4-multimodaal om afbeeldingen te lezen en code te genereren](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [Phi-4-mini demo's gehost op Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal demo's gehost op Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Visie Voorbeelden + - Phi-4 Voorbeelden + - [📓] [Gebruik Phi-4-multimodal om afbeeldingen te lezen en code te genereren](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 Voorbeelden - - [📓][Phi-3-vision-Afbeelding tekst naar tekst](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 Recyclen](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Visuele taalassistent - met Phi3-Vision en OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision multi-frame of multi-image voorbeeld](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision Lokaal ONNX Model met Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menu-gebaseerd Phi-3 Vision Lokaal ONNX Model met Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Redenering-Vision Voorbeelden - - Phi-4-Redenering-Vision-15B - - [📓] [Gebruik Phi-4-Redenering-Vision-15B om oversteken buiten het zebrapad te detecteren](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Gebruik Phi-4-Redenering-Vision-15B voor wiskunde](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Gebruik Phi-4-Redenering-Vision-15B om UI te detecteren](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓][Phi-3-visie-Afbeelding tekst naar tekst](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-visie-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-visie CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [DEMO: Phi-3 Recycling](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-visie - Visuele taalassistent - met Phi3-Vision en OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Visie Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Visie OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Visie multi-frame of multi-afbeelding voorbeeld](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Visie lokaal ONNX model met Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Menu-gebaseerd Phi-3 Visie lokaal ONNX model met Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Redenering-Visie Voorbeelden + - Phi-4-Redenering-Visie-15B + - [📓] [Gebruik Phi-4-Redenering-Visie-15B om overtredingen te detecteren](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Gebruik Phi-4-Redenering-Visie-15B voor wiskunde](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Gebruik Phi-4-Redenering-Visie-15B om UI te detecteren](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Wiskunde Voorbeelden - Phi-4-Mini-Flash-Redenering-Instruct Voorbeelden [Wiskunde Demo met Phi-4-Mini-Flash-Redenering-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - Audio Voorbeelden - - Phi-4 Voorbeelden - - [📓] [Audio transcripties extraheren met Phi-4-multimodaal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodaal Audio Voorbeeld](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodaal Spraakvertaling Voorbeeld](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET console applicatie met Phi-4-multimodaal Audio om een audiobestand te analyseren en transcript te genereren](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - Phi-4 Voorbeelden + - [📓] [Audio transcripties extraheren met Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal Audio Voorbeeld](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal Spraak Vertalingsvoorbeeld](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET console-applicatie die Phi-4-multimodal Audio gebruikt om een audiobestand te analyseren en transcript te genereren](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE Voorbeelden - Phi-3 / 3.5 Voorbeelden - - [📓] [Phi-3.5 Mixture of Experts Modellen (MoEs) Social Media Voorbeeld](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Een Retrieval-Augmented Generation (RAG) Pipeline bouwen met NVIDIA NIM Phi-3 MOE, Azure AI Search en LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 Mengeling van Experts Modellen (MoEs) Social Media Voorbeeld](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Bouwen van een Retrieval-Augmented Generation (RAG) Pipeline met NVIDIA NIM Phi-3 MOE, Azure AI Search, en LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - Function Calling Voorbeelden - Phi-4 Voorbeelden 🆕 - - [📓] [Function Calling gebruiken met Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Function Calling gebruiken om multi-agents te maken met Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Function Calling gebruiken met Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Function Calling gebruiken met ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - [📓] [Gebruik Function Calling met Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Gebruik Function Calling om multi-agenten te creëren met Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Gebruik Function Calling met Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Gebruik Function Calling met ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - Multimodale Meng Voorbeelden - Phi-4 Voorbeelden 🆕 - - [📓] [Phi-4-multimodaal gebruiken als technologiejournalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET console applicatie met Phi-4-multimodaal om beelden te analyseren](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [📓] [Gebruik Phi-4-multimodal als technologiejournalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET console-applicatie die Phi-4-multimodal gebruikt om afbeeldingen te analyseren](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - Fijn afstemmen Phi Voorbeelden - - [Fijn afstemmen scenario's](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Fijn afstemmen versus RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Fijn afstemmen Scenario's](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Fijn afstemmen vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - [Fijn afstemmen Laat Phi-3 een industrie-expert worden](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - [Fijn afstemmen Phi-3 met AI Toolkit voor VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [Fijn afstemmen Phi-3 met Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) @@ -202,123 +204,126 @@ Volg deze stappen om te beginnen met het gebruik van deze bronnen: - [Fijn afstemmen Phi-3 met Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - [Fijn afstemmen met Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Fijn afstemmen met Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - - [Fijn afstemmen Phi-3-vision met Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Fijn afstemmen Phi-3 met Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Fijn afstemmen Phi-3-vision (officiële ondersteuning)](./md/03.FineTuning/FineTuning_Vision.md) - - [Fijn afstemmen van Phi-3 met Kaito AKS, Azure Containers (officiële ondersteuning)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Fijn afstemmen Phi-3-visie met Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Fijn afstemmen van Phi-3 met Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Fijn afstemmen van Phi-3-vision (officiële ondersteuning)](./md/03.FineTuning/FineTuning_Vision.md) + - [Fijn afstemmen van Phi-3 met Kaito AKS , Azure Containers (officiële ondersteuning)](./md/03.FineTuning/FineTuning_Kaito.md) - [Fijn afstemmen van Phi-3 en 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) -- Praktijklab - - [Verkennen van geavanceerde modellen: LLM's, SLM's, lokale ontwikkeling en meer](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP-potentieel ontgrendelen: Fijn afstemmen met Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) +- Hands-on Lab + - [Verkennen van baanbrekende modellen: LLMs, SLMs, lokale ontwikkeling en meer](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Unlocking NLP Potential: Fijn afstemmen met Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Academische Onderzoeksartikelen en Publicaties - - [Leerboeken zijn alles wat je nodig hebt II: phi-1.5 technisch rapport](https://arxiv.org/abs/2309.05463) - - [Phi-3 Technisch Rapport: Een zeer capabel taalmodel lokaal op je telefoon](https://arxiv.org/abs/2404.14219) +- Academische onderzoeksartikelen en publicaties + - [Textbooks Are All You Need II: phi-1.5 technisch rapport](https://arxiv.org/abs/2309.05463) + - [Phi-3 Technisch Rapport: Een zeer bekwaam taalmodel lokaal op je telefoon](https://arxiv.org/abs/2404.14219) - [Phi-4 Technisch Rapport](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini Technisch Rapport: Compacte maar krachtige multimodale taalmodellen via Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Optimalisatie van kleine taalmodellen voor in-voertuig functie-aanroepen](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Fijn afstemmen van PHI-3 voor meerkeuzevragen: Methodologie, resultaten en uitdagingen](https://arxiv.org/abs/2501.01588) + - [Phi-4-Mini Technisch Rapport: Compacte maar krachtige multimodale taalmodellen via mengsel van LoRAs](https://arxiv.org/abs/2503.01743) + - [Optimaliseren van kleine taalmodellen voor in-voertuig functie-aanroepen](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Fijn afstemmen van PHI-3 voor multiple-choice vraag beantwoording: Methode, resultaten en uitdagingen](https://arxiv.org/abs/2501.01588) - [Phi-4-redenering Technisch Rapport](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Phi-4-mini-redenering Technisch Rapport](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Phi-modellen gebruiken +## Gebruik van Phi Modellen ### Phi op Microsoft Foundry -Je kunt leren hoe je Microsoft Phi gebruikt en hoe je end-to-end oplossingen bouwt op je verschillende hardware-apparaten. Om Phi zelf te ervaren, begin met het uitproberen van de modellen en het aanpassen van Phi aan jouw scenario’s via de [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Je kunt meer leren in de introductie met [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Je kunt leren hoe je Microsoft Phi gebruikt en hoe je end-to-end oplossingen bouwt op je verschillende hardware-apparaten. Om Phi zelf te ervaren, begin met spelen met de modellen en het aanpassen van Phi voor jouw scenario’s via de [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) je kunt meer leren bij Aan de slag met [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Speelomgeving** -Elk model heeft een speciale speelomgeving om het model te testen [Azure AI Playground](https://aka.ms/try-phi3). +**Speeltuin** +Elk model heeft een speciale speeltuin om het model te testen [Azure AI Playground](https://aka.ms/try-phi3). ### Phi op GitHub Modellen -Je kunt leren hoe je Microsoft Phi gebruikt en hoe je end-to-end oplossingen bouwt op je verschillende hardware-apparaten. Om Phi zelf te ervaren, begin met het uitproberen van het model en het aanpassen van Phi aan jouw scenario’s via de [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Je kunt meer leren in de introductie met [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Je kunt leren hoe je Microsoft Phi gebruikt en hoe je end-to-end oplossingen bouwt op je verschillende hardware-apparaten. Om Phi zelf te ervaren, begin met spelen met het model en het aanpassen van Phi voor jouw scenario’s via de [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) je kunt meer leren bij Aan de slag met [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Speelomgeving** -Elk model heeft een speciale [speelomgeving om het model te testen](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Speeltuin** +Elk model heeft een speciale [speeltuin om het model te testen](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi op Hugging Face -Je kunt het model ook vinden op [Hugging Face](https://huggingface.co/microsoft) +Je kunt het model ook vinden op de [Hugging Face](https://huggingface.co/microsoft) -**Speelomgeving** -[Hugging Chat speelomgeving](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Speeltuin** + [Hugging Chat speeltuin](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Andere cursussen + ## 🎒 Andere cursussen Ons team produceert ook andere cursussen! Bekijk: ### LangChain -[![LangChain4j voor beginners](https://img.shields.io/badge/LangChain4j%20voor%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js voor beginners](https://img.shields.io/badge/LangChain.js%20voor%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain voor beginners](https://img.shields.io/badge/LangChain%20voor%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain4j voor beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js voor beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain voor beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- ### Azure / Edge / MCP / Agents -[![AZD voor beginners](https://img.shields.io/badge/AZD%20voor%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI voor beginners](https://img.shields.io/badge/Edge%20AI%20voor%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP voor beginners](https://img.shields.io/badge/MCP%20voor%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents voor beginners](https://img.shields.io/badge/AI%20Agents%20voor%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AZD voor beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI voor beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP voor beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents voor beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Generatieve AI-serie -[![Generatieve AI voor beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generatieve AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generatieve AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generatieve AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### Generative AI Serie +[![Generative AI voor beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- ### Kernleren -[![ML voor beginners](https://img.shields.io/badge/ML%20voor%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Datawetenschap voor beginners](https://img.shields.io/badge/Data%20Science%20voor%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI voor beginners](https://img.shields.io/badge/AI%20voor%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity voor beginners](https://img.shields.io/badge/Cybersecurity%20voor%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Webontwikkeling voor beginners](https://img.shields.io/badge/Web%20Dev%20voor%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT voor beginners](https://img.shields.io/badge/IoT%20voor%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR-ontwikkeling voor beginners](https://img.shields.io/badge/XR%20Development%20voor%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![ML voor beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Datawetenschap voor beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI voor beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersecurity voor beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Webontwikkeling voor beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT voor beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR-ontwikkeling voor beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Copilot-reeks -[![Copilot voor AI-gepaard programmeren](https://img.shields.io/badge/Copilot%20voor%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot voor C#/.NET](https://img.shields.io/badge/Copilot%20voor%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +### Copilot Serie +[![Copilot voor AI-gepaired programmeren](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot voor C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot Avontuur](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Verantwoorde AI +## Verantwoord AI + +Microsoft zet zich in om onze klanten te helpen onze AI-producten verantwoordelijk te gebruiken, onze inzichten te delen en op vertrouwen gebaseerde partnerschappen op te bouwen door middel van tools zoals Transparantienota’s en Impactbeoordelingen. Veel van deze bronnen zijn te vinden op [https://aka.ms/RAI](https://aka.ms/RAI). +De benadering van Microsoft voor verantwoord AI is gebaseerd op onze AI-principes van rechtvaardigheid, betrouwbaarheid en veiligheid, privacy en beveiliging, inclusiviteit, transparantie en verantwoording. + +Groot-schalige taal-, beeld- en spraakmodellen - zoals die gebruikt worden in dit voorbeeld - kunnen zich mogelijk op manieren gedragen die oneerlijk, onbetrouwbaar of aanstootgevend zijn, en daarmee schade veroorzaken. Raadpleeg de [Azure OpenAI service Transparantienota](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) om geïnformeerd te zijn over risico’s en beperkingen. -Microsoft zet zich in om onze klanten te helpen onze AI-producten verantwoordelijk te gebruiken, onze ervaringen te delen en op vertrouwen gebaseerde partnerschappen op te bouwen via tools zoals Transparantienotities en Impactbeoordelingen. Veel van deze bronnen zijn te vinden op [https://aka.ms/RAI](https://aka.ms/RAI). -De aanpak van Microsoft voor verantwoorde AI is gebaseerd op onze AI-principes van eerlijkheid, betrouwbaarheid en veiligheid, privacy en beveiliging, inclusiviteit, transparantie en verantwoordelijkheid. -Grootschalige natuurlijke taal-, beeld- en spraakmodellen - zoals degene die in dit voorbeeld worden gebruikt - kunnen zich mogelijk op een oneerlijke, onbetrouwbare of aanstootgevende manier gedragen, wat kan leiden tot schade. Raadpleeg de [Azure OpenAI service Transparantienotitie](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) om geïnformeerd te worden over risico’s en beperkingen. -De aanbevolen aanpak om deze risico's te beperken is het opnemen van een veiligheidssysteem in uw architectuur dat schadelijk gedrag kan detecteren en voorkomen. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) biedt een onafhankelijke beschermingslaag, die schadelijke door gebruikers en AI gegenereerde inhoud in toepassingen en diensten kan detecteren. Azure AI Content Safety bevat tekst- en afbeeldings-API's die u in staat stellen materiaal te detecteren dat schadelijk is. Binnen Microsoft Foundry stelt de Content Safety-service u in staat voorbeeldcode te bekijken, te verkennen en uit te proberen voor het detecteren van schadelijke inhoud over verschillende modaliteiten. De volgende [quickstart documentatie](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) begeleidt u bij het maken van verzoeken aan de service. +De aanbevolen aanpak om deze risico’s te verminderen is het opnemen van een veiligheidssysteem in uw architectuur dat schadelijk gedrag kan detecteren en voorkomen. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) biedt een onafhankelijke beschermingslaag die in staat is schadelijke door gebruikers gegenereerde en door AI gegenereerde inhoud in applicaties en diensten te detecteren. Azure AI Content Safety bevat tekst- en beeld-API’s waarmee u schadelijk materiaal kunt opsporen. Binnen Microsoft Foundry kunt u met de Content Safety-service voorbeeldcode bekijken, verkennen en uitproberen voor het detecteren van schadelijke inhoud in verschillende modaliteiten. De volgende [quickstart-documentatie](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) begeleidt u bij het maken van verzoeken aan de service. -Een ander aspect om rekening mee te houden is de algehele prestatie van de applicatie. Bij multi-modale en multi-modelapplicaties verstaan we onder prestatie dat het systeem presteert zoals u en uw gebruikers verwachten, inclusief het niet genereren van schadelijke outputs. Het is belangrijk de prestaties van uw algehele applicatie te beoordelen met behulp van [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). U heeft ook de mogelijkheid om te creëren en evalueren met [aangepaste evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Een ander aspect om rekening mee te houden is de algehele prestaties van de applicatie. Bij multimodale en multimodelapplicaties zien we prestaties als het systeem presteert zoals u en uw gebruikers verwachten, waaronder het niet genereren van schadelijke outputs. Het is belangrijk om de prestaties van uw gehele applicatie te beoordelen met behulp van [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). U hebt ook de mogelijkheid om te creëren en te evalueren met [aangepaste evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -U kunt uw AI-toepassing evalueren in uw ontwikkelomgeving met behulp van de [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Gegeven een testdataset of een doel worden uw generatieve AI-toepassingsgeneraties kwantitatief gemeten met ingebouwde evaluators of aangepaste evaluators naar keuze. Om aan de slag te gaan met de azure ai evaluation sdk om uw systeem te evalueren, kunt u de [quickstart gids](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) volgen. Zodra u een evaluatieronde uitvoert, kunt u de resultaten [visualiseren in Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +U kunt uw AI-applicatie evalueren in uw ontwikkelomgeving met de [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Op basis van een testdataset of een doel worden de generaties van uw generatieve AI-applicatie kwantitatief gemeten met ingebouwde evaluators of aangepaste evaluators naar keuze. Om te beginnen met de azure ai evaluation sdk om uw systeem te evalueren, kunt u de [quickstart-gids](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) volgen. Nadat u een evaluatierun hebt uitgevoerd, kunt u de [resultaten visualiseren in Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Handelsmerken -Dit project kan handelsmerken of logo's bevatten voor projecten, producten of diensten. Het geautoriseerde gebruik van Microsoft-handelsmerken of logo's is onderworpen aan en moet voldoen aan de [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Gebruik van Microsoft-handelsmerken of logo's in gewijzigde versies van dit project mag geen verwarring veroorzaken of impliceren dat Microsoft het sponsort. Elk gebruik van handelsmerken of logo's van derden is onderworpen aan het beleid van die derden. +Dit project kan handelsmerken of logo’s bevatten van projecten, producten of diensten. Het geautoriseerde gebruik van Microsoft-handelsmerken of logo’s is onderworpen aan en moet voldoen aan de [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Het gebruik van Microsoft-handelsmerken of logo’s in gewijzigde versies van dit project mag geen verwarring veroorzaken of Microsoft-sponsoring impliceren. Elk gebruik van handelsmerken of logo’s van derden is onderworpen aan het beleid van die derden. ## Hulp krijgen -Als u vastloopt of vragen hebt over het bouwen van AI-apps, sluit u aan bij: +Als u vastloopt of vragen heeft over het bouwen van AI-apps, sluit dan aan bij: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Als u productfeedback of fouten hebt tijdens het bouwen, bezoek dan: +Als u productfeedback of fouten heeft tijdens het bouwen, bezoek dan: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Disclaimer**: -Dit document is vertaald met behulp van de AI vertaaldienst [Co-op Translator](https://github.com/Azure/co-op-translator). Hoewel we streven naar nauwkeurigheid, dient u er rekening mee te houden dat automatische vertalingen fouten of onnauwkeurigheden kunnen bevatten. Het originele document in de oorspronkelijke taal moet als de gezaghebbende bron worden beschouwd. Voor kritieke informatie wordt professionele menselijke vertaling aanbevolen. Wij zijn niet aansprakelijk voor misverstanden of verkeerde interpretaties die voortvloeien uit het gebruik van deze vertaling. +**Disclaimer**: +Dit document is vertaald met behulp van de AI vertaaldienst [Co-op Translator](https://github.com/Azure/co-op-translator). Hoewel we streven naar nauwkeurigheid, dient u er rekening mee te houden dat geautomatiseerde vertalingen fouten of onnauwkeurigheden kunnen bevatten. Het originele document in de oorspronkelijke taal moet worden beschouwd als de gezaghebbende bron. Voor kritieke informatie wordt professionele menselijke vertaling aanbevolen. Wij zijn niet aansprakelijk voor eventuele misverstanden of verkeerde interpretaties die voortvloeien uit het gebruik van deze vertaling. \ No newline at end of file diff --git a/translations/nl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/nl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..575006737 --- /dev/null +++ b/translations/nl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybride RAG met Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Auteur:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Focus:** Kleine Taalmodellen (SLM's), SQLite FTS5 BM25, Dense Embeddings, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivatie: Het Probleem van Keyword Recall in Lokale SLM's\n", + "Standaard RAG-architecturen die puur op dense vector embeddings vertrouwen, slagen er vaak niet in om exacte numerieke tokens op te halen (bijv. `2.340.000 TL`, contractcodes, rekeningnummers). \n", + "Daarentegen mist sparse lexicaal zoeken (BM25) semantische synoniemen en geparafraseerde vragen.\n", + "\n", + "Deze handleiding laat zien hoe je een **hoge-snelheid, in-memory Hybride Retrieval-engine** implementeert die combineert:\n", + "1. **Dense Vectoren** (Cosinus Similariteit)\n", + "2. **Sparse Lexicaal Zoeken** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Gegronde Citaatgeneratie (`[1]`, `[2]`)** met Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Dubbel SQLite-schema (Dense Vectors + Virtuele FTS5 BM25-tabel)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Voorbeeld Inname & Uitvoeringsbenchmark\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Gefundeerde Promptformulering voor Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Disclaimer**:\nDit document is vertaald met behulp van de AI vertaaldienst [Co-op Translator](https://github.com/Azure/co-op-translator). Hoewel we streven naar nauwkeurigheid, dient u er rekening mee te houden dat geautomatiseerde vertalingen fouten of onnauwkeurigheden kunnen bevatten. Het originele document in de oorspronkelijke taal moet worden beschouwd als de gezaghebbende bron. Voor kritieke informatie wordt professionele menselijke vertaling aanbevolen. Wij zijn niet aansprakelijk voor eventuele misverstanden of verkeerde interpretaties die voortvloeien uit het gebruik van deze vertaling.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/no/.co-op-translator.json b/translations/no/.co-op-translator.json index f5cd5fcb4..460b0be59 100644 --- a/translations/no/.co-op-translator.json +++ b/translations/no/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "no" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:06:58+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:15:59+00:00", "source_file": "README.md", "language_code": "no" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "no" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:19:22+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "no" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:15+00:00", diff --git a/translations/no/README.md b/translations/no/README.md index 77f6951ee..80d1270ff 100644 --- a/translations/no/README.md +++ b/translations/no/README.md @@ -3,40 +3,40 @@ [![Åpne og bruk eksemplene i GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Åpne i Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub-bidragsytere](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub-problemer](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub bidragsytere](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub issues](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub pull-requests](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Velkommen](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub watchers](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub overvåkere](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub stjerner](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi er en serie med åpne AI-modeller utviklet av Microsoft. +Phi er en serie med åpen kildekode AI-modeller utviklet av Microsoft. -Phi er for øyeblikket den mest kraftfulle og kostnadseffektive lille språkmodellen (SLM), med svært gode benchmarks innen flerspråklighet, resonnering, tekst-/chatgenerering, koding, bilder, lyd og andre scenarier. +Phi er for øyeblikket den mest kraftfulle og kostnadseffektive lille språkmodellen (SLM), med svært gode resultater i flerspråklige, resonnement, tekst/chat-generering, koding, bilder, lyd og andre scenarier. -Du kan distribuere Phi i skyen eller til edge-enheter, og du kan enkelt bygge generative AI-applikasjoner med begrenset datakraft. +Du kan distribuere Phi til skyen eller til kant-enheter, og du kan enkelt bygge generative AI-applikasjoner med begrenset datakraft. Følg disse stegene for å komme i gang med å bruke disse ressursene: -1. **Fork depotet**: Klikk [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Klon depotet**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Bli med i Microsoft AI Discord-community og møt eksperter og andre utviklere**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +1. **Fork repositoriet**: Klikk [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Klon repositoriet**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Bli med i Microsoft AI Discord Community og møt eksperter og andre utviklere**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/no/cover.eb18d1b9605d754b.webp) ### 🌐 Flerspråklig støtte -#### Støttet via GitHub Action (Automatisert & alltid oppdatert) +#### Støttet via GitHub Action (Automatisk & Alltid oppdatert) -[Arabisk](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarsk](../bg/README.md) | [Burmesisk (Myanmar)](../my/README.md) | [Kinesisk (forenklet)](../zh-CN/README.md) | [Kinesisk (tradisjonell, Hong Kong)](../zh-HK/README.md) | [Kinesisk (tradisjonell, Macau)](../zh-MO/README.md) | [Kinesisk (tradisjonell, Taiwan)](../zh-TW/README.md) | [Kroatisk](../hr/README.md) | [Tsjekkisk](../cs/README.md) | [Dansk](../da/README.md) | [Nederlandsk](../nl/README.md) | [Estisk](../et/README.md) | [Finsk](../fi/README.md) | [Fransk](../fr/README.md) | [Tysk](../de/README.md) | [Gresk](../el/README.md) | [Hebraisk](../he/README.md) | [Hindi](../hi/README.md) | [Ungarsk](../hu/README.md) | [Indonesisk](../id/README.md) | [Italiensk](../it/README.md) | [Japansk](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreansk](../ko/README.md) | [Litauisk](../lt/README.md) | [Malayisk](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigeriansk pidgin](../pcm/README.md) | [Norsk](./README.md) | [Persisk (Farsi)](../fa/README.md) | [Polsk](../pl/README.md) | [Portugisisk (Brasil)](../pt-BR/README.md) | [Portugisisk (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumensk](../ro/README.md) | [Russisk](../ru/README.md) | [Serbisk (kyrillisk)](../sr/README.md) | [Slovakisk](../sk/README.md) | [Slovensk](../sl/README.md) | [Spansk](../es/README.md) | [Swahili](../sw/README.md) | [Svensk](../sv/README.md) | [Tagalog (Filippinsk)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Tyrkisk](../tr/README.md) | [Ukrainsk](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamesisk](../vi/README.md) +[Arabisk](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarsk](../bg/README.md) | [Burmesisk (Myanmar)](../my/README.md) | [Kinesisk (Forenklet)](../zh-CN/README.md) | [Kinesisk (Tradisjonell, Hong Kong)](../zh-HK/README.md) | [Kinesisk (Tradisjonell, Macau)](../zh-MO/README.md) | [Kinesisk (Tradisjonell, Taiwan)](../zh-TW/README.md) | [Kroatisk](../hr/README.md) | [Tsjekkisk](../cs/README.md) | [Dansk](../da/README.md) | [Nederlandsk](../nl/README.md) | [Estisk](../et/README.md) | [Finsk](../fi/README.md) | [Fransk](../fr/README.md) | [Tysk](../de/README.md) | [Gresk](../el/README.md) | [Hebraisk](../he/README.md) | [Hindi](../hi/README.md) | [Ungarsk](../hu/README.md) | [Indonesisk](../id/README.md) | [Italiensk](../it/README.md) | [Japansk](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreansk](../ko/README.md) | [Litauisk](../lt/README.md) | [Malaysisk](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalesisk](../ne/README.md) | [Nigeriansk Pidgin](../pcm/README.md) | [Norsk](./README.md) | [Persisk (Farsi)](../fa/README.md) | [Polsk](../pl/README.md) | [Portugisisk (Brasil)](../pt-BR/README.md) | [Portugisisk (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumensk](../ro/README.md) | [Russisk](../ru/README.md) | [Serbisk (kyrillisk)](../sr/README.md) | [Slovakisk](../sk/README.md) | [Slovensk](../sl/README.md) | [Spansk](../es/README.md) | [Swahili](../sw/README.md) | [Svensk](../sv/README.md) | [Tagalog (Filippinsk)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Tyrkisk](../tr/README.md) | [Ukrainsk](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamesisk](../vi/README.md) -> **Foretrekker du å klone lokalt?** +> **Foretrekker å klone lokalt?** > -> Dette depotet inkluderer 50+ språkoversettelser som betydelig øker nedlastingsstørrelsen. For å klone uten oversettelser, bruk sparse checkout: +> Dette repositoriet inkluderer 50+ språköversettelser som betydelig øker nedlastingsstørrelsen. For å klone uten oversettelser, bruk sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,7 +52,7 @@ Følg disse stegene for å komme i gang med å bruke disse ressursene: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Dette gir deg alt du trenger for å fullføre kurset med mye raskere nedlasting. +> Dette gir deg alt du trenger for å fullføre kurset med en mye raskere nedlasting. ## Innholdsfortegnelse @@ -60,16 +60,16 @@ Følg disse stegene for å komme i gang med å bruke disse ressursene: - Introduksjon - [Velkommen til Phi-familien](./md/01.Introduction/01/01.PhiFamily.md) - [Sette opp miljøet ditt](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Forstå viktige teknologier](./md/01.Introduction/01/01.Understandingtech.md) + - [Forståelse av nøkkelteknologier](./md/01.Introduction/01/01.Understandingtech.md) - [AI-sikkerhet for Phi-modeller](./md/01.Introduction/01/01.AISafety.md) - [Phi maskinvarestøtte](./md/01.Introduction/01/01.Hardwaresupport.md) - [Phi-modeller og tilgjengelighet på tvers av plattformer](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Bruke Guidance-ai og Phi](./md/01.Introduction/01/01.Guidance.md) + - [Bruk av Guidance-ai og Phi](./md/01.Introduction/01/01.Guidance.md) - [GitHub Marketplace-modeller](https://github.com/marketplace/models) - - [Azure AI modellkatalog](https://ai.azure.com) + - [Azure AI-modellkatalog](https://ai.azure.com) -- Inferens Phi i forskjellige miljøer - - [Hugging face](./md/01.Introduction/02/01.HF.md) +- Inference Phi i ulike miljøer + - [Hugging Face](./md/01.Introduction/02/01.HF.md) - [GitHub-modeller](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry modellkatalog](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) @@ -77,123 +77,125 @@ Følg disse stegene for å komme i gang med å bruke disse ressursene: - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Inferens Phi-familien - - [Inferens Phi på iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Inferens Phi på Android](./md/01.Introduction/03/Android_Inference.md) - - [Inferens Phi på Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Inferens Phi på AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Inferens Phi med Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Inferens Phi på lokal server](./md/01.Introduction/03/Local_Server_Inference.md) - - [Inferens Phi på ekstern server ved bruk av AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Inferens Phi med Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inferens Phi--Visjon lokalt](./md/01.Introduction/03/Vision_Inference.md) - - [Inferens Phi med Kaito AKS, Azure Containers (offisiell støtte)](./md/01.Introduction/03/Kaito_Inference.md) -- [Kvantifisere Phi-familien](./md/01.Introduction/04/QuantifyingPhi.md) - - [Kvantifisere Phi-3.5 / 4 ved bruk av llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Kvantifisere Phi-3.5 / 4 ved bruk av Generative AI-utvidelser for onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Kvantifisere Phi-3.5 / 4 ved bruk av Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Kvantifisere Phi-3.5 / 4 ved bruk av Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Evaluering Phi +- Inference Phi-familien + - [Inference Phi i iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Inference Phi i Android](./md/01.Introduction/03/Android_Inference.md) + - [Inference Phi i Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Inference Phi i AI-PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Inference Phi med Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Inference Phi i lokal server](./md/01.Introduction/03/Local_Server_Inference.md) + - [Inference Phi i ekstern server ved bruk av AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Inference Phi med Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Inference Phi--Vision lokalt](./md/01.Introduction/03/Vision_Inference.md) + - [Inference Phi med Kaito AKS, Azure Containers (offisiell støtte)](./md/01.Introduction/03/Kaito_Inference.md) +- [Kvantifisering av Phi-familien](./md/01.Introduction/04/QuantifyingPhi.md) + - [Kvantifisering av Phi-3.5 / 4 ved bruk av llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Kvantifisering av Phi-3.5 / 4 med Generative AI-utvidelser for onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Kvantifisering av Phi-3.5 / 4 ved bruk av Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Kvantifisering av Phi-3.5 / 4 ved bruk av Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- Evaluering av Phi - [Ansvarlig AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry for evaluering](./md/01.Introduction/05/AIFoundry.md) - - [Bruke Promptflow for evaluering](./md/01.Introduction/05/Promptflow.md) + - [Bruk av Promptflow for evaluering](./md/01.Introduction/05/Promptflow.md) - RAG med Azure AI Search - [Hvordan bruke Phi-4-mini og Phi-4-multimodal (RAG) med Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud lokal hybrid RAG med SQLite FTS5 og phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi applikasjonsutviklingseksempler - - Tekst- og chatteapplikasjoner - - Phi-4 eksempler +- Eksempler på Phi-applikasjonsutvikling + - Tekst- og chatte-applikasjoner + - Phi-4-eksempler - [📓] [Chat med Phi-4-mini ONNX-modell](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Chat med Phi-4 lokal ONNX-modell .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - [Chat .NET konsollapp med Phi-4 ONNX ved bruk av Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 eksempler - - [Lokal chatbot i nettleseren ved bruk av Phi3, ONNX Runtime Web og WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + + - Phi-3 / 3.5 Eksempler + - [Lokal Chatbot i nettleseren med Phi3, ONNX Runtime Web og WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - Interaktiv Phi-3-mini og OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [Multi Modell - Interaktiv Phi-3-mini og OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Lage en wrapper og bruke Phi-3 med MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Modelloptimalisering - Hvordan optimalisere Phi-3-minimodellen for ONNX Runtime Web med Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3-app med Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Multi Model AI-drevet Notat-app Eksempel](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Finjustere og integrere egendefinerte Phi-3-modeller med Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Finjustere og integrere egendefinerte Phi-3-modeller med Prompt flow i Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Evaluer den finjusterte Phi-3 / Phi-3.5-modellen i Microsoft Foundry med fokus på Microsofts prinsipper for ansvarlig AI](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct språkprediksjonsprøve (kinesisk/engelsk)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Modelloptimalisering - Hvordan optimalisere Phi-3-mini modellen for ONNX Runtime Web med Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 App med Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 Multi Modell AI-Drevet Notat App Eksempel](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Finjustere og integrere egendefinerte Phi-3 modeller med Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Finjustere og integrere egendefinerte Phi-3 modeller med Prompt flow i Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Evaluer den finjusterte Phi-3 / Phi-3.5 modellen i Microsoft Foundry med fokus på Microsofts ansvarlige AI-prinsipper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct språkprediksjonseksempel (kinesisk/engelsk)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Bruke Windows GPU for å lage Prompt flow-løsning med Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Bruke Microsoft Phi-3.5 tflite til å lage Android-app](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET-eksempel som bruker lokal ONNX Phi-3-modell med Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Konsoll-chatt .NET-app med Semantic Kernel og Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Bruke Microsoft Phi-3.5 tflite for å lage Android-app](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Q&A .NET-eksempel med lokal ONNX Phi-3 modell med Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Console chat .NET app med Semantic Kernel og Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK kodebaserte eksempler - - Phi-4 Eksempler - - [📓] [Generer prosjektkode med Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Azure AI Inference SDK Kodebaserte Eksempler + - Phi-4 Eksempler + - [📓] [Generere prosjektkode med Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 Eksempler - - [Bygg din egen Visual Studio Code GitHub Copilot Chat med Microsoft Phi-3-familien](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Lag din egen Visual Studio Code Chat Copilot Agent med Phi-3.5 ved bruk av GitHub-modeller](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Lag din egen Visual Studio Code GitHub Copilot Chat med Microsoft Phi-3-familien](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Lag din egen Visual Studio Code Chat Copilot Agent med Phi-3.5 fra GitHub-modeller](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Avanserte resonneringseksempler - - Phi-4 Eksempler - - [📓] [Phi-4-mini-resonnering eller Phi-4-resonneringseksempler](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Finjustering av Phi-4-mini-resonnering med Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Finjustering av Phi-4-mini-resonnering med Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-resonnering med GitHub-modeller](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-resonnering med Microsoft Foundry-modeller](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Phi-4 Eksempler + - [📓] [Phi-4-mini-reasoning eller Phi-4-reasoning eksempler](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Finjustering av Phi-4-mini-reasoning med Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Finjustering av Phi-4-mini-reasoning med Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-reasoning med GitHub Modeller](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-reasoning med Microsoft Foundry Modeller](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demoer - [Phi-4-mini demoer hostet på Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demoer hostet på Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Visjons-eksempler - - Phi-4 Eksempler - - [📓] [Bruk Phi-4-multimodal til å lese bilder og generere kode](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [Phi-4-multimodal demoer hostet på Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Syns Eksempler + - Phi-4 Eksempler + - [📓] [Bruke Phi-4-multimodal for å lese bilder og generere kode](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 Eksempler - - [📓][Phi-3-vision-Bilde tekst til tekst](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP-embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-syn-Bilde tekst til tekst](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-syn-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-syn CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [DEMO: Phi-3 Resirkulering](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Visuell språkhjelper - med Phi3-Vision og OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision multi-ramme eller multi-bilde eksempel](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision Lokal ONNX-modell med Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menybasert Phi-3 Vision Lokal ONNX-modell med Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Reasoning-Vision eksempler - - Phi-4-Reasoning-Vision-15B - - [📓] [Bruke Phi-4-Reasoning-Vision-15B for å oppdage ulovlig kryssing av vei](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Bruke Phi-4-Reasoning-Vision-15B for matematikk](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Bruke Phi-4-Reasoning-Vision-15B for å oppdage brukergrensesnitt](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - Matematikk-eksempler - - Phi-4-Mini-Flash-Reasoning-Instruct Eksempler [Matematikk-demo med Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - - Lyd-eksempler - - Phi-4 Eksempler - - [📓] [Uttrekke lydtranskripsjoner med Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal lydeksempel](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal taleoversettelseseksempel](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET konsollapplikasjon som bruker Phi-4-multimodal lyd for å analysere en lydfil og generere transkripsjon](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [Phi-3-syn - Visuell språkassistent - med Phi3-Vision og OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Syn Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Syn OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Syn multi-frame eller multi-bilde eksempel](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Syn Lokal ONNX Modell med Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Menybasert Phi-3 Syn Lokal ONNX Modell med Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Resonnerings-Syn Eksempler + - Phi-4-Resonnerings-Syn-15B + - [📓] [Bruke Phi-4-Resonnerings-Syn-15B for å oppdage fotgjengerovertredelser](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Bruke Phi-4-Resonnerings-Syn-15B til matematikk](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Bruke Phi-4-Resonnerings-Syn-15B for å oppdage UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - Matematikk Eksempler + - Phi-4-Mini-Flash-Resonnerings-Instruct Eksempler [Matematikk Demo med Phi-4-Mini-Flash-Resonnerings-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + + - Lyd Eksempler + - Phi-4 Eksempler + - [📓] [Uttrekk av lydtranskripsjoner med Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal Lyd Eksempel](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal Taleoversettelse Eksempel](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET konsollapplikasjon med Phi-4-multimodal Audio for å analysere en lydfil og generere transkripsjon](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE Eksempler - Phi-3 / 3.5 Eksempler - - [📓] [Phi-3.5 Mixture of Experts Models (MoEs) Sosiale Medier-eksempel](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Bygge en Retrieval-Augmented Generation (RAG) Pipeline med NVIDIA NIM Phi-3 MOE, Azure AI Search og LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 Mixture of Experts Modeller (MoEs) Sosiale Medier Eksempel](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Lage en Retrieval-Augmented Generation (RAG) Pipeline med NVIDIA NIM Phi-3 MOE, Azure AI Search og LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Funksjonskall-eksempler + - Funksjonskall Eksempler - Phi-4 Eksempler 🆕 - - [📓] [Bruke funksjonskall med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Bruke funksjonskall til å lage multi-agenter med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Bruke funksjonskall med Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Bruke funksjonskall med ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Multimodal Blandings-eksempler + - [📓] [Bruke Funksjonskall med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Bruke Funksjonskall til å lage multi-agenter med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Bruke Funksjonskall med Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Bruke Funksjonskall med ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Multimodal Blandet Eksempler - Phi-4 Eksempler 🆕 - - [📓] [Bruke Phi-4-multimodal som teknologi-journalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [📓] [Bruke Phi-4-multimodal som teknologijournalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [.NET konsollapplikasjon som bruker Phi-4-multimodal for å analysere bilder](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Finjustering av Phi-eksempler +- Finjustering av Phi Eksempler - [Finjusteringsscenarier](./md/03.FineTuning/FineTuning_Scenarios.md) - [Finjustering vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Finjustering: La Phi-3 bli en bransjeekspert](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Finjustering La Phi-3 bli en bransjeekspert](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - [Finjustering av Phi-3 med AI Toolkit for VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [Finjustering av Phi-3 med Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - [Finjustering av Phi-3 med Lora](./md/03.FineTuning/FineTuning_Lora.md) @@ -202,23 +204,24 @@ Følg disse stegene for å komme i gang med å bruke disse ressursene: - [Finjustering av Phi-3 med Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - [Finjustering med Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Finjustering med Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - - [Finjustering av Phi-3-vision med Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Finjustering av Phi-3-syn med Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Finjustering av Phi-3 med Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - [Finjustering av Phi-3-vision (offisiell støtte)](./md/03.FineTuning/FineTuning_Vision.md) - [Finjustering av Phi-3 med Kaito AKS, Azure Containers (offisiell støtte)](./md/03.FineTuning/FineTuning_Kaito.md) - [Finjustering av Phi-3 og 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Praktisk lab - - [Utforske banebrytende modeller: LLMer, SLMer, lokal utvikling og mer](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Lås opp NLP-potensial: Finjustering med Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Utforske banebrytende modeller: LLMs, SLMs, lokal utvikling og mer](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Låse opp NLP-potensial: finjustering med Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - Akademiske forskningsartikler og publikasjoner - [Textbooks Are All You Need II: phi-1.5 teknisk rapport](https://arxiv.org/abs/2309.05463) - - [Phi-3 teknisk rapport: En svært kapabel språkmodell lokalt på telefonen din](https://arxiv.org/abs/2404.14219) + - [Phi-3 teknisk rapport: en svært kapabel språkmodell lokalt på telefonen din](https://arxiv.org/abs/2404.14219) - [Phi-4 teknisk rapport](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini teknisk rapport: Kompakte men kraftige multimodale språkmodeller via Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Optimalisering av små språkmodeller for in-vehicle funksjons-kalling](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Finjustering av PHI-3 for flervalgsspørsmål: Metodikk, resultater og utfordringer](https://arxiv.org/abs/2501.01588) + - [Phi-4-Mini teknisk rapport: Kompakte, men kraftige multimodale språkmodeller via Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Optimalisering av små språkmodeller for funksjonsanrop i kjøretøy](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Finjustering av PHI-3 for flervalgsoppgaver: Metodikk, resultater og utfordringer](https://arxiv.org/abs/2501.01588) - [Phi-4-reasoning teknisk rapport](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Phi-4-mini-reasoning teknisk rapport](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) @@ -226,14 +229,14 @@ Følg disse stegene for å komme i gang med å bruke disse ressursene: ### Phi på Microsoft Foundry -Du kan lære hvordan du bruker Microsoft Phi og hvordan bygge ende-til-ende-løsninger på dine forskjellige maskinvareenheter. For å oppleve Phi selv, start med å leke med modellene og tilpasse Phi for dine scenarier ved hjelp av [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Du kan lære mer i Komme i gang med [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Du kan lære hvordan du bruker Microsoft Phi og hvordan du bygger ende-til-ende-løsninger på de forskjellige maskinvarene dine. For å prøve Phi selv, start med å leke med modellene og tilpasse Phi for dine scenarioer ved å bruke [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Du kan lære mer i Komme i gang med [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) **Lekeplass** Hver modell har en dedikert lekeplass for å teste modellen [Azure AI Playground](https://aka.ms/try-phi3). ### Phi på GitHub-modeller -Du kan lære hvordan du bruker Microsoft Phi og hvordan bygge ende-til-ende-løsninger på dine forskjellige maskinvareenheter. For å oppleve Phi selv, start med å leke med modellen og tilpasse Phi for dine scenarier ved hjelp av [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Du kan lære mer i Komme i gang med [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Du kan lære hvordan du bruker Microsoft Phi og hvordan du bygger ende-til-ende-løsninger på de forskjellige maskinvarene dine. For å prøve Phi selv, start med å leke med modellen og tilpasse Phi for dine scenarioer ved å bruke [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Du kan lære mer i Komme i gang med [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) **Lekeplass** Hver modell har en dedikert [lekeplass for å teste modellen](/md/02.QuickStart/GitHubModel_QuickStart.md). @@ -243,11 +246,11 @@ Hver modell har en dedikert [lekeplass for å teste modellen](/md/02.QuickStart/ Du kan også finne modellen på [Hugging Face](https://huggingface.co/microsoft) **Lekeplass** -[Hugging Chat lekeplass](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) + [Hugging Chat lekeplass](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Andre kurs + ## 🎒 Andre kurs -Vårt team produserer andre kurs! Ta en titt på: +Vårt team produserer andre kurs! Sjekk ut: ### LangChain @@ -263,7 +266,7 @@ Vårt team produserer andre kurs! Ta en titt på: [![AI-agenter for nybegynnere](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Generativ AI-serie [![Generativ AI for nybegynnere](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Generativ AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) @@ -271,7 +274,7 @@ Vårt team produserer andre kurs! Ta en titt på: [![Generativ AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - + ### Kjerneopplæring [![ML for nybegynnere](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Data Science for nybegynnere](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) @@ -282,37 +285,39 @@ Vårt team produserer andre kurs! Ta en titt på: [![XR-utvikling for nybegynnere](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Copilot-serie -[![Copilot for AI koordinert programmering](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot for AI-parprogrammering](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## Ansvarlig AI -Microsoft er forpliktet til å hjelpe våre kunder med å bruke våre AI-produkter ansvarlig, dele våre erfaringer, og bygge tillitsbaserte partnerskap gjennom verktøy som Transparency Notes og Impact Assessments. Mange av disse ressursene finnes på [https://aka.ms/RAI](https://aka.ms/RAI). -Microsofts tilnærming til ansvarlig AI er forankret i våre AI-prinsipper om rettferdighet, pålitelighet og sikkerhet, personvern og sikkerhet, inkludering, åpenhet og ansvarlighet. +Microsoft er forpliktet til å hjelpe kundene våre med å bruke AI-produktene våre på en ansvarlig måte, dele våre erfaringer, og bygge tillitsbaserte partnerskap gjennom verktøy som Transparency Notes og Impact Assessments. Mange av disse ressursene finnes på [https://aka.ms/RAI](https://aka.ms/RAI). +Microsofts tilnærming til ansvarlig AI er basert på våre AI-prinsipper om rettferdighet, pålitelighet og sikkerhet, personvern og sikkerhet, inkludering, åpenhet og ansvarlighet. + +Store modeller for naturlig språk, bilde og tale – som de som brukes i dette eksemplet – kan potensielt oppføre seg på måter som er urettferdige, upålitelige eller støtende, og dermed forårsake skade. Vennligst se gjennom [Azure OpenAI-tjenestens Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) for å bli informert om risikoer og begrensninger. + -Store naturlige språk-, bilde- og tale-modeller – som de som brukes i dette eksempelet – kan potensielt opptre på måter som er urettferdige, upålitelige eller støtende, og dermed forårsake skade. Vennligst konsulter [Azure OpenAI-tjenestens Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) for å bli informert om risiko og begrensninger. -Den anbefalte tilnærmingen for å redusere disse risikoene er å inkludere et sikkerhetssystem i arkitekturen din som kan oppdage og forhindre skadefull oppførsel. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) gir et uavhengig beskyttelseslag, i stand til å oppdage skadelig innhold generert av brukere og AI i applikasjoner og tjenester. Azure AI Content Safety inkluderer tekst- og bilde-APIer som lar deg oppdage materiell som er skadelig. Innen Microsoft Foundry lar Content Safety-tjenesten deg se, utforske og prøve ut eksempel-kode for å oppdage skadelig innhold på tvers av forskjellige modaliteter. Følgende [kom-i-gang-dokumentasjon](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) veileder deg gjennom det å sende forespørsler til tjenesten. +Den anbefalte tilnærmingen for å redusere disse risikoene er å inkludere et sikkerhetssystem i arkitekturen din som kan oppdage og forhindre skadelig oppførsel. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) tilbyr et uavhengig beskyttelseslag, som kan oppdage skadelig innhold generert av brukere og AI i applikasjoner og tjenester. Azure AI Content Safety inkluderer tekst- og bilde-APIer som lar deg oppdage materiale som er skadelig. Innen Microsoft Foundry gir Content Safety-tjenesten deg mulighet til å se, utforske og prøve ut eksempelkode for å oppdage skadelig innhold på tvers av ulike modaliteter. Følgende [raskstartdokumentasjon](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) veileder deg gjennom hvordan du sender forespørsler til tjenesten. -Et annet aspekt å ta i betraktning er den overordnede applikasjonsytelsen. Med multimodale og multimodells applikasjoner, anser vi ytelse som at systemet presterer slik du og brukerne dine forventer, inkludert at det ikke genererer skadelige utdata. Det er viktig å vurdere ytelsen til hele applikasjonen din ved å bruke [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Du har også muligheten til å lage og evaluere med [egendefinerte evaluatorer](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +En annen aspekt å ta i betraktning er den generelle ytelsen til applikasjonen. Med multimodale og multimodellsapplikasjoner mener vi at systemet presterer som du og dine brukere forventer, inkludert at det ikke genererer skadelige utdata. Det er viktig å vurdere ytelsen til hele applikasjonen din ved hjelp av [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Du har også muligheten til å lage og evaluere med [egendefinerte evalueringer](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Du kan evaluere AI-applikasjonen din i utviklingsmiljøet ved hjelp av [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Gitt enten et testdatasett eller et mål, blir generasjoner fra din generative AI-applikasjon kvantitativt målt med innebygde evaluatorer eller egendefinerte evaluatorer etter eget valg. For å komme i gang med Azure AI Evaluation SDK for å evaluere systemet ditt, kan du følge [kom-i-gang-guiden](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Når du har kjørt en evaluering, kan du [visualisere resultatene i Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Du kan evaluere AI-applikasjonen din i utviklingsmiljøet ved hjelp av [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Gitt enten et testdatasett eller et mål, blir genereringene fra den generative AI-applikasjonen din kvantitativt målt med innebygde evalueringer eller egendefinerte evalueringer etter eget valg. For å komme i gang med azure ai evaluation sdk for å evaluere systemet ditt, kan du følge [raskstartguiden](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Når du har kjørt en evaluering, kan du [visualisere resultatene i Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Varemerker -Dette prosjektet kan inneholde varemerker eller logoer for prosjekter, produkter eller tjenester. Autorisert bruk av Microsofts varemerker eller logoer er underlagt og må følge [Microsofts retningslinjer for varemerker og merkevare](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Bruk av Microsoft-varemerker eller logoer i modifiserte versjoner av dette prosjektet må ikke skape forvirring eller gi inntrykk av Microsoft-sponsing. Enhver bruk av tredjeparts varemerker eller logoer er underlagt de respektive tredjepartsretningslinjene. +Dette prosjektet kan inneholde varemerker eller logoer for prosjekter, produkter eller tjenester. Autorisert bruk av Microsofts varemerker eller logoer er underlagt og må følge [Microsofts varemerke- og merkevare-retningslinjer](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Bruk av Microsofts varemerker eller logoer i modifiserte versjoner av dette prosjektet må ikke skape forvirring eller antyde at Microsoft sponser. Enhver bruk av tredjeparts varemerker eller logoer er underlagt de tredjepartsenes retningslinjer. ## Få hjelp -Hvis du står fast eller har spørsmål om å bygge AI-apper, bli med i: +Hvis du står fast eller har spørsmål om å bygge AI-apper, bli med på: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Hvis du har produktfeedback eller feil under utvikling, besøk: +Har du produktfeedback eller feil under bygging, besøk: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) @@ -320,5 +325,5 @@ Hvis du har produktfeedback eller feil under utvikling, besøk: **Ansvarsfraskrivelse**: -Dette dokumentet har blitt oversatt ved hjelp av AI-oversettingstjenesten [Co-op Translator](https://github.com/Azure/co-op-translator). Selv om vi streber etter nøyaktighet, vær oppmerksom på at automatiserte oversettelser kan inneholde feil eller unøyaktigheter. Det opprinnelige dokumentet på sitt opprinnelige språk bør betraktes som den autoritative kilden. For kritisk informasjon anbefales profesjonell menneskelig oversettelse. Vi er ikke ansvarlige for misforståelser eller feiltolkninger som oppstår fra bruken av denne oversettelsen. +Dette dokumentet er oversatt ved hjelp av AI-oversettelsestjenesten [Co-op Translator](https://github.com/Azure/co-op-translator). Selv om vi streber etter nøyaktighet, vær oppmerksom på at automatiske oversettelser kan inneholde feil eller unøyaktigheter. Det opprinnelige dokumentet på originalspråket skal betraktes som den autoritative kilden. For kritisk informasjon anbefales profesjonell menneskelig oversettelse. Vi er ikke ansvarlige for eventuelle misforståelser eller feiltolkninger som oppstår ved bruk av denne oversettelsen. \ No newline at end of file diff --git a/translations/no/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/no/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..df3dd780c --- /dev/null +++ b/translations/no/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrid RAG med Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Forfatter:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Fokus:** Små språkmodeller (SLMs), SQLite FTS5 BM25, Tette innebygde representasjoner, Gjensidig rangfusjon (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivasjon: Problemet med nøkkelordgjenfinning i lokale SLM-er\n", + "Standard RAG-arkitekturer som utelukkende baserer seg på tette vektorinnbygginger klarer ofte ikke å hente opp eksakte numeriske tokens (f.eks. `2,340,000 TL`, kontraktkoder, kontonummer). \n", + "Motsatt savner sparsom leksikalsk søk (BM25) semantiske synonymer og omskrevne spørsmål.\n", + "\n", + "Denne oppskriftsboken viser hvordan man implementerer en **høyhastighets, minnebasert hybrid gjenfinningsmotor** som kombinerer:\n", + "1. **Tette vektorer** (kosinuslikhet)\n", + "2. **Sparsom leksikalsk søk** (SQLite FTS5 BM25)\n", + "3. **Gjensidig rangfusjon (RRF, $k=60$)**\n", + "4. **Forankret siteringsgenerering (`[1]`, `[2]`)** med Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Dobbel SQLite-skjema (Tette vektorer + Virtuell FTS5 BM25-tabell)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Prøveinntak og kjørebenchmark\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Grunnlaget promptformulering for Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Ansvarsfraskrivelse**:\nDette dokumentet er oversatt ved hjelp av AI-oversettelsestjenesten [Co-op Translator](https://github.com/Azure/co-op-translator). Selv om vi streber etter nøyaktighet, vær oppmerksom på at automatiske oversettelser kan inneholde feil eller unøyaktigheter. Det opprinnelige dokumentet på originalspråket skal betraktes som den autoritative kilden. For kritisk informasjon anbefales profesjonell menneskelig oversettelse. Vi er ikke ansvarlige for eventuelle misforståelser eller feiltolkninger som oppstår ved bruk av denne oversettelsen.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/pa/.co-op-translator.json b/translations/pa/.co-op-translator.json index b505bb2f1..94ea07362 100644 --- a/translations/pa/.co-op-translator.json +++ b/translations/pa/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "pa" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:40:12+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:12:38+00:00", "source_file": "README.md", "language_code": "pa" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "pa" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:18:34+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "pa" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:31+00:00", diff --git a/translations/pa/README.md b/translations/pa/README.md index c14ffe97d..9ffb62908 100644 --- a/translations/pa/README.md +++ b/translations/pa/README.md @@ -1,42 +1,42 @@ -# ਫਾਈ ਕੂਕਬੁੱਕ: ਮਾਇਕਰੋਸਾਫਟ ਦੇ ਫਾਈ ਮਾਡਲਾਂ ਨਾਲ ਹੱਥੋਂ-ਹੱਥ ਉਦਾਹਰਣਾਂ +# ਫਾਈ ਕੁਕਬੁੱਕ: ਮਾਇਕ੍ਰੋਸਾਫ਼ਟ ਦੇ ਫਾਈ ਮਾਡਲਾਂ ਨਾਲ ਹੱਥ-ਓਣ ਉਦਾਹਰਨਾਂ -[![GitHub ਕੋਡਸਪੇਸ ਵਿੱਚ ਉਦਾਹਰਣਾਂ ਖੋਲ੍ਹੋ ਅਤੇ ਵਰਤੋ](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![ਡੈਵ ਕੰਟੇਨਰਾਂ ਵਿੱਚ ਖੋਲ੍ਹੋ](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![ਗਿੱਟਹੱਬ ਕੋਡਸਪੇਸਿਜ਼ ਵਿੱਚ ਸੈਂਪਲ ਖੋਲ੍ਹੋ ਅਤੇ ਵਰਤੋਂ ਕਰੋ](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![ਡੈਵ ਕੰਟੇਨਰ ਵਿੱਚ ਖੋਲ੍ਹੋ](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub ਯੋਗਦਾਨਕਾਰ](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ਮੁੱਦੇ](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ਪੁੱਲ-ਰਿਕਵੈਸਟ](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![ਪੀਆਰਾਂ ਸਵਾਗਤਯੋਗ](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![ਗਿੱਟਹੱਬ ਯੋਗਦਾਨਕਾਰਕ](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![ਗਿੱਟਹੱਬ ਮਸਲੇ](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![ਗਿੱਟਹੱਬ ਪੂਲ-ਰਿਕਵੇਸਟ](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs ਸਵਾਗਤ ਹੈ](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ਵਾਚਰਜ਼](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ਫੋਰਕ](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ਸਟਾਰ](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![ਗਿੱਟਹੱਬ ਵਾਚਰਸ](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![ਗਿੱਟਹੱਬ ਫੋਰਕਸ](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![ਗਿੱਟਹੱਬ ਸਿਤਾਰੇ](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) -[![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) +[![ਮਾਇਕ੍ਰੋਸਾਫਟ ਫਾਊਂਡਰੀ ਡਿਸਕਾਰਡ](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -ਫਾਈ ਮਾਇਕਰੋਸਾਫਟ ਵੱਲੋਂ ਵਿਕਸਤ ਖੁੱਲ੍ਹਾ ਸਰੋਤ ਏਆਈ ਮਾਡਲਾਂ ਦੀ ਇੱਕ ਲੜੀ ਹੈ। +ਫਾਈ ਮਾਇਕ੍ਰੋਸਾਫਟ ਵੱਲੋਂ ਵਿਕਸਤ ਖੁੱਲ੍ਹਾ ਸਤਰ ਦਾ ਏਆਈ ਮਾਡਲਾਂ ਦਾ ਇੱਕ ਸਿਰੀਜ਼ ਹੈ। -ਫਾਈ ਇਸ ਸਮੇਂ ਸਭ ਤੋਂ ਸ਼ਕਤੀਸ਼ਾਲੀ ਅਤੇ ਖਰਚਾ-ਪ੍ਰਭਾਵੀ ਛੋਟਾ ਭਾਸ਼ਾਈ ਮਾਡਲ (ਐਸਐਲਐਮ) ਹੈ, ਜੋ ਬਹੁਭਾਸ਼ੀ, ਤਰਕ, ਪਾਠ/ਚੈਟ ਤਿਆਰ ਕਰਨ, ਕੋਡਿੰਗ, ਚਿੱਤਰ, ਆਡੀਓ ਅਤੇ ਹੋਰ ਸਥਿਤੀਆਂ ਵਿੱਚ ਬਹੁਤ ਵਧੀਆ ਬੈਂਚਮਾਰਕ ਪ੍ਰਦਰਸ਼ਿਤ ਕਰਦਾ ਹੈ। +ਫਾਈ ਵਰਤਮਾਨ ਵਿੱਚ ਸਭ ਤੋਂ ਤਾਕਤਵਰ ਅਤੇ ਕਿਫਾਇਤੀ ਛੋਟਾ ਭਾਸ਼ਾ ਮਾਡਲ (SLM) ਹੈ, ਜਿਸਦੇ ਬਹੁ-ਭਾਸ਼ਾਈ, ਤਰਕਸ਼ੀਲ, ਪਾਠ/ਚੈਟ ਜੈਨਰੇਸ਼ਨ, ਕੋਡਿੰਗ, ਚਿੱਤਰਾਂ, ਆਡੀਓ ਤੇ ਹੋਰ ਸਥਿਤੀਆਂ ਵਿੱਚ ਸੁਚੱਜੇ ਬੈਂਚਮਾਰਕ ਹਨ। -ਤੁਸੀਂ ਫਾਈ ਨੂੰ ਕਲਾਉਡ ਜਾਂ ਐਜ ਡਿਵਾਇਸਾਂ 'ਤੇ ਤੈਨਾਤ ਕਰ ਸਕਦੇ ਹੋ, ਅਤੇ ਸੀਮਤ ਕੰਪਿਊਟਿੰਗ ਸ਼ਕਤੀ ਨਾਲ ਆਸਾਨੀ ਨਾਲ ਜਨਰੇਟਿਵ ਏਆਈ ਐਪਲੀਕੇਸ਼ਨ ਬਣਾ ਸਕਦੇ ਹੋ। +ਤੁਸੀਂ ਫਾਈ ਨੂੰ ਕਲਾਉਡ ਜਾਂ ਐਜ ਡਿਵਾਈਸਾਂ 'ਤੇ ਡਿਪਲੌਇ ਕਰ ਸਕਦੇ ਹੋ, ਅਤੇ ਸੀਮਤ ਕੰਪਿਊਟਿੰਗ ਸ਼ਕਤੀ ਨਾਲ ਆਸਾਨੀ ਨਾਲ ਜਨਰੇਟਿਵ ਏਆਈ ਐਪਲੀਕੇਸ਼ਨਾਂ ਬਣਾ ਸਕਦੇ ਹੋ। -ਇਹ ਸਾਧਨਾਂ ਦੀ ਵਰਤੋਂ ਸ਼ੁਰੂ ਕਰਨ ਲਈ ਇਹ ਕਦਮ ਅਪਣਾਓ: -1. **ਰਿਪੋਜਿਟਰੀ ਨੂੰ ਫੋਰਕ ਕਰੋ**: ਕਲਿੱਕ ਕਰੋ [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **ਰਿਪੋਜਿਟਰੀ ਨੂੰ ਕਲੋਨ ਕਰੋ**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**ਮਾਇਕਰੋਸਾਫਟ ਏਆਈ ਡਿਸਕੋਰਡ ਸਮੁਦਾਇ ਵਿੱਚ ਸ਼ਾਮਲ ਹੋਵੋ ਅਤੇ ਮਾਹਿਰਾਂ ਤੇ ਹੋਰ ਵਿਕਾਸਕਾਰਾਂ ਨਾਲ ਮਿਲੋ**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +ਇਹਨਾਂ ਸਰੋਤਾਂ ਦੀ ਵਰਤੋਂ ਸ਼ੁਰੂ ਕਰਨ ਲਈ ਇਹ ਕਦਮ ਫਾਲੋ ਕਰੋ: +1. **ਰਿਪੋਜ਼ਿਟਰੀ ਨੂੰ ਫੋਰਕ ਕਰੋ**: ਕਲਿੱਕ ਕਰੋ [![ਗਿੱਟਹੱਬ ਫੋਰਕਸ](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **ਰਿਪੋਜ਼ਿਟਰੀ ਕਲੋਨ ਕਰੋ**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**ਮਾਇਕ੍ਰੋਸਾਫਟ ਏਆਈ ਡਿਸਕਾਰਡ ਕਮਿਊਨਿਟੀ ਨਾਲ ਜੁੜੋ ਅਤੇ ਮਾਹਿਰਾਂ ਅਤੇ ਹੋਰ ਡਿਵੈਲਪਰਾਂ ਨੂੰ ਮਿਲੋ**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/pa/cover.eb18d1b9605d754b.webp) +![ਕਵਰ](../../translated_images/pa/cover.eb18d1b9605d754b.webp) -### 🌐 ਬਹੁਭਾਸ਼ੀ ਸਹਾਇਤਾ +### 🌐 ਬਹੁ-ਭਾਸ਼ਾਈ ਸਹਾਇਤਾ -#### GitHub ਐਕਸ਼ਨ ਰਾਹੀਂ ਸਮਰਥਿਤ (ਸਵਚਾਲਿਤ ਅਤੇ ਸਦਾ ਅਪ-ਟੂ-ਡੇਟ) +#### ਗਿੱਟਹੱਬ ਐਕਸ਼ਨ ਰਾਹੀਂ ਸਪੋਰਟ ਕੀਤਾ ਗਿਆ (ਆਟੋਮੇਟੇਡ ਅਤੇ ਸਦਾ-ਅਪਡੇਟ) [Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](./README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **ਮੂਲ ਤੌਰ 'ਤੇ ਕਲੋਨ ਕਰਨਾ ਪਸੰਦ ਕਰੋ?** +> **ਕੀ ਤੁਸੀਂ ਸਥਾਨਕ ਤੌਰ 'ਤੇ ਕਲੋਨ ਕਰਨਾ ਪਸੰਦ ਕਰਦੇ ਹੋ?** > -> ਇਹ ਰਿਪੋਜਿਟਰੀ 50+ ਭਾਸ਼ਾਈ ਅਨੁਵਾਦ ਸ਼ਾਮਲ ਕਰਦਾ ਹੈ, ਜੋ ਡਾਊਨਲੋਡ ਸਾਈਜ਼ ਨੂੰ ਕਾਫੀ ਵਧਾ ਦਿੰਦਾ ਹੈ। ਬਿਨਾਂ ਅਨੁਵਾਦਾਂ ਦੇ ਕਲੋਨ ਕਰਨ ਲਈ sparse checkout ਵਰਤੋਂ ਕਰੋ: +> ਇਹ ਰਿਪੋਜ਼ਿਟਰੀ ਵਿੱਚ 50+ ਭਾਸ਼ਾ ਅਨੁਵਾਦ ਸ਼ਾਮਲ ਹਨ ਜੋ ਡਾਉਨਲੋਡ ਆਕਾਰ ਨੂੰ ਵੱਡਾ ਕਰ ਦਿੰਦے ਹਨ। ਬਿਨਾਂ ਅਨੁਵਾਦਾਂ ਦੇ ਕਲੋਨ ਕਰਨ ਲਈ, sparse checkout ਵਰਤੋਂ ਕਰੋ: > > **Bash / macOS / Linux:** > ```bash @@ -52,202 +52,205 @@ > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> ਇਹ ਤੁਹਾਨੂੰ ਕੋਰਸ ਨੂੰ ਮੂਲ ਤੌਰ 'ਤੇ ਤੇਜ਼ ਡਾਊਨਲੋਡ ਨਾਲ ਪੂਰਾ ਕਰਨ ਲਈ ਸਭ ਕੁਝ ਦਿੰਦਾ ਹੈ। +> ਇਹ ਤੁਹਾਨੂੰ ਕੋਰਸ ਨੂੰ ਪੂਰਾ ਕਰਨ ਲਈ ਸਾਰਾ ਕੁਝ ਮੁਹੱਈਆ ਕਰਵਾਉਂਦਾ ਹੈ ਤੇ ਡਾਉਨਲੋਡ ਤੇਜ਼ ਹੁੰਦਾ ਹੈ। -## ਸੂਚੀ ਸਿਰਲੇਖ +## ਸਮਗਰੀ ਸੂਚੀ - ਪਰਿਚਯ - - [ਫਾਈ ਪਰਿਵਾਰ ਵਿੱਚ ਸਵਾਗਤ ਹੈ](./md/01.Introduction/01/01.PhiFamily.md) - - [ਆਪਣਾ ਵਾਤਾਵਰਣ ਸੈੱਟ ਕਰਨਾ](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [ਮੁੱਖ ਤਕਨੀਕਾਂ ਦੀ ਸਮਝ](./md/01.Introduction/01/01.Understandingtech.md) + - [ਫਾਈ ਪਰਿਵਾਰ ਦਾ ਸਵਾਗਤ ਹੈ](./md/01.Introduction/01/01.PhiFamily.md) + - [ਆਪਣੇ ਵਾਤਾਵਰਨ ਨੂੰ ਸੈੱਟ ਕਰਨਾ](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [ਮੁੱਖ ਪ੍ਰੌਧੋਗਿਕੀਆਂ ਨੂੰ ਸਮਝਣਾ](./md/01.Introduction/01/01.Understandingtech.md) - [ਫਾਈ ਮਾਡਲਾਂ ਲਈ ਏਆਈ ਸੁਰੱਖਿਆ](./md/01.Introduction/01/01.AISafety.md) - [ਫਾਈ ਹਾਰਡਵੇਅਰ ਸਹਾਇਤਾ](./md/01.Introduction/01/01.Hardwaresupport.md) - - [ਫਾਈ ਮਾਡਲਾਂ ਅਤੇ ਪਲੇਟਫਾਰਮਾਂ 'ਤੇ ਉਪਲਬਧਤਾ](./md/01.Introduction/01/01.Edgeandcloud.md) - - [ਗਾਈਡੈਂਸ-ਏਆਈ ਅਤੇ ਫਾਈ ਦੀ ਵਰਤੋਂ](./md/01.Introduction/01/01.Guidance.md) - - [GitHub ਮਾਰਕੀਟਪਲੇਸ ਮਾਡਲ](https://github.com/marketplace/models) - - [ਆਜ਼ੁਰ ਏਆਈ ਮਾਡਲ ਕੈਟਾਲੌਗ](https://ai.azure.com) + - [ਫਾਈ ਮਾਡਲਾਂ ਅਤੇ ਪਲੇਟਫਾਰਮਾਂ ਉੱਤੇ ਉਪਲਬਧਤਾ](./md/01.Introduction/01/01.Edgeandcloud.md) + - [ਗਾਇਡੈਂਸ-ai ਅਤੇ ਫਾਈ ਦੀ ਵਰਤੋਂ](./md/01.Introduction/01/01.Guidance.md) + - [ਗਿੱਟਹੱਬ ਮਾਰਕੀਟਪਲੇਸ ਮਾਡਲ](https://github.com/marketplace/models) + - [ਅਜ਼ਯੂਰ ਏਆਈ ਮਾਡਲ ਕੈਟਲੌਗ](https://ai.azure.com) -- ਵੱਖ-ਵੱਖ ਵਾਤਾਵਰਣ ਵਿੱਚ ਫਾਈ ਇੰਫਰਨਸ +- ਵੱਖ-ਵੱਖ ਵਾਤਾਵਰਨ ਵਿੱਚ ਫਾਈ ਇਨਫਰੈਂਸ - [ਹੱਗਿੰਗ ਫੇਸ](./md/01.Introduction/02/01.HF.md) - - [GitHub ਮਾਡਲ](./md/01.Introduction/02/02.GitHubModel.md) - - [ਮਾਇਕਰੋਸਾਫਟ ਫਾਉਂਡਰੀ ਮਾਡਲ ਕੈਟਾਲੌਗ](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [ਗਿੱਟਹੱਬ ਮਾਡਲ](./md/01.Introduction/02/02.GitHubModel.md) + - [ਮਾਇਕ੍ਰੋਸਾਫਟ ਫਾਊਂਡਰੀ ਮਾਡਲ ਕੈਟਲੌਗ](./md/01.Introduction/02/03.AzureAIFoundry.md) - [ਓਲਾਮਾ](./md/01.Introduction/02/04.Ollama.md) - [ਏਆਈ ਟੂਲਕਿਟ VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - - [ਐਨਵੀਡੀਏ ਨਿੰਮ](./md/01.Introduction/02/06.NVIDIA.md) - - [ਫਾਉਂਡਰੀ ਲੋਕਲ](./md/01.Introduction/02/07.FoundryLocal.md) - -- ਫਾਈ ਪਰਿਵਾਰ ਦਾ ਇੰਫਰਨਸ - - [iOS ਵਿੱਚ ਫਾਈ ਇੰਫਰਨਸ](./md/01.Introduction/03/iOS_Inference.md) - - [ਐਂਡਰੌਇਡ ਵਿੱਚ ਫਾਈ ਇੰਫਰਨਸ](./md/01.Introduction/03/Android_Inference.md) - - [ਜੇਟਸਨ ਵਿੱਚ ਫਾਈ ਇੰਫਰਨਸ](./md/01.Introduction/03/Jetson_Inference.md) - - [ਏਆਈ ਪੀਸੀ ਵਿੱਚ ਫਾਈ ਇੰਫਰਨਸ](./md/01.Introduction/03/AIPC_Inference.md) - - [ਐਪਲ MLX ਫਰੇਮਵਰਕ ਨਾਲ ਫਾਈ ਇੰਫਰਨਸ](./md/01.Introduction/03/MLX_Inference.md) - - [ਲੋਕਲ ਸਰਵਰ ਵਿੱਚ ਫਾਈ ਇੰਫਰਨਸ](./md/01.Introduction/03/Local_Server_Inference.md) - - [ਏਆਈ ਟੂਲਕਿਟ ਵਰਤ ਕੇ ਰਿਮੋਟ ਸਰਵਰ ਵਿੱਚ ਫਾਈ ਇੰਫਰਨਸ](./md/01.Introduction/03/Remote_Interence.md) - - [ਰਸਟ ਨਾਲ ਫਾਈ ਇੰਫਰਨਸ](./md/01.Introduction/03/Rust_Inference.md) - - [ਲੋਕਲ ਵਿੱਚ ਫਾਈ-ਵਿਜ਼ਨ ਇੰਫਰਨਸ](./md/01.Introduction/03/Vision_Inference.md) - - [ਕੈਇਟੋ AKS, ਆਜ਼ੁਰ ਕੰਟੇਨਰਾਂ (ਅਧਿਕਾਰਤ ਸਹਾਇਤਾ) ਨਾਲ ਫਾਈ ਇੰਫਰਨਸ](./md/01.Introduction/03/Kaito_Inference.md) -- [ਫਾਈ ਪਰਿਵਾਰ ਦੀ ਮਾਤਰਾ](./md/01.Introduction/04/QuantifyingPhi.md) - - [ਲਾਮਾ.cpp ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਫਾਈ-3.5 / 4 ਨੂੰ ਕਵਾਂਟਾਈਜ਼ ਕਰਨਾ](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [ਓਨਐਕਸਰੰਟਾਈਮ ਲਈ ਜਨਰੇਟਿਵ ਏਆਈ ਵਿਸਥਾਰਨ ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਫਾਈ-3.5 / 4 ਨੂੰ ਕਵਾਂਟਾਈਜ਼ ਕਰਨਾ](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [ਇੰਟਲ OpenVINO ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਫਾਈ-3.5 / 4 ਨੂੰ ਕਵਾਂਟਾਈਜ਼ ਕਰਨਾ](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [ਐਪਲ MLX ਫਰੇਮਵਰਕ ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਫਾਈ-3.5 / 4 ਨੂੰ ਕਵਾਂਟਾਈਜ਼ ਕਰਨਾ](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- ਫਾਈ ਮੁਲਾਂਕਣ + - [ਐਨਵੀਆਨਵੀਡੀਆ NIM](./md/01.Introduction/02/06.NVIDIA.md) + - [ਫਾਊਂਡਰੀ ਲੋਕਲ](./md/01.Introduction/02/07.FoundryLocal.md) + +- ਫਾਈ ਪਰਿਵਾਰ ਦੀ ਇਨਫਰੈਂਸ + - [iOS ਵਿੱਚ ਫਾਈ ਇਨਫਰੈਂਸ](./md/01.Introduction/03/iOS_Inference.md) + - [ਐਂਡਰਾਇਡ ਵਿੱਚ ਫਾਈ ਇਨਫਰੈਂਸ](./md/01.Introduction/03/Android_Inference.md) + - [ਜੇਟਸਨ ਵਿੱਚ ਫਾਈ ਇਨਫਰੈਂਸ](./md/01.Introduction/03/Jetson_Inference.md) + - [ਏਆਈ ਪੀਸੀ ਵਿੱਚ ਫਾਈ ਇਨਫਰੈਂਸ](./md/01.Introduction/03/AIPC_Inference.md) + - [ਐਪਲ MLX ਫਰੇਮਵਰਕ ਨਾਲ ਫਾਈ ਇਨਫਰੈਂਸ](./md/01.Introduction/03/MLX_Inference.md) + - [ਲੋਕਲ ਸਰਵਰ ਵਿੱਚ ਫਾਈ ਇਨਫਰੈਂਸ](./md/01.Introduction/03/Local_Server_Inference.md) + - [ਏਆਈ ਟੂਲਕਿਟ ਨਾਲ ਰਿਮੋਟ ਸਰਵਰ ਵਿੱਚ ਫਾਈ ਇਨਫਰੈਂਸ](./md/01.Introduction/03/Remote_Interence.md) + - [ਰੱਸਟ ਨਾਲ ਫਾਈ ਇਨਫਰੈਂਸ](./md/01.Introduction/03/Rust_Inference.md) + - [ਲੋਕਲ ਵਿੱਚ ਫਾਈ--ਵਿਜ਼ਨ ਇਨਫਰੈਂਸ](./md/01.Introduction/03/Vision_Inference.md) + - [ਕੈਟੋ AKS, ਅਜ਼ਯੂਰ ਕੰਟੇਨਰਜ਼ (ਅਧਿਕਾਰਿਕ ਸਹਾਇਤਾ) ਨਾਲ ਫਾਈ ਇਨਫਰੈਂਸ](./md/01.Introduction/03/Kaito_Inference.md) +- [ਫਾਈ ਪਰਿਵਾਰ ਦਾ ਮਾਪ](./md/01.Introduction/04/QuantifyingPhi.md) + - [llama.cpp ਨਾਲ ਫਾਈ-3.5 / 4 ਕੋਅੰਟਾਈਜ਼ਿੰਗ](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntime ਲਈ ਜਨਰੇਟਿਵ ਏਆਈ ਐਕਸਟੈਂਸ਼ਨ ਨਾਲ ਫਾਈ-3.5 / 4 ਕੋਅੰਟਾਈਜ਼ਿੰਗ](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [ਇੰਟਲ ਓਪਨਵੀਨੋ ਨਾਲ ਫਾਈ-3.5 / 4 ਕੋਅੰਟਾਈਜ਼ਿੰਗ](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [ਐਪਲ MLX ਫਰੇਮਵਰਕ ਨਾਲ ਫਾਈ-3.5 / 4 ਕੋਅੰਟਾਈਜ਼ਿੰਗ](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- ਫਾਈ ਜ਼ਾਂਚ - [ਜ਼ਿੰਮੇਵਾਰ ਏਆਈ](./md/01.Introduction/05/ResponsibleAI.md) - - [ਮਾਇਕਰੋਸਾਫਟ ਫਾਉਂਡਰੀ ਮੁਲਾਂਕਣ ਲਈ](./md/01.Introduction/05/AIFoundry.md) - - [ਪ੍ਰਾਂਪਟਫਲੋ ਦੀ ਵਰਤੋਂ ਮੁਲਾਂਕਣ ਲਈ](./md/01.Introduction/05/Promptflow.md) + - [ਮਾਇਕ੍ਰੋਸਾਫਟ ਫਾਊਂਡਰੀ ਖਾਤਮੇ ਲਈ](./md/01.Introduction/05/AIFoundry.md) + - [ਖਾਤਮੇ ਲਈ ਪ੍ਰੰਪਟਫਲੋ ਦੀ ਵਰਤੋਂ](./md/01.Introduction/05/Promptflow.md) -- ਆਜ਼ੁਰ ਏਆਈ ਖੋਜ ਨਾਲ RAG - - [ਕਿਵੇਂ ਫਾਈ-4-ਮੀਨੀ ਅਤੇ ਫਾਈ-4-ਮਲਟੀਮੋਡਲ (RAG) ਨੂੰ ਆਜ਼ੁਰ ਏਆਈ ਖੋਜ ਨਾਲ ਵਰਤਣਾ ਹੈ](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- ਅਜ਼ਯੂਰ ਏਆਈ ਖੋਜ ਨਾਲ RAG + - [ਅਜ਼ਯੂਰ ਏਆਈ ਖੋਜ ਨਾਲ ਫਾਈ-4-ਮਿਨੀ ਅਤੇ ਫਾਈ-4-ਮਲਟੀਮੋਡਲ (RAG) ਦੀ ਵਰਤੋਂ ਕਿਵੇਂ ਕਰੀਏ](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [ਜ਼ੀਰੋ-ਕਲਾਉਡ ਲੋਕਲ ਹਾਈਬ੍ਰਿਡ RAG SQLite FTS5 ਅਤੇ phi-4-mini ਨਾਲ](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) + +- ਫਾਈ ਐਪਲੀਕੇਸ਼ਨ ਵਿਕਾਸ ਸੈਂਪਲ + - ਟੈਕਸਟ ਅਤੇ ਚੈਟ ਐਪਲੀਕੇਸ਼ਨ + - ਫਾਈ-4 ਸੈਂਪਲ + - [📓] [ਫਾਈ-4-ਮਿਨੀ ONNX ਮਾਡਲ ਨਾਲ ਚੈਟ](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [ਫਾਈ-4 ਲੋਕਲ ONNX ਮਾਡਲ .NET ਨਾਲ ਚੈਟ](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [ਸੇਂਮੈਂਟਿਕ ਕਰਨੇਲ ਦੀ ਵਰਤੋਂ ਨਾਲ ਫਾਈ-4 ONNX ਚੈਟ .NET ਕੰਸੋਲ ਐਪ](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) -- ਫਾਈ ਐਪਲੀਕੇਸ਼ਨ ਵਿਕਾਸ ਨਮੂਨੇ - - ਪਾਠ ਅਤੇ ਚੈਟ ਐਪਲੀਕੇਸ਼ਨ - - ਫਾਈ-4 ਨਮੂਨੇ - - [📓] [ਫਾਈ-4-ਮੀਨੀ ONNX ਮਾਡਲ ਨਾਲ ਚੈਟ ਕਰੋ](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [ਫਾਈ-4 ਲੋਕਲ ONNX ਮਾਡਲ ਨਾਲ .NET ਵਿੱਚ ਚੈਟ](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [ਸੇਮੰਟਿਕ ਕਰਨਲ ਦੀ ਵਰਤੋਂ ਨਾਲ ਫਾਈ-4 ONNX ਨਾਲ .NET ਕੰਸੋਲ ਐਪ ਚੈਟ](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - ਫਾਈ-3 / 3.5 ਨਮੂਨੇ - - [ਫਾਈ3, ONNX ਰੰਟਾਈਮ ਵੈੱਬ ਅਤੇ WebGPU ਵਰਤ ਕੇ ਬ੍ਰਾਊਜ਼ਰ ਵਿੱਚ ਲੋਕਲ ਚੈਟਬੋਟ](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [ਫਾਈ3, ONNX ਰਨਟਾਈਮ ਵੈੱਬ ਅਤੇ WebGPU ਦੀ ਵਰਤੋਂ ਕਰਦਿਆਂ ਬ੍ਰਾਉਜ਼ਰ ਵਿੱਚ ਸਥਾਨਕ ਚੈਟਬੋਟ](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino ਚੈਟ](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [ਮਲਟੀ ਮਾਡਲ - ਇੰਟਰਐਕਟਿਵ Phi-3-mini ਅਤੇ OpenAI ਵਿਸਪਰ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - ਵ੍ਰੈਪਰ ਬਣਾਉਣਾ ਅਤੇ Phi-3 ਨੂੰ MLFlow ਨਾਲ ਵਰਤਣਾ](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [ਮਾਡਲ ਆਪਟੀਮਾਈਜੇਸ਼ਨ - ONNX ਰਨਟਾਈਮ ਵੈੱਬ ਲਈ Phi-3-mini ਮਾਡਲ ਨੂੰ Olive ਨਾਲ ਕਿਵੇਂ ਆਪਟੀਮਾਈਜ਼ ਕਰਨਾ](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 ਐਪ ਨਾਲ Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 ਮਲਟੀ ਮਾਡਲ AI ਪਾਵਰਡ ਨੋਟਸ ਐਪ ਸੈਂਪਲ](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [ਫਾਈਨ-ਟਿਊਨ ਅਤੇ ਪ੍ਰੋਪਟ ਫਲੋ ਨਾਲ ਕਸਟਮ Phi-3 ਮਾਡਲ ਇਕਠੇ ਕਰਨਾ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [ਮਾਈਕ੍ਰੋਸਾਫਟ ਫਾਊਂਡਰੀ ਵਿੱਚ ਪ੍ਰੋਪਟ ਫਲੋ ਨਾਲ ਕਸਟਮ Phi-3 ਮਾਡਲ ਫਾਈਨ-ਟਿਊਨ ਅਤੇ ਇਕਠੇ ਕਰਨਾ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [ਮਾਈਕ੍ਰੋਸਾਫਟ ਦੇ ਜ਼ਿੰਮੇਵਾਰ AI ਸਿਧਾਂਤਾਂ ’ਤੇ ਧਿਆਨ ਕੇਂਦਰਿਤ ਕਰਦਿਆਂ ਮਾਈਕ੍ਰੋਸਾਫਟ ਫਾਊਂਡਰੀ ਵਿੱਚ ਫਾਈਨ-ਟਿਊਨ ਕੀਤੇ Phi-3 / Phi-3.5 ਮਾਡਲ ਦਾ ਮੁਲਾਂਕਣ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct ਭਾਸ਼ਾ ਅਨੁਮਾਨ ਸੈਂਪਲ (ਚੀਨੀ/ਅੰਗਰੇਜ਼ੀ)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG ਚੈਟਬੋਟ](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [ਵਿੰਡੋਜ਼ GPU ਦੀ ਵਰਤੋਂ ਕਰਕੇ Phi-3.5-Instruct ONNX ਨਾਲ ਪ੍ਰੋਪਟ ਫਲੋ ਸਮਾਧਾਨ ਬਣਾਉਣਾ](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [ਮਾਈਕ੍ਰੋਸਾਫਟ Phi-3.5 tflite ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਐਂਡਰਾਇਡ ਐਪ ਬਣਾਉਣਾ](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET ਉਦਾਹਰਨ ਜੋ ਸਥਾਨਕ ONNX Phi-3 ਮਾਡਲ ਅਤੇ Microsoft.ML.OnnxRuntime ਦੀ ਵਰਤੋਂ ਕਰਦੀ ਹੈ](../../md/04.HOL/dotnet/src/LabsPhi301) - - [ਕਨਸੋਲ ਚੈਟ .NET ਐਪ Semantic Kernel ਅਤੇ Phi-3 ਨਾਲ](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI ਇਨਫਰੈਂਸ SDK ਕੋਡ ਆਧਾਰਿਤ ਸੈਂਪਲ - - Phi-4 ਸੈਂਪਲ - - [📓] [Phi-4-multimodal ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਪ੍ਰੋਜੈਕਟ ਕੋਡ ਬਣਾਉਣਾ](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 ਸੈਂਪਲ - - [ਆਪਣਾ ਆਪਣਾ Visual Studio Code GitHub Copilot ਚੈਟ Microsoft Phi-3 ਪਰਿਵਾਰ ਨਾਲ ਬਣਾਓ](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub ਮਾਡਲ ਨਾਲ Phi-3.5 ਦੁਆਰਾ ਆਪਣਾ Visual Studio Code ਚੈਟ ਕੋਪਾਇਲਟ ਏਜेंट ਬਣਾਓ](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - ਉन्नਤ ਤਰਕ ਸੈਂਪਲ - - Phi-4 ਸੈਂਪਲ - - [📓] [Phi-4-mini-reasoning ਜਾਂ Phi-4-reasoning ਸੈਂਪਲ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Microsoft Olive ਨਾਲ Phi-4-mini-reasoning ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Apple MLX ਨਾਲ Phi-4-mini-reasoning ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [GitHub ਮਾਡਲ ਨਾਲ Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Microsoft Foundry ਮਾਡਲ ਨਾਲ Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - ਡੈਮੋਜ਼ - - [Phi-4-mini ਡੈਮੋਜ਼ ਜੋ Hugging Face Spaces ’ਤੇ ਹੋਸਟ ਕੀਤੇ ਗਏ ਹਨ](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal ਡੈਮੋਜ਼ ਜੋ Hugginge Face Spaces ’ਤੇ ਹੋਸਟ ਕੀਤੇ ਗਏ ਹਨ](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - ਦ੍ਰਿਸ਼ ਸੈਂਪਲ - - Phi-4 ਸੈਂਪਲ - - [📓] [Phi-4-multimodal ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਤਸਵੀਰਾਂ ਪੜ੍ਹੋ ਅਤੇ ਕੋਡ ਬਣਾਓ](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 ਸੈਂਪਲ - - [📓][Phi-3-vision-ਛਬੀ ਤੋਂ ਲਿਖਤ ਤਕ ਲਿਖਤ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP ਐਮਬੇਡਿੰਗ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [ਡੈਮੋ: Phi-3 ਰਿਸਾਇਕਲਿੰਗ](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - ਵਿਜ਼ੂਅਲ ਭਾਸ਼ਾ ਸਹਾਇਕ - Phi3-Vision ਅਤੇ OpenVINO ਨਾਲ](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 ਵਿਜ਼ਨ Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 ਵਿਜ਼ਨ OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 ਵਿਜ਼ਨ ਮਲਟੀ-ਫਰੇਮ ਜਾਂ ਮਲਟੀ-ਛਬੀ ਸੈਂਪਲ](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 ਵਿਜ਼ਨ ਸਥਾਨਕ ONNX ਮਾਡਲ Microsoft.ML.OnnxRuntime .NET ਦੀ ਵਰਤੋਂ ਕਰਦੇ ਹੋਏ](../../md/04.HOL/dotnet/src/LabsPhi303) - - [ਮੇਨੂ ਅਧਾਰਿਤ Phi-3 ਵਿਜ਼ਨ ਸਥਾਨਕ ONNX ਮਾਡਲ Microsoft.ML.OnnxRuntime .NET ਦੀ ਵਰਤੋਂ ਕਰਦੇ ਹੋਏ](../../md/04.HOL/dotnet/src/LabsPhi304) - - - ਤਰਕ-ਦ੍ਰਿਸ਼ ਸੈਂਪਲ - - Phi-4-ਤਰਕ-ਦ੍ਰਿਸ਼-15B - - [📓] [Phi-4-ਤਰਕ-ਦ੍ਰਿਸ਼-15B ਨੂੰ ਜੇਵਾਕਿੰਗ ਦਾ ਪਤਾ ਲਗਾਉਣ ਲਈ ਵਰਤਣਾ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Phi-4-ਤਰਕ-ਦ੍ਰਿਸ਼-15B ਨੂੰ ਗਣਿਤ ਲਈ ਵਰਤਣਾ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-ਤਰਕ-ਦ੍ਰਿਸ਼-15B ਨੂੰ UI ਦਾ ਪਤਾ ਲਗਾਉਣ ਲਈ ਵਰਤਣਾ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - ਗਣਿਤ ਸੈਂਪਲ - - Phi-4-ਮਿਨੀ-ਫਲੈਸ਼-ਤਰਕ-ਹਿਦਾਇਤ ਸੈਂਪਲ [Phi-4-ਮਿਨੀ-ਫਲੈਸ਼-ਤਰਕ-ਹਿਦਾਇਤ ਨਾਲ ਗਣਿਤ ਡੈਮੋ](./md/02.Application/09.Math/MathDemo.ipynb) - - - ਆਡੀਓ ਸੈਂਪਲ - - Phi-4 ਸੈਂਪਲ - - [📓] [Phi-4-multimodal ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਆਡੀਓ ਟ੍ਰਾਂਸਕ੍ਰਿਪਟ ਬਾਹਰ ਕੱਢਣਾ](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal ਆਡੀਓ ਸੈਂਪਲ](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal ਵਆਕ ਅਨੁਵਾਦ ਸੈਂਪਲ](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET ਕਨਸੋਲ ਐਪਲੀਕੇਸ਼ਨ ਜੋ Phi-4-multimodal ਆਡੀਓ ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਆਡੀਓ ਫਾਈਲ ਦਾ ਵਿਸ਼ਲੇਸ਼ਣ ਅਤੇ ਟ੍ਰਾਂਸਕ੍ਰਿਪਟ ਬਣਾਉਂਦੀ ਹੈ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - MOE ਸੈਂਪਲ - - Phi-3 / 3.5 ਸੈਂਪਲ - - [📓] [Phi-3.5 ਮਿਕਸਚਰ ਆਫ਼ ਐਕਸਪਰਟਸ ਮਾਡਲਜ਼ (MoEs) ਸੋਸ਼ਲ ਮੀਡੀਆ ਸੈਂਪਲ](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search ਅਤੇ LlamaIndex ਨਾਲ ਰਿਟਰੀਵਲ-ਅੱਗਮੈਂਟੈਡ ਜਨਰੇਸ਼ਨ (RAG) ਪਾਈਪਲਾਈਨ ਬਣਾਉਣਾ](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [ਮਲਟੀ ਮਾਡਲ - ਇੰਟਰਐਕਟਿਵ ਫਾਈ-3-ਮਿਨੀ ਅਤੇ OpenAI ਵਿਸਪਰ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - ਇੱਕ ਰੈਪਰ ਬਣਾਉਣਾ ਅਤੇ ਫਾਈ-3 ਨੂੰ MLFlow ਨਾਲ ਵਰਤਣਾ](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [ਮਾਡਲ ਆਪਟੀਮਾਈਜ਼ੇਸ਼ਨ - ONNX ਰਨਟਾਈਮ ਵੈੱਬ ਲਈ ਫਾਈ-3-ਮਿਨੀ ਮਾਡਲ ਨੂੰ Olive ਨਾਲ ਕਿਵੇਂ ਆਪਟੀਮਾਈਜ਼ ਕਰਨਾ](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 ਐਪ ਫਾਈ-3 ਮਿਨੀ-4k-instruct-onnx ਨਾਲ](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 ਮਲਟੀ ਮਾਡਲ ਏआई ਪਾਵਰਡ ਨੋਟਸ ਐਪ ਨਮੂਨਾ](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [ਕਸਟਮ ਫਾਈ-3 ਮਾਡਲਾਂ ਨੂੰ ਪ੍ਰੋੰਪਟ ਫਲੋ ਨਾਲ ਫਾਈਨ-ਟਿਊਨ ਅਤੇ ਇੰਟੀਗ੍ਰੇਟ ਕਰੋ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [ਮਾਈਕ੍ਰੋਸਾਫਟ Foundry ਵਿੱਚ ਪ੍ਰੋੰਪਟ ਫਲੋ ਨਾਲ ਕਸਟਮ ਫਾਈ-3 ਮਾਡਲਾਂ ਨੂੰ ਫਾਈਨ-ਟਿਊਨ ਅਤੇ ਇੰਟੀਗ੍ਰੇਟ ਕਰੋ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [ਮਾਈਕ੍ਰੋਸਾਫਟ ਦੀਜ਼ ਜ਼ਿੰਮੇਵਾਰ AI ਸਿਧਾਂਤਾਂ 'ਤੇ ਧਿਆਨ ਕੇਂਦਰਿਤ ਕਰਦਿਆਂ ਮਾਈਕ੍ਰੋਸਾਫਟ Foundry ਵਿੱਚ ਫਾਈਨ-ਟਿਊਨ ਕੀਤੇ ਗਏ ਫਾਈ-3 / ਫਾਈ-3.5 ਮਾਡਲ ਦਾ ਮੁੱਲਾਂਕਣ ਕਰੋ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [ਫਾਈ-3.5-ਮਿਨੀ-ਇੰਸਟ੍ਰੱਕਟ ਭਾਸ਼ਾ ਭਵਿੱਖਬਾਣੀ ਨਮੂਨਾ (ਚੀਨੀ/ਅੰਗਰੇਜ਼ੀ)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [ਫਾਈ-3.5-ਇੰਸਟ੍ਰੱਕਟ WebGPU RAG ਚੈਟਬੋਟ](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [ਫਾਈ-3.5-ਇੰਸਟ੍ਰੱਕਟ ONNX ਨਾਲ ਪ੍ਰੋੰਪਟ ਫਲੋ ਸਉਲਿਊਸ਼ਨ ਬਣਾਉਣ ਲਈ Windows GPU ਦੀ ਵਰਤੋਂ](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Android ਐਪ ਬਣਾਉਣ ਲਈ ਮਾਈਕ੍ਰੋਸਾਫਟ ਫਾਈ-3.5 tflite ਦੀ ਵਰਤੋਂ](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Microsoft.ML.OnnxRuntime ਦੀ ਵਰਤੋਂ ਕਰਦਿਆਂ ਸਥਾਨਕ ONNX ਫਾਈ-3 ਮਾਡਲ ਨਾਲ Q&A .NET ਉਦਾਹਰਨ](../../md/04.HOL/dotnet/src/LabsPhi301) + - [ਸੈਮੈਂਟਿਕ ਕਰਨਲ ਅਤੇ ਫਾਈ-3 ਨਾਲ ਕੰਸੋਲ ਚੈਟ .NET ਐਪ](../../md/04.HOL/dotnet/src/LabsPhi302) + + - ਅਜ਼ੁਰ ਏਆਈ ਇੰਫਰੈਂਸ SDK ਕੋਡ ਅਧਾਰਿਤ ਨਮੂਨੇ + - ਫਾਈ-4 ਨਮੂਨੇ + - [📓] [ਫਾਈ-4-ਮਲਟੀਮੋਡਲ ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਪ੍ਰੋਜੈਕਟ ਕੋਡ ਜਨਰੇਟ ਕਰੋ](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - ਫਾਈ-3 / 3.5 ਨਮੂਨੇ + - [ਮਾਈਕ੍ਰੋਸਾਫਟ ਫਾਈ-3 ਪਰਿਵਾਰ ਨਾਲ ਆਪਣਾ ਵਿਜ਼ੂਅਲ ਸਟੂਡੀਓ ਕੋਡ GitHub ਕੋਪਾਈਲਟ ਚੈਟ ਬਣਾਓ](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub ਮਾਡਲਾਂ ਦੁਆਰਾ ਫਾਈ-3.5 ਨਾਲ ਆਪਣਾ ਵਿਜ਼ੂਅਲ ਸਟੂਡੀਓ ਕੋਡ ਚੈਟ ਕੋਪਾਈਲਟ ਏਜੰਟ ਬਣਾਓ](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - ਉੱਚ ਤਰ੍ਹਾਂ ਦਾ ਤਰਕਸ਼ਾਸਤਰ ਨਮੂਨੇ + - ਫਾਈ-4 ਨਮੂਨੇ + - [📓] [ਫਾਈ-4-ਮਿਨੀ-ਤਕਰਾਰ ਜਾਂ ਫਾਈ-4-ਤਕਰਾਰ ਨਮੂਨੇ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [ਮਾਈਕ੍ਰੋਸਾਫਟ Olive ਨਾਲ ਫਾਈ-4-ਮਿਨੀ-ਤਕਰਾਰ ਨੂੰ ਫਾਈਨ-ਟਿਊਨ ਕਰਨਾ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLX ਨਾਲ ਫਾਈ-4-ਮਿਨੀ-ਤਕਰਾਰ ਨੂੰ ਫਾਈਨ-ਟਿਊਨ ਕਰਨਾ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [GitHub ਮਾਡਲਾਂ ਨਾਲ ਫਾਈ-4-ਮਿਨੀ-ਤਕਰਾਰ](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [ਮਾਈਕ੍ਰੋਸਾਫਟ Foundry ਮਾਡਲਾਂ ਨਾਲ ਫਾਈ-4-ਮਿਨੀ-ਤਕਰਾਰ](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - ਡੈਮੋ + - [ਫਾਈ-4-ਮਿਨੀ ਡੈਮੋਜ਼ Hugging Face ਸਪੇਸز ਤੇ ਹੋਸਟ ਕੀਤੇ ਗਏ](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [ਫਾਈ-4-ਮਲਟੀਮੋਡਲ ਡੈਮੋਜ਼ Hugginge Face ਸਪੇਸਜ਼ ਤੇ ਹੋਸਟ ਕੀਤੇ ਗਏ](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - ਵਿਜ਼ਨ ਨਮੂਨੇ + - ਫਾਈ-4 ਨਮੂਨੇ + - [📓] [ਫਾਈ-4-ਮਲਟੀਮੋਡਲ ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਚਿੱਤਰ ਪੜ੍ਹੋ ਅਤੇ ਕੋਡ ਜਨਰੇਟ ਕਰੋ](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - ਫਾਈ-3 / 3.5 ਨਮੂਨੇ + - [📓][ਫਾਈ-3-ਵਿਜ਼ਨ-ਚਿੱਤਰ ਲਿਖਤ ਤੋਂ ਲਿਖਤ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [ਫਾਈ-3-ਵਿਜ਼ਨ-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][ਫਾਈ-3-ਵਿਜ਼ਨ CLIP ਐਂਬੈਡਿੰਗ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [ਡੈਮੋ: ਫਾਈ-3 ਰੀਸਾਈਕਲਿੰਗ](https://github.com/jennifermarsman/PhiRecycling/) + - [ਫਾਈ-3-ਵਿਜ਼ਨ - ਦ੍ਰਿਸ਼ਟੀ ਭਾਸ਼ਾ ਸਹਾਇਕ - ਫਾਈ3-ਵਿਜ਼ਨ ਅਤੇ OpenVINO ਨਾਲ](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [ਫਾਈ-3 ਵਿਜ਼ਨ Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [ਫਾਈ-3 ਵਿਜ਼ਨ OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][ਫਾਈ-3.5 ਵਿਜ਼ਨ ਮਲਟੀ-ਫ੍ਰੇਮ ਜਾਂ ਮਲਟੀ-ਚਿੱਤਰ ਨਮੂਨਾ](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [ਮਾਈਕ੍ਰੋਸਾਫਟ.ML.OnnxRuntime .NET ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਫਾਈ-3 ਵਿਜ਼ਨ ਸਥਾਨਕ ONNX ਮਾਡਲ](../../md/04.HOL/dotnet/src/LabsPhi303) + - [ਮੇਨੂ ਅਧਾਰਿਤ ਫਾਈ-3 ਵਿਜ਼ਨ ਸਥਾਨਕ ONNX ਮਾਡਲ Microsoft.ML.OnnxRuntime .NET ਦੀ ਵਰਤੋਂ ਨਾਲ](../../md/04.HOL/dotnet/src/LabsPhi304) + + - ਤਰਕਸ਼ਾਸਤਰ-ਵਿਜ਼ਨ ਨਮੂਨੇ + - ਫਾਈ-4-ਤਰਕਸ਼ਾਸਤਰ-ਵਿਜ਼ਨ-15B + - [📓] [ਫਾਈ-4-ਤਰਕਸ਼ਾਸਤਰ-ਵਿਜ਼ਨ-15B ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਜੇਵਾਕਿੰਗ ਦੀ ਪਹਿਚਾਣ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [ਫਾਈ-4-ਤਰਕਸ਼ਾਸਤਰ-ਵਿਜ਼ਨ-15B ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਗਣਿਤ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [ਫਾਈ-4-ਤਰਕਸ਼ਾਸਤਰ-ਵਿਜ਼ਨ-15B ਦੀ ਵਰਤੋਂ ਕਰਕੇ UI ਦੀ ਪਹਿਚਾਣ](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - ਗਣਿਤ ਨਮੂਨੇ + - ਫਾਈ-4-ਮਿਨੀ-ਫਲੈਸ਼-ਤਰਕਸ਼ਾਸਤਰ-ਇਨਸਟ੍ਰੱਕਟ ਨਮੂਨੇ [ਫਾਈ-4-ਮਿਨੀ-ਫਲੈਸ਼-ਤਰਕਸ਼ਾਸਤਰ-ਇਨਸਟ੍ਰੱਕਟ ਨਾਲ ਗਣਿਤ ਡੈਮੋ](./md/02.Application/09.Math/MathDemo.ipynb) + + - ਆਡੀਓ ਨਮੂਨੇ + - ਫਾਈ-4 ਨਮੂਨੇ + - [📓] [ਫਾਈ-4-ਮਲਟੀਮੋਡਲ ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਆਡੀਓ ਟ੍ਰਾਂਸਕ੍ਰਿਪਟ ਨਿਕਾਲਣਾ](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [ਫਾਈ-4-ਮਲਟੀਮੋਡਲ ਆਡੀਓ ਨਮੂਨਾ](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [ਫਾਈ-4-ਮਲਟੀਮੋਡਲ ਸਪੀਚ ਅਨੁਵਾਦ ਨਮੂਨਾ](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET ਕੰਸੋਲ ਐਪਲੀਕੇਸ਼ਨ ਜੋ ਫਾਈ-4-ਮਲਟੀਮੋਡਲ ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਆਡੀਓ ਫਾਈਲ ਦਾ ਵਿਸ਼ਲੇਸ਼ਣ ਅਤੇ ਟ੍ਰਾਂਸਕ੍ਰਿਪਟ ਜਨਰੇਟ ਕਰਦਾ ਹੈ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE ਨਮੂਨੇ + - ਫਾਈ-3 / 3.5 ਨਮੂਨੇ + - [📓] [ਫਾਈ-3.5 ਮਿਕਸਚਰ ਆਫ਼ ਐਕਸਪਟਰਟ ਮਾਡਲਜ਼ (MoEs) ਸੋਸ਼ਲ ਮੀਡੀਆ ਨਮੂਨਾ](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM ਫਾਈ-3 MOE, ਅਜ਼ੁਰ ਏਆਈ ਸਹਾਇਕ ਅਤੇ LlamaIndex ਨਾਲ ਰੀਟਰੀਵਲ-ਆਗਮੈਂਟਡ ਜਨਰੇਸ਼ਨ (RAG) ਪਾਈਪਲਾਈਨ ਬਣਾਉਣਾ](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - ਫੰਕਸ਼ਨ ਕਾਲਿੰਗ ਸੈਂਪਲ - - Phi-4 ਸੈਂਪਲ 🆕 - - [📓] [Phi-4-mini ਨਾਲ ਫੰਕਸ਼ਨ ਕਾਲਿੰਗ ਦੀ ਵਰਤੋਂ](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Phi-4-mini ਨਾਲ ਮਲਟੀ-ਏਜੰਟ ਬਣਾਉਣ ਲਈ ਫੰਕਸ਼ਨ ਕਾਲਿੰਗ ਦੀ ਵਰਤੋਂ](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Ollama ਨਾਲ ਫੰਕਸ਼ਨ ਕਾਲਿੰਗ ਦੀ ਵਰਤੋਂ ਕਰਨਾ](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - ਫੰਕਸ਼ਨ ਕਾਲਿੰਗ ਨਮੂਨੇ + - ਫਾਈ-4 ਨਮੂਨੇ 🆕 + - [📓] [ਫਾਈ-4-ਮਿਨੀ ਨਾਲ ਫੰਕਸ਼ਨ ਕਾਲਿੰਗ ਦੀ ਵਰਤੋਂ](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [ਫਾਈ-4-ਮਿਨੀ ਨਾਲ ਕਈ ਏਜੰਟ ਬਣਾਉਣ ਲਈ ਫੰਕਸ਼ਨ ਕਾਲਿੰਗ ਦੀ ਵਰਤੋਂ](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Ollama ਨਾਲ ਫੰਕਸ਼ਨ ਕਾਲਿੰਗ ਦੀ ਵਰਤੋਂ](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [ONNX ਨਾਲ ਫੰਕਸ਼ਨ ਕਾਲਿੰਗ ਦੀ ਵਰਤੋਂ](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - ਮਲਟੀਮੋਡਲ ਮਿਕਸਿੰਗ ਸੈਂਪਲ - - Phi-4 ਸੈਂਪਲ 🆕 - - [📓] [ਤਕਨਾਲੋਜੀ ਪੱਤਰਕਾਰ ਵਜੋਂ Phi-4-multimodal ਦੀ ਵਰਤੋਂ](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET ਕਨਸੋਲ ਐਪਲੀਕੇਸ਼ਨ ਜੋ Phi-4-multimodal ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਛਬੀਆਂ ਦਾ ਵਿਸ਼ਲੇਸ਼ਣ ਕਰਦਾ ਹੈ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- ਫਾਈਨ-ਟਿਊਨਿੰਗ Phi ਸੈਂਪਲ - - [ਫਾਈਨ-ਟਿਊਨਿੰਗ ਸਿਨਾਰੀਓਜ਼](./md/03.FineTuning/FineTuning_Scenarios.md) - - [ਫਾਈਨ-ਟਿਊਨਿੰਗ ਬਨਾਮ RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [ਫਾਈਨ-ਟਿਊਨਿੰਗ Phi-3 ਨੂੰ ਉਦਯੋਗ ਵਿਸ਼ੇਸ਼ਜ्ञ ਬਣਾਉਣਾ](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [VS ਕੋਡ ਲਈ AI ਟੂਲਕਿੱਟ ਨਾਲ Phi-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Azure ਮਸ਼ੀਨ ਲਰਨਿੰਗ ਸੇਵਾ ਨਾਲ Phi-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/Introduce_AzureML.md) - - [ਲੋਰਾ ਨਾਲ Phi-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora ਨਾਲ Phi-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_Qlora.md) - - [Microsoft Foundry ਨਾਲ Phi-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Azure ML CLI/SDK ਨਾਲ Phi-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Microsoft Olive ਨਾਲ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive ਦੀ ਹੈਂਡਸ-ਆਨ ਲੈਬ ਨਾਲ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/olive-lab/readme.md) - - [Weights and Bias ਨਾਲ Phi-3-vision ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Apple MLX ਫਰੇਮਵਰਕ ਨਾਲ Phi-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-vision (ਅਧਿਕਾਰਕ ਸਮਰਥਨ) ਨਾਲ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_Vision.md) - - [Fine-Tuning Phi-3 with Kaito AKS , Azure Containers(official Support)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Fine-Tuning Phi-3 and 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - -- Hands on Lab - - [Exploring cutting-edge models: LLMs, SLMs, local development and more](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Unlocking NLP Potential: Fine-Tuning with Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - -- Academic Research Papers and Publications - - [Textbooks Are All You Need II: phi-1.5 technical report](https://arxiv.org/abs/2309.05463) - - [Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone](https://arxiv.org/abs/2404.14219) - - [Phi-4 Technical Report](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Optimizing Small Language Models for In-Vehicle Function-Calling](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Fine-Tuning PHI-3 for Multiple-Choice Question Answering: Methodology, Results, and Challenges](https://arxiv.org/abs/2501.01588) - - [Phi-4-reasoning Technical Report](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini-reasoning Technical Report](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) - -## Using Phi Models - -### Phi on Microsoft Foundry - -ਤੁਸੀਂ ਸਿੱਖ ਸਕਦੇ ਹੋ ਕਿ ਮਾਇਕਰੋਸੋਫਟ ਫਾਈ ਨੂੰ ਕਿਵੇਂ ਵਰਤਣਾ ਹੈ ਅਤੇ ਆਪਣੇ ਵੱਖ-ਵੱਖ ਹਾਰਡਵੇਅਰ ਡਿਵਾਇਸز ਵਿੱਚ E2E ਹੱਲ ਕਿਵੇਂ ਬਣਾਉਣੇ ਹਨ। ਆਪਣੇ ਲਈ ਫਾਈ ਦਾ ਅਨੁਭਵ ਕਰਨ ਲਈ, ਮਾਡਲਾਂ ਨਾਲ ਖੇਡਣ ਦੀ ਸ਼ੁਰੂਆਤ ਕਰੋ ਅਤੇ ਆਪਣੇ ਦ੍ਰਿਸ਼ਾਂ ਲਈ ਫਾਈ ਨੂੰ ਕਸਟਮਾਈਜ਼ ਕਰੋ ਵਰਤ ਕੇ [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai)। ਤੁਸੀਂ Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) 'ਤੇ ਹੋਰ ਸਿੱਖ ਸਕਦੇ ਹੋ। - -**Playground** -ਹਰ ਮਾਡਲ ਲਈ ਇੱਕ ਮਖ਼ਸੂਸ ਪਲੇਗ੍ਰਾਊਂਡ ਹੁੰਦਾ ਹੈ ਮਾਡਲ ਦੀ ਜਾਂਚ ਲਈ [Azure AI Playground](https://aka.ms/try-phi3)। - -### Phi on GitHub Models - -ਤੁਸੀਂ ਸਿੱਖ ਸਕਦੇ ਹੋ ਕਿ ਮਾਇਕਰੋਸੋਫਟ ਫਾਈ ਨੂੰ ਕਿਵੇਂ ਵਰਤਣਾ ਹੈ ਅਤੇ ਆਪਣੇ ਵੱਖ-ਵੱਖ ਹਾਰਡਵੇਅਰ ਡਿਵਾਇਸਜ਼ ਵਿੱਚ E2E ਹੱਲ ਕਿਵੇਂ ਬਣਾਉਣੇ ਹਨ। ਆਪਣੇ ਲਈ ਫਾਈ ਦਾ ਅਨੁਭਵ ਕਰਨ ਲਈ, ਮਾਡਲ ਨਾਲ ਖੇਡਣ ਅਤੇ ਆਪਣੇ ਦ੍ਰਿਸ਼ਾਂ ਲਈ ਫਾਈ ਨੂੰ ਕਸਟਮਾਈਜ਼ ਕਰਨ ਦੀ ਸ਼ੁਰੂਆਤ ਕਰੋ ਵਰਤ ਕੇ [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo)। ਤੁਸੀਂ Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) 'ਤੇ ਹੋਰ ਸਿੱਖ ਸਕਦੇ ਹੋ। - -**Playground** -ਹਰ ਮਾਡਲ ਲਈ ਇੱਕ ਮਖ਼ਸੂਸ [playground to test the model](/md/02.QuickStart/GitHubModel_QuickStart.md) ਹੁੰਦਾ ਹੈ। - -### Phi on Hugging Face + - ਮਲਟੀਮੋਡਲ ਮਿਕਸਿੰਗ ਨਮੂਨੇ + - ਫਾਈ-4 ਨਮੂਨੇ 🆕 + - [📓] [ਫਾਈ-4-ਮਲਟੀਮੋਡਲ ਨੂੰ ਇੱਕ ਤਕਨਾਲੋਜੀ ਪੱਤਰਕਾਰ ਵਜੋਂ ਵਰਤਣਾ](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET ਕੰਸੋਲ ਐਪਲੀਕੇਸ਼ਨ ਜੋ ਫਾਈ-4-ਮਲਟੀਮੋਡਲ ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਚਿੱਤਰਾਂ ਦਾ ਵਿਸ਼ਲੇਸ਼ਣ ਕਰਦਾ ਹੈ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- ਫਾਈਨ-ਟਿਊਨਿੰਗ ਫਾਈ ਨਮੂਨੇ + - [ਫਾਈਨ-ਟਿਊਨਿੰਗ ਸੀਨਾਰਿਓਜ਼](./md/03.FineTuning/FineTuning_Scenarios.md) + - [ਫਾਈਨ-ਟਿਊਨਿੰਗ ਵਿਰੁੱਧ RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [ਫਾਈਨ-ਟਿਊਨਿੰਗ ਫਾਈ-3 ਨੂੰ ਉਦਯੋਗ ਵਿਸ਼ੇਸ਼ਜ্ঞান ਬਣਾਉਣਾ](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [VS ਕੋਡ ਲਈ AI ਟੂਲਕਿੱਟ ਨਾਲ ਫਾਈ-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [ਅਜ਼ੁਰ ਮਸ਼ੀਨ ਲਰਨਿੰਗ ਸਰਵਿਸ ਨਾਲ ਫਾਈ-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/Introduce_AzureML.md) + - [ਲੋਰਾ ਨਾਲ ਫਾਈ-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_Lora.md) + - QLora ਨਾਲ ਫਾਈ-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_Qlora.md) + - [ਮਾਈਕ੍ਰੋਸਾਫਟ Foundry ਨਾਲ ਫਾਈ-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [ਅਜ਼ੁਰ ML CLI/SDK ਨਾਲ ਫਾਈ-3 ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_MLSDK.md) + - [ਮਾਈਕ੍ਰੋਸਾਫਟ Olive ਨਾਲ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [ਮਾਈਕ੍ਰੋਸਾਫਟ Olive ਹੈਂਡਜ਼-ਆਨ ਲੈਬ ਨਾਲ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/olive-lab/readme.md) + - [Weights and Bias ਨਾਲ ਫਾਈ-3-ਵਿਜ਼ਨ ਦੀ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Apple MLX Framework ਨਾਲ ਫਾਈ-3 ਦਾ ਫਾਈਨ-ਟਿਊਨਿੰਗ](./md/03.FineTuning/FineTuning_MLX.md) + - [ਫਾਈ-3-ਵਿਜ਼ਨ ਦਾ ਫਾਈਨ-ਟਿਊਨਿੰਗ (ਅਧਿਕਾਰਕ ਸਪੋਰਟ)](./md/03.FineTuning/FineTuning_Vision.md) + - [ਫਾਈ-3 ਦਾ ਕਾਈਟੋ AKS, Azure ਕਨਟੇਨਰਾਂ ਨਾਲ ਫਾਈਨ-ਟਿਊਨਿੰਗ (ਅਧਿਕਾਰਕ ਸਹਾਇਤਾ)](./md/03.FineTuning/FineTuning_Kaito.md) + - [ਫਾਈ-3 ਅਤੇ 3.5 ਵਿਜ਼ਨ ਦਾ ਫਾਈਨ-ਟਿਊਨਿੰਗ](https://github.com/2U1/Phi3-Vision-Finetune) + +- ਹੈਂਡਸ-ਆਨ ਲੈਬ + - [ਅਗੇਤੁਰ ਮਾਡਲ ਦੀ ਪੜਚੋਲ: LLMs, SLMs, ਸਥਾਨਕ ਵਿਕਾਸ ਅਤੇ ਹੋਰ](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP ਸਮਰੱਥਾ ਖੋਲ੍ਹਣਾ: Microsoft Olive ਨਾਲ ਫਾਈਨ-ਟਿਊਨਿੰਗ](https://github.com/azure/Ignite_FineTuning_workshop) + +- ਅਕਾਦਮਿਕ ਖੋਜ ਪੱਤਰ ਅਤੇ ਪ੍ਰਕਾਸ਼ਨ + - [Textbooks Are All You Need II: phi-1.5 ਤਕਨੀਕੀ ਰਿਪੋਰਟ](https://arxiv.org/abs/2309.05463) + - [Phi-3 ਤਕਨੀਕੀ ਰਿਪੋਰਟ: ਤੁਹਾਡੇ ਫੋਨ 'ਤੇ ਇੱਕ ਬਹੁਤ ਸਮਰੱਥ ਭਾਸ਼ਾ ਮਾਡਲ](https://arxiv.org/abs/2404.14219) + - [Phi-4 ਤਕਨੀਕੀ ਰਿਪੋਰਟ](https://arxiv.org/abs/2412.08905) + - [Phi-4-Mini ਤਕਨੀਕੀ ਰਿਪੋਰਟ: ਮਿਕਸਚਰ-ਆਫ-ਲੋRAs ਰਾਹੀਂ ਘਣਿਸ਼ਠ ਅਤੇ ਸ਼ਕਤੀਸ਼ਾਲੀ ਮਲਟੀਮੋਡਲ ਭਾਸ਼ਾ ਮਾਡਲ](https://arxiv.org/abs/2503.01743) + - [ਵਾਹਨ ਵਿੱਚ ਫੰਕਸ਼ਨ-ਕਾਲਿੰਗ ਲਈ ਛੋਟੇ ਭਾਸ਼ਾ ਮਾਡਲਾਂ ਦਾ ਆਪਟੀਮਾਈਜ਼ੇਸ਼ਨ](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) ਬਹੁ-ਚੋਣ ਸਵਾਲ ਜਵਾਬ ਲਈ PHI-3 ਦਾ ਫਾਈਨ-ਟਿਊਨਿੰਗ: ਢੰਗ, ਨਤੀਜੇ ਅਤੇ ਚੁਣੌਤੀਆਂ](https://arxiv.org/abs/2501.01588) + - [Phi-4-ਵੀਚਾਰਧਾਰਾ ਤਕਨੀਕੀ ਰਿਪੋਰਟ](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-ਵੀਚਾਰਧਾਰਾ ਤਕਨੀਕੀ ਰਿਪੋਰਟ](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + +## ਫਾਈ ਮਾਡਲਸ ਦੀ ਵਰਤੋਂ + +### Microsoft Foundry 'ਤੇ ਫਾਈ + +ਤੁਸੀਂ Microsoft Phi ਦੀ ਵਰਤੋਂ ਕਿਵੇਂ ਕਰਨ ਦੀ ਸਿੱਖਿਆ ਪ੍ਰਾਪਤ ਕਰ ਸਕਦੇ ਹੋ ਅਤੇ ਵੱਖ-ਵੱਖ ਹਾਰਡਵੇਅਰ ਡਿਵਾਈਸਾਂ ਵਿੱਚ E2E ਸੋਲਿਊਸ਼ਨ ਬਣਾਉਣ ਦੇ ਤਰੀਕੇ ਸਿੱਖ ਸਕਦੇ ਹੋ। Phi ਦਾ ਅਨੁਭਵ ਲੈਣ ਲਈ, models ਨਾਲ ਖੇਡੋ ਅਤੇ ਆਪਣੇ ਮਾਮਲਿਆਂ ਲਈ Phi ਨੂੰ ਕਸਟਮਾਈਜ਼ ਕਰੋ, [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) ਦੇ ਜ਼ਰੀਏ। ਜ਼ਿਆਦਾ ਜਾਣਕਾਰੀ ਲਈ [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) ਦੇ ਗੈਟਿੰਗ ਸਟਾਰਟ ਪੇਜ 'ਤੇ ਜਾਓ। + +**ਪਲੇਗ੍ਰਾਉਂਡ** +ਹਰ ਮਾਡਲ ਲਈ ਇਕ ਖਾਸ ਪਲੇਗ੍ਰਾਉਂਡ ਹੈ ਜਿੱਥੇ ਮਾਡਲ ਦੀ ਟੈਸਟਿੰਗ ਕਰ ਸਕਦੇ ਹੋ [Azure AI Playground](https://aka.ms/try-phi3)। + +### GitHub Models 'ਤੇ ਫਾਈ + +ਤੁਸੀਂ Microsoft Phi ਦੀ ਵਰਤੋਂ ਕਿਵੇਂ ਕਰਨ ਦੀ ਸਿੱਖਿਆ ਪ੍ਰਾਪਤ ਕਰ ਸਕਦੇ ਹੋ ਅਤੇ ਵੱਖ-ਵੱਖ ਹਾਰਡਵੇਅਰ ਡਿਵਾਈਸਾਂ ਵਿੱਚ E2E ਸੋਲਿਊਸ਼ਨ ਬਣਾਉਣ ਦੇ ਤਰੀਕੇ ਸਿੱਖ ਸਕਦੇ ਹੋ। Phi ਦਾ ਅਨੁਭਵ ਲੈਣ ਲਈ, ਮਾਡਲ ਨਾਲ ਖੇਡੋ ਅਤੇ ਆਪਣੇ ਮਾਮਲਿਆਂ ਲਈ Phi ਨੂੰ ਕਸਟਮਾਈਜ਼ ਕਰੋ, [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ਦੇ ਜ਼ਰੀਏ। ਜ਼ਿਆਦਾ ਜਾਣਕਾਰੀ ਲਈ [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) ਦੇ ਗੈਟਿੰਗ ਸਟਾਰਟ ਪੇਜ 'ਤੇ ਜਾਓ। + +**ਪਲੇਗ੍ਰਾਉਂਡ** +ਹਰ ਮਾਡਲ ਲਈ ਇਕ ਖਾਸ [ਪਲੇਗ੍ਰਾਉਂਡ ਹੈ ਜਿੱਥੇ ਮਾਡਲ ਨੂੰ ਟੈਸਟ ਕੀਤਾ ਜਾ ਸਕਦਾ ਹੈ](/md/02.QuickStart/GitHubModel_QuickStart.md)। + +### Hugging Face 'ਤੇ ਫਾਈ ਤੁਸੀਂ ਮਾਡਲ ਨੂੰ [Hugging Face](https://huggingface.co/microsoft) 'ਤੇ ਵੀ ਲੱਭ ਸਕਦੇ ਹੋ। -**Playground** +**ਪਲੇਗ੍ਰਾਉਂਡ** [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) - ## 🎒 ਹੋਰ ਕੋਰਸز + ## 🎒 ਹੋਰ ਕੋਰਸ -ਸਾਡੀ ਟੀਮ ਹੋਰ ਕੋਰਸਜ਼ ਬਣਾਉਂਦੀ ਹੈ! ਵੇਖੋ: +ਅਸੀਂ ਹੋਰ ਕੋਰਸ ਵੀ ਬਣਾਉਂਦੇ ਹਾਂ! ਵੇਖੋ: ### LangChain @@ -272,7 +275,7 @@ --- -### Core Learning +### ਕੋਰ ਸਿੱਖਿਆ [![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) @@ -283,41 +286,44 @@ --- -### Copilot Series +### ਕੋ-ਪਾਇਲਟ ਸੀਰੀਜ਼ [![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Responsible AI +## ਜ਼ਿੰਮੇਵਾਰ AI + +Microsoft ਆਪਣੇ ਗਾਹਕਾਂ ਨੂੰ ਸਾਡੇ AI ਉਤਪਾਦਾਂ ਦੀ ਜ਼ਿੰਮੇਵਾਰੀ ਨਾਲ ਵਰਤੋਂ ਵਿੱਚ ਮਦਦ ਕਰਨ, ਸਾਡੇ ਸਿੱਖਿਆਵਾਂ ਸਾਂਝੇ ਕਰਨ, ਅਤੇ ਟਰਾਂਸਪੇਰੈਂਸੀ ਨੋਟਸ ਅਤੇ ਪ੍ਰਭਾਵ ਮੁਲਾਂਕਣ ਵਰਗੇ ਸੰਦਾਂ ਰਾਹੀਂ ਭਰੋਸੇਯੋਗ ਭਾਈਚਾਰੇ ਨਿਰਮਾਣ 'ਤੇ ਬਹੁਤ ਜ਼ੋਰ ਦਿੰਦਾ ਹੈ। ਇਨ੍ਹਾਂ ਵਿੱਚੋਂ ਕਈ ਸਾਧਨ [https://aka.ms/RAI](https://aka.ms/RAI) 'ਤੇ ਮਿਲ ਸਕਦੇ ਹਨ। +Microsoft ਦਾ ਜ਼ਿੰਮੇਵਾਰ AI ਦਾ ਤਰੀਕਾ ਸਾਡੇ AI ਦੇ ਅਦਾਲਤੀ, ਭਰੋਸੇਯੋਗਤਾ ਅਤੇ ਸੁਰੱਖਿਆ, ਪ੍ਰਾਈਵੇਸੀ ਅਤੇ ਸੁਰੱਖਿਆ, ਸ਼ਾਮਿਲਤਾ, ਪਾਰਦਰਸ਼ਤਾ ਅਤੇ ਜਵਾਬਦੇਹੀ ਦੇ ਸਿਧਾਂਤਾਂ ਤੇ ਆਧਾਰਿਤ ਹੈ। + +ਵੱਡੇ ਪੈਮانے 'ਤੇ ਨੇਚੁਰਲ ਲੈਂਗਵੇਜ਼, ਚਿੱਤਰ ਅਤੇ ਬੋਲ ਚਾਲ ਦੇ ਮਾਡਲ - ਜਿਵੇਂ ਕਿ ਇਸ ਨਮੂਨੇ ਵਿੱਚ ਵਰਤੇ ਗਏ - ਅਕਸਰ ਅਣਦੇਖਿਆਂ, ਅਣਭਰੋਸੇਯੋਗ ਜਾਂ ਅਪਮਾਨਜਨਕ ਵਿਹਾਰ ਕਰ ਸਕਦੇ ਹਨ, ਜੋ ਕਿ ਨੁਕਸਾਨ ਪਹੁੰਚਾ ਸਕਦਾ ਹੈ। ਕਿਰਪਾ ਕਰਕੇ ਜੋਖਮ ਅਤੇ ਸੀਮਾਵਾਂ ਬਾਰੇ ਜਾਣਕਾਰੀ ਲਈ [Azure OpenAI ਸੇਵਾ ਟਰਾਂਸਪੇਰੈਂਸੀ ਨੋਟ](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ਨੂੰ ਵੇਖੋ। -ਮਾਇਕਰੋਸੋਫਟ ਸਾਡੇ ਗਾਹਕਾਂ ਨੂੰ ਸਾਡੇ AI ਉਤਪਾਦਾਂ ਨੂੰ ਜ਼ਿੰਮੇਵਾਰੀ ਨਾਲ ਵਰਤਣ ਵਿੱਚ ਮਦਦ ਕਰਨ ਲਈ ਬੱਝਾ ਹੈ, ਸਾਡੇ ਸਿਖਿਆ ਸਾਂਝੀਆਂ ਕਰਨ ਅਤੇ ਟਰਾਂਸਪੇਰੰਸੀ ਨੋਟਸ ਅਤੇ ਇੰਪੈਕਟ ਅਸੈਸਮੈਂਟ ਜਿਹੇ ਟੂਲਸ ਰਾਹੀਂ ਭਰੋਸਾ-ਅਧਾਰਿਤ ਭਾਈਚਾਰੇ ਬਣਾਉਣ ਲਈ। ਬਹੁਤ ਸਾਰੇ ਇਹ ਸਰੋਤ [https://aka.ms/RAI](https://aka.ms/RAI) 'ਤੇ ਮਿਲ ਸਕਦੇ ਹਨ। -ਮਾਇਕਰੋਸੋਫਟ ਦਾ ਜ਼ਿੰਮੇਵਾਰ AI ਲਈ ਰਵੱਈਆ ਸਾਡੇ ਨਿਆਂ, ਭਰੋਸੇਯੋਗਤਾ ਅਤੇ ਸੁਰੱਖਿਆ, ਪ੍ਰਾਈਵੇਸੀ ਅਤੇ ਸੁਰੱਖਿਆ, ਸ਼ਾਮਲਤਾ, ਪਾਰਦਰਸ਼ਤਾ ਅਤੇ ਜ਼ਿੰਮੇਦਾਰੀ ਦੇ AI ਨੀਤੀਆਂ 'ਤੇ ਆਧਾਰਿਤ ਹੈ। -ਵੱਡੇ ਪੱਧਰ ਦੇ ਕੁਦਰਤੀ ਭਾਸ਼ਾ, ਚਿੱਤਰ ਅਤੇ ਬੋਲਚਾਲ ਮਾਡਲ - ਜਿਵੇਂ ਕਿ ਇਸ ਉਦਾਹਰਨ ਵਿੱਚ ਵਰਤੇ ਗਏ ਹਨ - ਸੰਭਾਵਿਤ ਤੌਰ 'ਤੇ ਅਸਮਾਨ, ਅਭਰਾਮਕ ਜਾਂ ਅਪਮਾਨਜਨਕ ਤਰੀਕੇ ਨਾਲ ਵਰਤੋਂ ਕਰ ਸਕਦੇ ਹਨ, ਜਿਸ ਨਾਲ ਨੁਕਸਾਨ ਹੋ ਸਕਦਾ ਹੈ। ਸੂਚਿਤ ਹੋਣ ਲਈ ਕ੍ਰਿਪਾ ਕਰਕੇ [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) 'ਤੇ ਜਾਓ ਅਤੇ ਜੋਖਮ ਅਤੇ ਸੀਮਾਵਾਂ ਬਾਰੇ ਜਾਣੋ। -ਇਨ੍ਹਾਂ ਖਤਰਿਆਂ ਨੂੰ ਘਟਾਉਣ ਲਈ ਸਿਫਾਰਸ਼ ਕੀਤੀ ਗਈ ਰਣਨੀਤੀ ਇਹ ਹੈ ਕਿ ਤੁਹਾਡੇ ਆਰਕੀਟੈਕਚਰ ਵਿੱਚ ਇੱਕ ਸੁਰੱਖਿਆ ਪ੍ਰਣਾਲੀ ਸ਼ਾਮਿਲ ਕੀਤੀ ਜਾਵੇ ਜੋ ਹਾਨੀਕਾਰਕ ਵਿਹਾਰ ਦਾ ਪਤਾ ਲਗਾ ਸਕੇ ਅਤੇ ਉਸ ਨੂੰ ਰੋਕ ਸਕੇ। [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ਇਕ ਸਵਤੰਤਰ ਸੁਰੱਖਿਆ ਪਰਤ ਮੁਹੱਈਆ ਕਰਵਾਉਂਦਾ ਹੈ, ਜੋ ਐਪਲੀਕੇਸ਼ਨਾਂ ਅਤੇ ਸੇਵਾਵਾਂ ਵਿੱਚ ਹਾਨੀਕਾਰਕ ਉਪਭੋਗਤਾ-ਜਨਰੇਟ ਅਤੇ AI-ਜਨਰੇਟ ਸਮੱਗਰੀ ਦਾ ਪਤਾ ਲਗਾਉਣ ਯੋਗ ਹੈ। Azure AI Content Safety ਵਿੱਚ ਟੈਕਸਟ ਅਤੇ ਚਿੱਤਰ API ਸ਼ਾਮਿਲ ਹਨ ਜੋ ਤੁਹਾਨੂੰ ਹਾਨੀਕਾਰਕ ਸਮੱਗਰੀ ਦਾ ਪਤਾ ਲਗਾਉਣ ਦੇ ਯੋਗ ਬਣਾਉਂਦੇ ਹਨ। Microsoft Foundry ਦੇ ਅੰਦਰ, Content Safety ਸੇਵਾ ਤੁਹਾਨੂੰ ਵੱਖ-ਵੱਖ ਮਾਡਲਿਟੀਜ਼ ਵਿੱਚ ਹਾਨੀਕਾਰਕ ਸਮੱਗਰੀ ਖੋਜਣ ਲਈ ਨਮੂਨੇ ਦੇ ਕੋਡ ਨੂੰ ਵੇਖਣ, ਖੋਜਣ ਅਤੇ ਅਜ਼ਮਾਉਣ ਦੀ ਆਗਿਆ ਦਿੰਦੀ ਹੈ। ਹੇਠਾਂ ਦਿੱਤਾ [quickstart ਦਸਤਾਵੇਜ਼](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ਤੁਹਾਨੂੰ ਸੇਵਾ ਨੂੰ ਬੇਨਤੀ ਕਰਨ ਵਿਚ ਰਾਹਨੁਮਾ ਕਰਦਾ ਹੈ। +ਇਹ ਜੋਖਮਾਂ ਨੂੰ ਘਟਾਉਣ ਲਈ ਸਿਫਾਰਸ਼ੀ ਕੀਤੀ ਗਈ ਪদ্ধਤੀ ਇਹ ਹੈ ਕਿ ਤੁਹਾਡੇ ਸਾਂਰਚਨਾ ਵਿੱਚ ਇੱਕ ਸੁਰੱਖਿਆ ਪ੍ਰਣਾਲੀ ਸ਼ਾਮਲ ਕੀਤੀ ਜਾਵੇ ਜੋ ਖਤਰਨਾਕ ਵਰਤਾਰ ਨੂੰ ਪਛਾਣ ਅਤੇ ਰੋਕ ਸਕੇ। [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ਇੱਕ ਸਵਤੰਤਰ ਸੁਰੱਖਿਆ ਪਰਤ ਪ੍ਰਦਾਨ ਕਰਦਾ ਹੈ, ਜੋ ਐਪਲੀਕੇਸ਼ਨਾਂ ਅਤੇ ਸੇਵਾਵਾਂ ਵਿੱਚ ਖਤਰਨਾਕ ਉਪਭੋਗਤਾ-ਜਨਰੀਤ ਅਤੇ ਏਆਈ-ਜਨਰੀਤ ਸਮੱਗਰੀ ਦਾ ਪਤਾ ਲਗਾਉਣ ਦੇ ਯੋਗ ਹੈ। Azure AI Content Safety ਵਿੱਚ ਟੈਕਸਟ ਅਤੇ ਚਿੱਤਰ ਏਪੀਆਈਜ਼ ਸ਼ਾਮਲ ਹਨ ਜੋ ਤੁਹਾਨੂੰ ਖਤਰਨਾਕ ਸਮੱਗਰੀ ਦਾ ਪਤਾ ਲਾਉਣ ਦੀ ਆਗਿਆ ਦਿੰਦੇ ਹਨ। ਮਾਇਕਰੋਸਾਫਟ ਫਾਊਂਡਰੀ ਦੇ ਅੰਦਰ, Content Safety ਸੇਵਾ ਤੁਹਾਨੂੰ ਵੱਖ-ਵੱਖ ਮੋਡਾਲਿਟੀਜ਼ ਵਿੱਚ ਖਤਰਨਾਕ ਸਮੱਗਰੀ ਪਛਾਣ ਲਈ ਨਮੂਨਾ ਕੋਡ ਦੇਖਣ, ਖੰਗਾਲਣ ਅਤੇ ਕੋਸ਼ਿਸ਼ ਕਰਨ ਦੀ ਆਗਿਆ ਦਿੰਦੀ ਹੈ। ਹੇਠਾਂ ਦਿੱਤੀ ਗਈ [quickstart ਦਸਤਾਵੇਜ਼](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ਤੁਹਾਡੇ ਲਈ ਸੇਵਾ ਨੂੰ ਬੇਨਤੀਆਂ ਭੇਜਣ ਦੀ ਪ੍ਰਕਿਰਿਆ ਦਿਖਾਉਂਦੀ ਹੈ। -ਇੱਕ ਹੋਰ ਪਹਲੂ ਜੋ ਧਿਆਨ ਵਿੱਚ ਰੱਖਣ ਯੋਗ ਹੈ, ਉਹ ਹੈ ਸਮੁੱਚੇ ਐਪਲੀਕੇਸ਼ਨ ਦੀ ਕਾਰਗੁਜ਼ਾਰੀ। ਬਹੁ-ਮਾਡਲ ਅਤੇ ਬਹੁ-ਮਾਡਲ ਵਾਲੀਆਂ ਐਪਲੀਕੇਸ਼ਨਾਂ ਨਾਲ, ਅਸੀਂ ਕਾਰਗੁਜ਼ਾਰੀ ਦਾ ਮਤਲਬ ਇਹ ਰੱਖਦੇ ਹਾਂ ਕਿ ਪ੍ਰਣਾਲੀ ਤੁਹਾਨੂੰ ਅਤੇ ਤੁਹਾਡੇ ਉਪਭੋਗਤਿਆਂ ਦੀ ਉਮੀਦਾਂ ਦੇ ਅਨੁਸਾਰ ਕੰਮ ਕਰੇ, ਜਿਸ ਵਿੱਚ ਹਾਨੀਕਾਰਕ ਨਤੀਜੇ ਨਾ ਬਣਾਉਣਾ ਸ਼ਾਮਿਲ ਹੈ। ਇਹ ਜਰੂਰੀ ਹੈ ਕਿ ਤੁਸੀਂ ਆਪਣੇ ਸਮੁੱਚੇ ਐਪਲੀਕੇਸ਼ਨ ਦੀ ਕਾਰਗੁਜ਼ਾਰੀ ਦੀ ਮੁਲਾਂਕਣ ਕਰੋ [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) ਦੀ ਵਰਤੋਂ ਕਰਕੇ। ਤੁਹਾਡੇ ਕੋਲ [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) ਨਾਲ ਬਣਾਉਣ ਅਤੇ ਮੁਲਾਂਕਣ ਕਰਨ ਦੀ ਸਮਰੱਥਾ ਵੀ ਹੈ। +ਇੱਕ ਹੋਰ پهلو ਜੋ ਧਿਆਨ ਵਿੱਚ ਰੱਖਣਾ ਚਾਹੀਦਾ ਹੈ ਉਹ ਕੁੱਲ ਐਪਲੀਕੇਸ਼ਨ ਦੀ ਪ੍ਰਦਰਸ਼ਨਸ਼ੀਲਤਾ ਹੈ। ਬਹੁ-ਮੋਡਲ ਅਤੇ ਬਹੁ-ਮਾਡਲ ਐਪਲੀਕੇਸ਼ਨਾਂ ਨਾਲ, ਅਸੀਂ ਪ੍ਰਦਰਸ਼ਨਸ਼ੀਲਤਾ ਦੱਸਦੇ ਹਾਂ ਕਿ ਸਿਸਟਮ ਉਸ ਤਰ੍ਹਾਂ ਕੰਮ ਕਰਦਾ ਹੈ ਜਿਵੇਂ ਤੂੰ ਅਤੇ ਤੇਰੇ ਵਰਤੋਂਕਾਰ ਉਮੀਦ ਕਰਦੇ ਹਨ, ਜਿਸ ਵਿੱਚ ਖਤਰਨਾਕ ਨਤੀਜਿਆਂ ਦਾ ਉਤਪਾਤਨ ਸ਼ਾਮਲ ਨਹੀਂ ਹੈ। ਤੁਹਾਡੇ ਲਈ ਇਹ ਜਰੂਰੀ ਹੈ ਕਿ ਤੁਸੀਂ ਆਪਣੇ ਕੁੱਲ ਐਪਲੀਕੇਸ਼ਨ ਦੀ ਪ੍ਰਦਰਸ਼ਨਸ਼ੀਲਤਾ ਦਾ ਮੁਲਾਂਕਣ ਕਰੋ [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) ਦੀ ਵਰਤੋਂ ਕਰਕੇ। ਤੁਹਾਡੇ ਕੋਲ [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) ਬਣਾਉਣ ਅਤੇ ਮੁਲਾਂਕਣ ਕਰਨ ਦੀ ਸਮਰੱਥਾ ਵੀ ਹੈ। -ਤੁਸੀਂ ਆਪਣੇ ਵਿਕਾਸੀਲੇਵਲ.Environment ਵਿੱਚ ਆਪਣੇ AI ਐਪਲੀਕੇਸ਼ਨ ਦਾ ਮੁਲਾਂਕਣ [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਕਰ ਸਕਦੇ ਹੋ। ਕਿਵੇਂ ਇੱਕ ਟੈਸਟ ਡੈਟਾ ਸੈੱਟ ਜਾਂ ਟਾਰਗੇਟ ਦਿੱਤਾ ਜਾਵੇ, ਤਾਂ ਤੁਹਾਡੇ ਜਨਰੇਟਿਵ AI ਐਪਲੀਕੇਸ਼ਨ ਦੀਆਂ ਪੈਦਾਵਾਰਾਂ ਨੂੰ ਕੁਆੰਟੀਟੇਟਿਵ ਤੌਰ 'ਤੇ ਮਾਪਿਆ ਜਾਂਦਾ ਹੈ ਵਿਟ ਬਿਲਟ-ਇਨ ਜਾਂ ਕਸਟਮ ਇਵੈਲੂਏਟਰ ਜਾਂ ਤੁਹਾਡੇ ਚੋਣ ਦੇ ਮੁਤਾਬਕ। azure ai evaluation sdk ਨਾਲ ਸ਼ੁਰੂਆਤ ਕਰਨ ਲਈ ਅਤੇ ਆਪਣੀ ਪ੍ਰਣਾਲੀ ਦਾ ਮੁਲਾਂਕਣ ਕਰਨ ਲਈ ਤੁਸੀਂ [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) ਦੀ ਪਾਲਣਾ ਕਰ ਸਕਦੇ ਹੋ। ਇੱਕ ਵਾਰੀ ਤੁਹਾਡੇ ਕੋਲ ਮੁਲਾਂਕਣ ਚਲਾਇਆ ਜਾਵੇ ਤਾਂ ਤੁਸੀਂ [Microsoft Foundry ਵਿੱਚ ਨਤੀਜੇ ਦਰਸਾ ਸਕਦੇ ਹੋ](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)। +ਤੁਸੀਂ ਆਪਣੇ ਵਿਕਾਸ ਦੇ ਵਾਤਾਵਰਣ ਵਿੱਚ ਆਪਣੀ ਏਆਈ ਐਪਲੀਕੇਸ਼ਨ ਦਾ ਮੁਲਾਂਕਣ [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ਦੀ ਵਰਤੋਂ ਨਾਲ ਕਰ ਸਕਦੇ ਹੋ। ਚਾਹੇ ਇੱਕ ਟੈਸਟ ਡੇਟਾਸੇਟ ਹੋਵੇ ਜਾਂ ਟਾਰਗੇਟ, ਤੁਹਾਡੇ ਜਨਰੇਟਿਵ ਏਆਈ ਐਪਲੀਕੇਸ਼ਨ ਦੇ ਜਨਰੇਸ਼ਨਾਂ ਦੀ ਮਾਤਰਾਤਮਕ ਮਾਪ ਪੇਸ਼ਗੀ ਤਿਆਰ ਕੀਤੇ ਇਵੈਲੂਏਟਰ ਜਾਂ ਤੁਹਾਡੇ ਚੋਣ ਦੇ ਅਨੁਕੂਲ ਇਵੈਲੂਏਟਰ ਨਾਲ ਕੀਤੀ ਜਾਂਦੀ ਹੈ। ਆਪਣੇ ਸਿਸਟਮ ਦੇ ਮੁਲਾਂਕਣ ਲਈ azure ai evaluation sdk ਨਾਲ ਸ਼ੁਰੂਆਤ ਕਰਨ ਲਈ, ਤੁਸੀਂ [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) ਦੀ ਪਾਲਣਾ ਕਰ ਸਕਦੇ ਹੋ। ਇੱਕ ਵਾਰੀ ਤੁਸੀਂ ਮੁਲਾਂਕਣ ਚਲਾਉਂਦੇ ਹੋ, ਤੁਸੀਂ [Microsoft Foundry ਵਿੱਚ ਨਤੀਜਿਆਂ ਦਾ ਦਰਸ਼ਨ ਕਰ ਸਕਦੇ ਹੋ](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## ਟ੍ਰੇਡਮਾਰਕ +## ਟਰੇਡਮਾਰਕ -ਇਹ ਪ੍ਰੋਜੈਕਟ ਪ੍ਰੋਜੈਕਟਾਂ, ਉਤਪਾਦਾਂ ਜਾਂ ਸੇਵਾਵਾਂ ਲਈ ਟ੍ਰੇਡਮਾਰਕ ਜਾਂ ਲੋਗੋ ਸ਼ਾਮਿਲ ਕਰ ਸਕਦਾ ਹੈ। Microsoft ਦੇ ਟ੍ਰੇਡਮਾਰਕ ਜਾਂ ਲੋਗੋ ਦਾ ਪ੍ਰਮਾਣਿਤ ਵਰਤੋਂ [Microsoft ਦੀ ਟ੍ਰੇਡਮਾਰਕ ਅਤੇ ਬ੍ਰਾਂਡ ਗਾਈਡਲਾਈਨਾਂ](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) ਦੇ ਅਧੀਨ ਹੈ ਅਤੇ ਉਹਨਾਂ ਦੀ ਪਾਲਣਾ ਕਰਨੀ ਜਰੂਰੀ ਹੈ। ਇਸ ਪ੍ਰੋਜੈਕਟ ਦੀ ਤਬਦੀਲ ਕੀਤੀ ਗਈ ਵਰਜਨ ਵਿੱਚ Microsoft ਦੇ ਟ੍ਰੇਡਮਾਰਕ ਜਾਂ ਲੋਗੋ ਦੀ ਵਰਤੋਂ ਨਾਲ ਕੋਈ ਭ੍ਰਮ ਜਾਂ Microsoft ਦੀ ਸਪਾਂਸਰਸ਼ਿਪ ਦਾ ਭਾਵ ਨਹੀਂ ਹੋਣਾ ਚਾਹੀਦਾ। ਤੀਸਰੀ-ਪੱਖ ਦੇ ਟ੍ਰੇਡਮਾਰਕਾਂ ਜਾਂ ਲੋਗੋ ਦੀ ਵਰਤੋਂ ਉਸ ਤੀਸਰੇ-ਪੱਖ ਦੀਆਂ ਨੀਤੀਆਂ ਦੇ ਅਧੀਨ ਹੈ। +ਇਸ ਪ੍ਰੋਜੈਕਟ ਵਿੱਚ ਪ੍ਰੋਜੈਕਟਾਂ, ਉਤਪਾਦਾਂ ਜਾਂ ਸੇਵਾਵਾਂ ਲਈ ਟਰੇਡਮਾਰਕ ਜਾਂ ਲੋਗੋਜ਼ ਹੋ ਸਕਦੇ ਹਨ। Microsoft ਟਰੇਡਮਾਰਕ ਜਾਂ ਲੋਗੋਜ਼ ਦੀ ਮਨਜ਼ੂਰਸ਼ੁਦਾ ਵਰਤੋਂ [Microsoft ਦੇ ਟਰੇਡਮਾਰਕ ਅਤੇ ਬ੍ਰਾਂਡ ਦਿਸ਼ਾ-ਨਿਰਦੇਸ਼](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) ਦੇ ਅਧੀਨ ਹੈ ਅਤੇ ਇਸਦੀ ਪਾਲਣਾ ਕਰਨੀ ਜਰੂਰੀ ਹੈ। +Microsoft ਟਰੇਡਮਾਰਕ ਜਾਂ ਲੋਗੋਜ਼ ਦੀ ਇਹ ਪ੍ਰੋਜੈਕਟ ਦੇ ਸੰਜੋੜੇ ਸੰਸਕਰਨਾਂ ਵਿੱਚ ਵਰਤੋਂ Microsoft ਦੀ ਸਹਿਯੋਗੀਤਾ ਦਰਸਾਉਣ ਜਾਂ ਗਲਤੀ ਪੈਦਾ ਕਰਨ ਵਾਲੀ ਨਹੀਂ ਹੋਣੀ ਚਾਹੀਦੀ। ਕਿਸੇ ਵੀ ਤੀਸਰੇ ਧਿਰ ਦੇ ਟਰੇਡਮਾਰਕ ਜਾਂ ਲੋਗੋਜ਼ ਦੀ ਵਰਤੋਂ ਉਹਨਾਂ ਤੀਸਰੇ ਧਿਰ ਦੀਆਂ ਨੀਤੀਆਂ ਦੇ ਅਧੀਨ ਹੁੰਦੀ ਹੈ। -## ਮਦਦ ਪ੍ਰਾਪਤ ਕਰੋ +## ਸਹਾਇਤਾ ਪ੍ਰਾਪਤ ਕਰਨਾ -ਜੇ ਤੁਸੀਂ ਫੱਸ ਜਾਂਦੇ ਹੋ ਜਾਂ AI ਐਪ ਬਨਾਉਣ ਬਾਰੇ ਕੋਈ ਸਵਾਲ ਹੈ, ਤਾਂ ਜੁੜੋ: +ਜੇ ਤੁਸੀਂ ਫਸ ਜਾਂਦੇ ਹੋ ਜਾਂ ਏਆਈ ਐਪ ਬਣਾਉਣ ਬਾਰੇ ਕੋਈ ਸਵਾਲ ਹੈ, ਤਦ ਸ਼ਾਮਿਲ ਹੋਵੋ: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -ਜੇ ਤੁਹਾਡੇ ਕੋਲ ਉਤਪਾਦ ਸਬੰਧੀ ਪ੍ਰਤੀਕ੍ਰਿਆ ਜਾਂ ਬਿਲਡਿੰਗ ਦੌਰਾਨ ਕੋਈ ਗਲਤੀਆਂ ਹਨ ਤਾਂ ਮਿਲੋ: +ਜੇ ਤੁਹਾਨੂੰ ਉਤਪਾਦ ਫੀਡਬੈਕ ਜਾਂ ਤਿਆਰ ਕਰਦੇ ਸਮੇਂ ਕੋਈ ਗਲਤੀ ਮਿਲਦੀ ਹੈ, ਤਾਂ ਜਾਓ: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**ਦਾਇਤਵ-ਤਿਆਗ:** -ਇਹ ਦਸਤਾਵੇਜ਼ AI ਅਨੁਵਾਦ ਸੇਵਾ [Co-op Translator](https://github.com/Azure/co-op-translator) ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਅਨੁਵਾਦਿਤ ਕੀਤਾ ਗਿਆ ਹੈ। ਜਦੋਂ ਕਿ ਅਸੀਂ ਸਹੀਤਾ ਲਈ ਕੋਸ਼ਿਸ਼ ਕਰਦੇ ਹਾਂ, ਕਿਰਪਾ ਕਰਕੇ ਵੇਖੋ ਕਿ ਸੁਚਾਲਿਤ ਅਨੁਵਾਦਾਂ ਵਿੱਚ ਗਲਤੀਆਂ ਜਾਂ ਅਸਮਰੱਥਾ ਹੋ ਸਕਦੀ ਹੈ। ਮੂਲ ਦਸਤਾਵੇਜ਼ ਆਪਣੀ ਮੂਲ ਭਾਸ਼ਾ ਵਿੱਚ ਪ੍ਰਮਾਣਿਕ ਸਰੋਤ ਵਜੋਂ ਮੰਨਿਆ ਜਾਣਾ ਚਾਹੀਦਾ ਹੈ। ਮਹੱਤਵਪੂਰਣ ਜਾਣਕਾਰੀ ਲਈ, ਪੇਸ਼ੇਵਰ ਮਨੁੱਖੀ ਅਨੁਵਾਦ ਦੀ ਸਿਫ਼ਾਰਿਸ਼ ਕੀਤੀ ਜਾਂਦੀ ਹੈ। ਅਸੀਂ ਇਸ ਅਨੁਵਾਦ ਦੀ ਵਰਤੋਂ ਤੋਂ ਪੈਦਾ ਹੋਣ ਵਾਲੀਆਂ ਕਿਸੇ ਵੀ ਭੁੱਲਾਂ ਜਾਂ ਗਲਤਫਹਮੀਆਂ ਲਈ ਜ਼ਿੰਮੇਵਾਰ ਨਹੀਂ ਹਾਂ। +**ਅਸਵੀਕਾਰੋਪਣ**: +ਇਸ ਦਸਤਾਵੇਜ਼ ਦਾ ਅਨੁਵਾਦ ਏਆਈ ਅਨੁਵਾਦ ਸੇਵਾ [Co-op Translator](https://github.com/Azure/co-op-translator) ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਕੀਤਾ ਗਿਆ ਹੈ। ਜਦੋਂ ਕਿ ਅਸੀਂ ਸਹੀਤਾਵਾਂ ਲਈ ਯਤਨਸ਼ੀਲ ਹਾਂ, ਕਿਰਪਾ ਕਰਕੇ ਧਿਆਨ ਰੱਖੋ ਕਿ ਸਵੈਚਾਲਿਤ ਅਨੁਵਾਦਾਂ ਵਿੱਚ ਗਲਤੀਆਂ ਜਾਂ ਅਸਮੱਤਿਆਵਾਂ ਹੋ ਸਕਦੀਆਂ ਹਨ। ਮੂਲ ਦਸਤਾਵੇਜ਼ ਆਪਣੀ ਮੂਲ ਭਾਸ਼ਾ ਵਿੱਚ ਅਧਿਕਾਰਕ ਸਰੋਤ ਮੰਨਿਆ ਜਾਣਾ ਚਾਹੀਦਾ ਹੈ। ਜਰੂਰੀ ਜਾਣਕਾਰੀ ਲਈ, ਪੇਸ਼ੇਵਰ ਮਨੁੱਖੀ ਅਨੁਵਾਦ ਦੀ ਸਿਫ਼ਾਰਸ਼ ਕੀਤੀ ਜਾਂਦੀ ਹੈ। ਅਸੀਂ ਇਸ ਅਨੁਵਾਦ ਦੇ ਉਪਯੋਗ ਤੋਂ ਪੈਦਾ ਹੋਣ ਵਾਲੀਆਂ ਕਿਸੇ ਵੀ ਗਲਤਫਹਿਮੀਆਂ ਜਾਂ ਗਲਤ ਵਿਆਖਿਆਵਾਂ ਲਈ ਜਵਾਬਦੇਹ ਨਹੀਂ ਹਾਂ। \ No newline at end of file diff --git a/translations/pa/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/pa/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..979e061b8 --- /dev/null +++ b/translations/pa/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 ਮਾਈਕਰੋਸੌਫਟ phi-4-mini ਅਤੇ SQLite FTS5 ਨਾਲ ਹਾਈਬ੍ਰਿਡ RAG (ਜ਼ੀਰੋ-ਕਲਾਉਡ SLM)\n", + "\n", + "> **ਲੇਖਕ:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **ਕੇਂਦਰ:** ਛੋਟੇ ਭਾਸ਼ਾ ਮਾਡਲ (SLMs), SQLite FTS5 BM25, ਡੈਨਸ ਐਂਬੈਡਿੰਗਜ਼, ਰੀਸੀਪ੍ਰੋਕਲ ਰੈਂਕ ਫਿਊਜ਼ਨ (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. ਪ੍ਰੇਰਣਾ: ਸਥਾਨਕ SLMs ਵਿੱਚ ਕੀਵਰਡ ਰੀਕਾਲ ਦੀ ਸਮੱਸਿਆ\n", + "ਸਧਾਰਨ RAG ਆਰਕੀਟੈਕਚਰ ਜੋ ਸਿਰਫ਼ ਡੈਨਸ ਵੇਕਟਰ ਐਂਬੈਡਿੰਗਜ਼ 'ਤੇ ਆਧਾਰਿਤ ਹੁੰਦੇ ਹਨ, ਅਕਸਰ ਸਹੀ ਨੰਬਰਾਤਮਕ ਟੋਕਨ (ਜਿਵੇਂ `2,340,000 TL`, ਕਾਂਟ੍ਰੈਕਟ ਕੋਡ, ਖਾਤਾ ਨੰਬਰ) ਪ੍ਰਾਪਤ ਕਰਨ ਵਿਚ ਫੇਲ ਜਾਂਦੇ ਹਨ। \n", + "ਵਿਰੋਧ ਵਿੱਚ, ਸਪੈਰਸ ਲੈਕਸੀਕਲ ਸਰਚ (BM25) ਸੈਮਾਂਟਿਕ ਸਮਾਨਾਰਥਕ ਅਤੇ ਵੱਖ-ਵੱਖ ਢੰਗ ਨਾਲ ਪੁੱਛੇ ਗਏ ਪ੍ਰਸ਼ਨਾਂ ਨੂੰ ਨਹੀਂ ਪਕੜ ਪਾਉਂਦੀ।\n", + "\n", + "ਇਹ ਕੂਕਬੁੱਕ ਦਿਖਾਉਂਦਾ ਹੈ ਕਿ ਕਿਵੇਂ ਇੱਕ **ਉੱਚ-ਗਤੀ, ਇਨ-ਮੇਮੋਰੀ ਹਾਈਬ੍ਰਿਡ ਰੀਟੀਵਲ ਇੰਜਣ** ਬਣਾਇਆ ਜਾ ਸਕਦਾ ਹੈ ਜੋ ਕਿ:\n", + "1. **ਡੈਨਸ ਵੇਕਟਰਜ਼** (ਕੋਜ਼ਾਈਨ ਸਮਾਨਤਾ)\n", + "2. **ਸਪੈਰਸ ਲੈਕਸੀਕਲ ਸਰਚ** (SQLite FTS5 BM25)\n", + "3. **ਰੀਸੀਪ੍ਰੋਕਲ ਰੈਂਕ ਫਿਊਜ਼ਨ (RRF, $k=60$)**\n", + "4. **ਗ੍ਰਾਊਂਡਡ ਸਿਟੇਸ਼ਨ ਜਨਰੇਸ਼ਨ (`[1]`, `[2]`)** ਮਾਈਕਰੋਸੌਫਟ `phi-4-mini` ਨਾਲ।\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. ਦੁਵੀਂ SQLite ਸਕੀਮਾ (ਘਣੇ ਵੇਕਟਰ + ਵਰਚੁਅਲ FTS5 BM25 ਟੇਬਲ)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. ਨਮੂਨਾ ਸੰਗ੍ਰਹਿ ਅਤੇ ਕਾਰਜਨਵਾਈ ਬੈਂਚਮਾਰਕ\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. ਮਾਈਕ੍ਰੋਸਾਫਟ phi-4-mini ਲਈ ਗਰਾਊਂਡਡ ਪ੍ਰੌਮਪਟ ਫਾਰਮੂਲੇਸ਼ਨ\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**ਅਸਵੀਕਾਰੋਪਣ**:\nਇਸ ਦਸਤਾਵੇਜ਼ ਦਾ ਅਨੁਵਾਦ ਏਆਈ ਅਨੁਵਾਦ ਸੇਵਾ [Co-op Translator](https://github.com/Azure/co-op-translator) ਦੀ ਵਰਤੋਂ ਕਰਕੇ ਕੀਤਾ ਗਿਆ ਹੈ। ਜਦੋਂ ਕਿ ਅਸੀਂ ਸਹੀਤਾਵਾਂ ਲਈ ਯਤਨਸ਼ੀਲ ਹਾਂ, ਕਿਰਪਾ ਕਰਕੇ ਧਿਆਨ ਰੱਖੋ ਕਿ ਸਵੈਚਾਲਿਤ ਅਨੁਵਾਦਾਂ ਵਿੱਚ ਗਲਤੀਆਂ ਜਾਂ ਅਸਮੱਤਿਆਵਾਂ ਹੋ ਸਕਦੀਆਂ ਹਨ। ਮੂਲ ਦਸਤਾਵੇਜ਼ ਆਪਣੀ ਮੂਲ ਭਾਸ਼ਾ ਵਿੱਚ ਅਧਿਕਾਰਕ ਸਰੋਤ ਮੰਨਿਆ ਜਾਣਾ ਚਾਹੀਦਾ ਹੈ। ਜਰੂਰੀ ਜਾਣਕਾਰੀ ਲਈ, ਪੇਸ਼ੇਵਰ ਮਨੁੱਖੀ ਅਨੁਵਾਦ ਦੀ ਸਿਫ਼ਾਰਸ਼ ਕੀਤੀ ਜਾਂਦੀ ਹੈ। ਅਸੀਂ ਇਸ ਅਨੁਵਾਦ ਦੇ ਉਪਯੋਗ ਤੋਂ ਪੈਦਾ ਹੋਣ ਵਾਲੀਆਂ ਕਿਸੇ ਵੀ ਗਲਤਫਹਿਮੀਆਂ ਜਾਂ ਗਲਤ ਵਿਆਖਿਆਵਾਂ ਲਈ ਜਵਾਬਦੇਹ ਨਹੀਂ ਹਾਂ।\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/pcm/.co-op-translator.json b/translations/pcm/.co-op-translator.json index a7b09f809..e514ac848 100644 --- a/translations/pcm/.co-op-translator.json +++ b/translations/pcm/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "pcm" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T07:57:14+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:30:45+00:00", "source_file": "README.md", "language_code": "pcm" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "pcm" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:31:43+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "pcm" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-12-21T15:48:52+00:00", diff --git a/translations/pcm/README.md b/translations/pcm/README.md index 1abe58783..a57f001d8 100644 --- a/translations/pcm/README.md +++ b/translations/pcm/README.md @@ -1,4 +1,4 @@ -# Phi Cookbook: Hands-On Examples wit Microsoft Phi Models +# Phi Cookbook: Hands-On Examples wit Microsoft's Phi Models [![Open and use the samples in GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Open in Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) @@ -14,16 +14,16 @@ [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi na series of open source AI models wey Microsoft develop. +Phi na series of open source AI models wey Microsoft develop. -Phi na di most powerful and cost-effective small language model (SLM) now, get beta beta benchmarks for multi-language, reasoning, text/chat generation, coding, images, audio and oda scenarios. +Phi na di most powerful and cost-effective small language model (SLM) now, get better benchmarks for multi-language, reasoning, text/chat generation, coding, images, audio and oda tins dem. -You fit deploy Phi go cloud or edge devices, and you fit easy build generative AI applications wit limited computing power. +You fit deploy Phi for cloud or edge devices, and you fit build generative AI apps quick quick wit small computing power. -Make you follow these steps to start use these resource: +Follow these steps to start use dis resource: 1. **Fork di Repository**: Click [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Clone di Repository**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Join Di Microsoft AI Discord Community and meet experts and fellow developers**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Join The Microsoft AI Discord Community and meet experts and fellow developers**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/pcm/cover.eb18d1b9605d754b.webp) @@ -34,9 +34,9 @@ Make you follow these steps to start use these resource: [Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](./README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **You prefer to Clone for your machine?** +> **You prefer to clone am locally?** > -> Dis repository get 50+ language translations wey heavy for download size. To clone witout translations, use sparse checkout: +> Dis repository get over 50 language translations wey make di download size big. If you wan clone without di translations, use sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,7 +52,7 @@ Make you follow these steps to start use these resource: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Dis one go give you everything wey you need to finish di course with fast download. +> Dis one go give you everything you need to complete di course quick quick wit faster download. ## Table of Contents @@ -101,15 +101,17 @@ Make you follow these steps to start use these resource: - RAG with Azure AI Search - [How to use Phi-4-mini and Phi-4-multimodal(RAG) with Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud Local Hybrid RAG with SQLite FTS5 and phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Phi application development samples - Text & Chat Applications - - Phi-4 Samples + - Phi-4 Samples - [📓] [Chat With Phi-4-mini ONNX Model](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Chat with Phi-4 local ONNX Model .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - [Chat .NET Console App with Phi-4 ONNX using Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 Samples - - [Local Chatbot in the browser using Phi3, ONNX Runtime Web and WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + + - Phi-3 / 3.5 Sampul Dem + - [Local Chatbot for browser wey dey use Phi3, ONNX Runtime Web and WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [Multi Model - Interactive Phi-3-mini and OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Building a wrapper and using Phi-3 with MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) @@ -126,27 +128,27 @@ Make you follow these steps to start use these resource: - [Q&A .NET Example using local ONNX Phi-3 model using the Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - [Console chat .NET app with Semantic Kernel and Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK Code Based Samples - - Phi-4 Samples + - Azure AI Inference SDK Code Based Sampul Dem + - Phi-4 Sampul Dem - [📓] [Generate project code using Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 Samples + - Phi-3 / 3.5 Sampul Dem - [Build your own Visual Studio Code GitHub Copilot Chat with Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - [Create your own Visual Studio Code Chat Copilot Agent with Phi-3.5 by GitHub Models](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - Advanced Reasoning Samples - - Phi-4 Samples - - [📓] [Phi-4-mini-reasoning or Phi-4-reasoning Samples](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - Advanced Reasoning Sampul Dem + - Phi-4 Sampul Dem + - [📓] [Phi-4-mini-reasoning or Phi-4-reasoning Sampul Dem](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - [📓] [Fine-tuning Phi-4-mini-reasoning with Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Fine-tuning Phi-4-mini-reasoning with Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-reasoning with GitHub Models](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-reasoning with Microsoft Foundry Models](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demos - - [Phi-4-mini demos hosted on Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demos hosted on Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Vision Samples - - Phi-4 Samples - - [📓] [Use Phi-4-multimodal to read images and generate code](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 Samples + - [Phi-4-mini demos wey dem host for Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal demos wey dem host for Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Vision Sampul Dem + - Phi-4 Sampul Dem + - [📓] [Use Phi-4-multimodal to read pictures and generate code](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 Sampul Dem - [📓][Phi-3-vision-Image text to text](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) @@ -158,39 +160,39 @@ Make you follow these steps to start use these resource: - [Phi-3 Vision Local ONNX Model using the Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - [Menu based Phi-3 Vision Local ONNX Model using the Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - Reasoning-Vision Samples + - Reasoning-Vision Sampul Dem - Phi-4-Reasoning-Vision-15B - [📓] [Using Phi-4-Reasoning-Vision-15B to detect jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [Using Phi-4-Reasoning-Vision-15B to math](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - [📓] [Using Phi-4-Reasoning-Vision-15B to detect UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - Math Samples - - Phi-4-Mini-Flash-Reasoning-Instruct Samples [Math Demo with Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Math Sampul Dem + - Phi-4-Mini-Flash-Reasoning-Instruct Sampul Dem [Math Demo with Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - Audio Samples - - Phi-4 Samples + - Audio Sampul Dem + - Phi-4 Sampul Dem - [📓] [Extracting audio transcripts using Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal Audio Sample](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal Speech Translation Sample](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET console application using Phi-4-multimodal Audio to analyze an audio file and generate transcript](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [📓] [Phi-4-multimodal Audio Sampul Dem](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal Speech Translation Sampul Dem](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET console application using Phi-4-multimodal Audio to analyze audio file and generate transcript](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - MOE Samples - - Phi-3 / 3.5 Samples - - [📓] [Phi-3.5 Mixture of Experts Models (MoEs) Social Media Sample](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - MOE Sampul Dem + - Phi-3 / 3.5 Sampul Dem + - [📓] [Phi-3.5 Mixture of Experts Models (MoEs) Social Media Sampul Dem](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - [📓] [Building a Retrieval-Augmented Generation (RAG) Pipeline with NVIDIA NIM Phi-3 MOE, Azure AI Search, and LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Function Calling Samples - - Phi-4 Samples 🆕 + - Function Calling Sampul Dem + - Phi-4 Sampul Dem 🆕 - [📓] [Using Function Calling With Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - [📓] [Using Function Calling to create multi-agents With Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Using Function Calling with Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [Using Function Calling with ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Multimodal Mixing Samples - - Phi-4 Samples 🆕 + - Multimodal Mixing Sampul Dem + - Phi-4 Sampul Dem 🆕 - [📓] [Using Phi-4-multimodal as a Technology journalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET console application using Phi-4-multimodal to analyze images](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [.NET console application using Phi-4-multimodal to analyze pictures](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Fine-tuning Phi Samples +- Fine-tuning Phi Sampul Dem - [Fine-tuning Scenarios](./md/03.FineTuning/FineTuning_Scenarios.md) - [Fine-tuning vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - [Fine-tuning Make Phi-3 become industry expert](./md/03.FineTuning/LetPhi3gotoIndustriy.md) @@ -203,20 +205,21 @@ Make you follow these steps to start use these resource: - [Fine-tuning with Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Fine-tuning with Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - [Fine-tuning Phi-3-vision with Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Fine-tuning Phi-3 with Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + + - [Fine-tuning Phi-3 wit Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - [Fine-tuning Phi-3-vision (official support)](./md/03.FineTuning/FineTuning_Vision.md) - [Fine-Tuning Phi-3 wit Kaito AKS , Azure Containers(official Support)](./md/03.FineTuning/FineTuning_Kaito.md) - [Fine-Tuning Phi-3 and 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Hands on Lab - - [Explorin di latest models: LLMs, SLMs, local development and more](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Exploring cutting-edge models: LLMs, SLMs, local development and more](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - [Unlocking NLP Potential: Fine-Tuning wit Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - Academic Research Papers and Publications - [Textbooks Na All You Need II: phi-1.5 technical report](https://arxiv.org/abs/2309.05463) - - [Phi-3 Technical Report: One Big Language Model Wey Dey inside Your Phone](https://arxiv.org/abs/2404.14219) + - [Phi-3 Technical Report: A Highly Capable Language Model Locally for Your Phone](https://arxiv.org/abs/2404.14219) - [Phi-4 Technical Report](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini Technical Report: Small but Strong Multimodal Language Models Through Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Phi-4-Mini Technical Report: Compact but Powerful Multimodal Language Models via Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - [Optimizing Small Language Models for In-Vehicle Function-Calling](https://arxiv.org/abs/2501.02342) - [(WhyPHI) Fine-Tuning PHI-3 for Multiple-Choice Question Answering: Methodology, Results, and Challenges](https://arxiv.org/abs/2501.01588) - [Phi-4-reasoning Technical Report](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) @@ -226,21 +229,21 @@ Make you follow these steps to start use these resource: ### Phi on Microsoft Foundry -You fit learn how to use Microsoft Phi and how to build E2E solutions for your different hardware devices. To try Phi by yourself, start by playing wit di models and customize Phi for your own wahala using di [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) you fit learn more at Getting Started wit [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +You fit learn how to use Microsoft Phi and how to build E2E solutions for your different hardware dem. To experience Phi yourself, start by to dey play wit the models and customize Phi for your own scenarios using the [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) you fit learn more for Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) **Playground** -Each model get im own playground to test am [Azure AI Playground](https://aka.ms/try-phi3). +Every model get im own playground to test the model [Azure AI Playground](https://aka.ms/try-phi3). ### Phi on GitHub Models -You fit learn how to use Microsoft Phi and how to build E2E solutions for your different hardware devices. To try Phi by yourself, start by playing wit di model and customize Phi for your own wahala using di [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) you fit learn more at Getting Started wit [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +You fit learn how to use Microsoft Phi and how to build E2E solutions for your different hardware dem. To experience Phi yourself, start by to dey play wit the model and customize Phi for your own scenarios using the [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) you fit learn more for Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) **Playground** -Each model get im own [playground to test di model](/md/02.QuickStart/GitHubModel_QuickStart.md). +Every model get im own [playground to test the model](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi on Hugging Face -You fit still find di model on di [Hugging Face](https://huggingface.co/microsoft) +You fit also find the model for [Hugging Face](https://huggingface.co/microsoft) **Playground** [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) @@ -291,34 +294,36 @@ Our team dey produce other courses! Check am out: ## Responsible AI -Microsoft dey committed to help our customers use our AI products responsibly, share our learnings, and build trust-based partnerships through tools like Transparency Notes and Impact Assessments. Plenty of these resources dey for [https://aka.ms/RAI](https://aka.ms/RAI). -Microsoft approach to responsible AI na based on our AI principles of fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability. +Microsoft dey committed to help our customers use our AI products responsible, dey share our learnings, and dey build trust-based partnerships through tools like Transparency Notes and Impact Assessments. Plenti of these resources you fit find for [https://aka.ms/RAI](https://aka.ms/RAI). +Microsoft approach to responsible AI base on our AI principles of fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability. + +Big-big natural language, image, and speech models - like the one wey dem use for this sample - fit behave sometimes for way wey no fair, unreliable, or offensive, and fit cause wahala. Abeg check the [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) make you sabi the risks and limitations. + -Big natural language, image, and speech models - like di ones wey dem use for dis sample - fit behave in ways wey no too fair, no reliable, or fit offend people, wey fit cause wahala. Abeg check di [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) to sabi all di risks and limits. -Di recommended way wey fit help reduce dese risks na to put one safety system for your architecture wey fit detect and stop bad behavior. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) dey provide one independent protection layer wey fit detect bad content wey users or AI generate for different apps and services. Azure AI Content Safety get text and image APIs wey make you fit detect harmful material. Inside Microsoft Foundry, di Content Safety service go allow you see, explore and try sample code wey fit detect bad content for different type modality. Di following [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) go guide you how to make requests to di service. +Di recommended way to reduce dese risks na to add safety system for your architecture wey fit detect and stop bad behavior. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) dey provide independent layer of protection, wey fit detect bad user-generated and AI-generated content for apps and services. Azure AI Content Safety get text and image APIs wey allow you detect harmful matter. Inside Microsoft Foundry, Content Safety service allow you see, explore and try sample code to detect bad content for different modalities. Dis [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) go guide you on how to make requests to di service. -Another tin to consider na di general application performance. For multi-modal and multi-models applications, we mean say di system go work as you and your users expect, including say e no go generate bad outputs. E important to check di performance of your whole application using [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). You fit also create and evaluate using [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Another tin to consider na di overall application performance. With multi-modal and multi-models applications, we mean say di system dey work as you and your users dey expect, including not to produce bad outputs. E dey important to check how your overall application dey perform by using [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). You fit also create and check with [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -You fit evaluate your AI application inside your development environment using di [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). If you get test dataset or target, your generative AI application generations go get quantitative measurement with built-in evaluators or custom evaluators wey you pick. To start with di azure ai evaluation sdk to evaluate your system, you fit follow di [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). After you run evaluation, you fit [visualize di results in Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +You fit check your AI application for your development environment using di [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Whether na test dataset or target, your generative AI application generations go get quantitative measure with built-in evaluators or custom evaluators wey you choose. To start to use di azure ai evaluation sdk to check your system, you fit follow di [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). After you run evaluation, you fit [visualize di results for Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Trademarks -Dis project fit get trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos mean say you gats follow [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -If you dey use Microsoft trademarks or logos for changed versions of dis project, e no suppose cause confusion or make person think say Microsoft sponsor am. Any use of third-party trademarks or logos go dey under those third-party policies. +Dis project fit get trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos dey bajo condition and you gots follow [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Use of Microsoft trademarks or logos for modified versions of dis project no suppose cause confusion or make people reason say Microsoft sponsor am. Any use of third-party trademarks or logos dey subject to dem third-party policy. ## Getting Help -If you jam wahala or get any question about how to build AI apps, join: +If you jam wahala or get any questions about building AI apps, come join: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -If you get product feedback or error while you dey build, visit: +If you get product feedback or errors while you dey build, visit: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Disclaimer**: -Dis dokment don translate wit AI translation service [Co-op Translator](https://github.com/Azure/co-op-translator). Even though we dey try make am correct, abeg sabi say automated translation fit get errors or mistakes. Di original dokment wey e dey for im own language na di correct source. For important matter, make person wey sabi translate am well translate am. We no go carry any yawa wey fit happen if pesin no understand well or interpret am wrong because of dis translation. +**Disclaimer**: +Dis document don translate wit AI translation service [Co-op Translator](https://github.com/Azure/co-op-translator). Even tho we dey try make am correct, abeg make you know say automated translation fit get errors or mistakes. Di original document for dia own language na im be di correct source. For important info, make person wey sabi human translation do am. We no go responsible for any misunderstanding or wrong understanding wey fit happen because of dis translation. \ No newline at end of file diff --git a/translations/pcm/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/pcm/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..9b47b08bc --- /dev/null +++ b/translations/pcm/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrid RAG wit Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Author:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Focus:** Small Language Models (SLMs), SQLite FTS5 BM25, Dense Embeddings, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivation: Di Keyword Recall Wahala for Local SLMs\n", + "Standard RAG architectures wey depend fully on dense vector embeddings dey often fail for get exact numerical tokens (like `2,340,000 TL`, contract codes, account numbers). \n", + "For di oda hand, sparse lexical search (BM25) no dey catch semantic synonyms and paraphrased questions.\n", + "\n", + "Dis cookbook show how to implement **high-speed, in-memory Hybrid Retrieval engine** wey combine:\n", + "1. **Dense Vectors** (Cosine Similarity)\n", + "2. **Sparse Lexical Search** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Grounded Citation Generation (`[1]`, `[2]`)** wit Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Dual SQLite Schema (Dense Vectors + Virtual FTS5 BM25 Table)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Sample Ingestion & Execution Benchmark\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Grounded Prompt Formulation for Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Disclaimer**:\nDis document don translate wit AI translation service [Co-op Translator](https://github.com/Azure/co-op-translator). Even tho we dey try make am correct, abeg make you know say automated translation fit get errors or mistakes. Di original document for dia own language na im be di correct source. For important info, make person wey sabi human translation do am. We no go responsible for any misunderstanding or wrong understanding wey fit happen because of dis translation.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/pl/.co-op-translator.json b/translations/pl/.co-op-translator.json index 69e002788..cc00452f7 100644 --- a/translations/pl/.co-op-translator.json +++ b/translations/pl/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "pl" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:50:41+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:03:33+00:00", "source_file": "README.md", "language_code": "pl" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "pl" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:06:59+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "pl" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:48+00:00", diff --git a/translations/pl/README.md b/translations/pl/README.md index c8fd80e6f..9592378bf 100644 --- a/translations/pl/README.md +++ b/translations/pl/README.md @@ -1,42 +1,42 @@ # Phi Cookbook: Praktyczne przykłady z modelami Phi firmy Microsoft -[![Otwórz i korzystaj z przykładów w GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Otwórz i używaj przykładów w GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Otwórz w Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![Współtwórcy GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![Zgłoszenia GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![Problemy GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) [![Pull requesty GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![Obserwujący GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![Forki GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![Rozwidlenia GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![Gwiazdki GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) Phi to seria otwartych modeli AI opracowanych przez Microsoft. -Phi jest obecnie najpotężniejszym i najbardziej opłacalnym małym modelem językowym (SLM), z bardzo dobrymi wynikami w wielu językach, wnioskowaniu, generowaniu tekstu/czatów, kodowaniu, obrazach, dźwięku i innych scenariuszach. +Phi jest obecnie najsilniejszym i najbardziej opłacalnym modelem małego języka (SLM), z bardzo dobrymi benchmarkami w wielojęzyczności, rozumowaniu, generowaniu tekstu/czatu, kodowaniu, obrazach, dźwięku i innych scenariuszach. -Możesz wdrożyć Phi w chmurze lub na urządzeniach edge i łatwo tworzyć aplikacje generatywnego AI przy ograniczonej mocy obliczeniowej. +Możesz wdrożyć Phi w chmurze lub na urządzeniach brzegowych i łatwo tworzyć aplikacje generatywnej AI przy ograniczonej mocy obliczeniowej. -Wykonaj poniższe kroki, aby rozpocząć korzystanie z tych zasobów: -1. **Rozgałęź repozytorium**: Kliknij [![Forki GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +Postępuj zgodnie z tymi krokami, aby rozpocząć korzystanie z tych zasobów: +1. **Utwórz fork repozytorium**: Kliknij [![Rozwidlenia GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Sklonuj repozytorium**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Dołącz do Microsoft AI Discord Community i spotkaj ekspertów oraz innych deweloperów**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Dołącz do społeczności Microsoft AI na Discordzie i spotkaj ekspertów i innych deweloperów**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/pl/cover.eb18d1b9605d754b.webp) ### 🌐 Wsparcie wielojęzyczne -#### Wspierane przez GitHub Action (zautomatyzowane i zawsze aktualne) +#### Obsługiwane przez GitHub Action (Automatyczne i zawsze aktualne) -[Arabski](../ar/README.md) | [Bengalski](../bn/README.md) | [Bułgarski](../bg/README.md) | [Birmański (Myanmar)](../my/README.md) | [Chiński (uproszczony)](../zh-CN/README.md) | [Chiński (tradycyjny, Hongkong)](../zh-HK/README.md) | [Chiński (tradycyjny, Makau)](../zh-MO/README.md) | [Chiński (tradycyjny, Tajwan)](../zh-TW/README.md) | [Chorwacki](../hr/README.md) | [Czeski](../cs/README.md) | [Duński](../da/README.md) | [Holenderski](../nl/README.md) | [Estoński](../et/README.md) | [Fiński](../fi/README.md) | [Francuski](../fr/README.md) | [Niemiecki](../de/README.md) | [Grecki](../el/README.md) | [Hebrajski](../he/README.md) | [Hindi](../hi/README.md) | [Węgierski](../hu/README.md) | [Indonezyjski](../id/README.md) | [Włoski](../it/README.md) | [Japoński](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreański](../ko/README.md) | [Litewski](../lt/README.md) | [Malajski](../ms/README.md) | [Malajalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalski](../ne/README.md) | [Nigeryjski Pidgin](../pcm/README.md) | [Norweski](../no/README.md) | [Perski (Farsi)](../fa/README.md) | [Polski](./README.md) | [Portugalski (Brazylia)](../pt-BR/README.md) | [Portugalski (Portugalia)](../pt-PT/README.md) | [Pendżabski (Gurmukhi)](../pa/README.md) | [Rumuński](../ro/README.md) | [Rosyjski](../ru/README.md) | [Serbski (cyrylica)](../sr/README.md) | [Słowacki](../sk/README.md) | [Słoweński](../sl/README.md) | [Hiszpański](../es/README.md) | [Suahili](../sw/README.md) | [Szwedzki](../sv/README.md) | [Tagalog (Filipiński)](../tl/README.md) | [Tamilski](../ta/README.md) | [Telugu](../te/README.md) | [Tajski](../th/README.md) | [Turecki](../tr/README.md) | [Ukraiński](../uk/README.md) | [Urdu](../ur/README.md) | [Wietnamski](../vi/README.md) +[Arabski](../ar/README.md) | [Bengalski](../bn/README.md) | [Bułgarski](../bg/README.md) | [Birmański (Myanmar)](../my/README.md) | [Chiński (uproszczony)](../zh-CN/README.md) | [Chiński (tradycyjny, Hong Kong)](../zh-HK/README.md) | [Chiński (tradycyjny, Makau)](../zh-MO/README.md) | [Chiński (tradycyjny, Tajwan)](../zh-TW/README.md) | [Chorwacki](../hr/README.md) | [Czeski](../cs/README.md) | [Duński](../da/README.md) | [Niderlandzki](../nl/README.md) | [Estoński](../et/README.md) | [Fiński](../fi/README.md) | [Francuski](../fr/README.md) | [Niemiecki](../de/README.md) | [Grecki](../el/README.md) | [Hebrajski](../he/README.md) | [Hindi](../hi/README.md) | [Węgierski](../hu/README.md) | [Indonezyjski](../id/README.md) | [Włoski](../it/README.md) | [Japoński](../ja/README.md) | [Kannada](../kn/README.md) | [Khmerski](../km/README.md) | [Koreański](../ko/README.md) | [Litewski](../lt/README.md) | [Malajski](../ms/README.md) | [Malajalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalski](../ne/README.md) | [Pidgin Nigeryjski](../pcm/README.md) | [Norweski](../no/README.md) | [Perski (Farsi)](../fa/README.md) | [Polski](./README.md) | [Portugalski (Brazylia)](../pt-BR/README.md) | [Portugalski (Portugalia)](../pt-PT/README.md) | [Pendżabski (Gurmukhi)](../pa/README.md) | [Rumuński](../ro/README.md) | [Rosyjski](../ru/README.md) | [Serbski (cyrylica)](../sr/README.md) | [Słowacki](../sk/README.md) | [Słoweński](../sl/README.md) | [Hiszpański](../es/README.md) | [Suahili](../sw/README.md) | [Szwedzki](../sv/README.md) | [Tagalog (Filipiński)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Tajski](../th/README.md) | [Turecki](../tr/README.md) | [Ukraiński](../uk/README.md) | [Urdu](../ur/README.md) | [Wietnamski](../vi/README.md) > **Wolisz klonować lokalnie?** > -> To repozytorium zawiera ponad 50 tłumaczeń, co znacząco zwiększa rozmiar pobierania. Aby sklonować bez tłumaczeń, użyj sparse checkout: +> To repozytorium zawiera ponad 50 tłumaczeń językowych, co znacznie zwiększa rozmiar pobierania. Aby sklonować bez tłumaczeń, użyj sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,7 +52,7 @@ Wykonaj poniższe kroki, aby rozpocząć korzystanie z tych zasobów: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Dzięki temu zyskasz wszystko, co potrzebne do ukończenia kursu, przy znacznie szybszym pobieraniu. +> To daje wszystko, czego potrzebujesz, aby ukończyć kurs z dużo szybszym pobieraniem. ## Spis treści @@ -61,130 +61,132 @@ Wykonaj poniższe kroki, aby rozpocząć korzystanie z tych zasobów: - [Witamy w rodzinie Phi](./md/01.Introduction/01/01.PhiFamily.md) - [Konfiguracja środowiska](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Zrozumienie kluczowych technologii](./md/01.Introduction/01/01.Understandingtech.md) - - [Bezpieczeństwo AI w modelach Phi](./md/01.Introduction/01/01.AISafety.md) - - [Wsparcie sprzętowe Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Modele Phi i ich dostępność na platformach](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Używanie Guidance-ai i Phi](./md/01.Introduction/01/01.Guidance.md) - - [Modele na GitHub Marketplace](https://github.com/marketplace/models) - - [Katalog modeli AI Azure](https://ai.azure.com) + - [Bezpieczeństwo AI dla modeli Phi](./md/01.Introduction/01/01.AISafety.md) + - [Wsparcie sprzętu Phi](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Modele Phi i dostępność na różnych platformach](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Korzystanie z Guidance-ai i Phi](./md/01.Introduction/01/01.Guidance.md) + - [Modele w GitHub Marketplace](https://github.com/marketplace/models) + - [Katalog modeli Azure AI](https://ai.azure.com) - Wnioskowanie Phi w różnych środowiskach - [Hugging face](./md/01.Introduction/02/01.HF.md) - [Modele GitHub](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry Model Catalog](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Katalog modeli Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [Narzędzia AI VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [Zestaw narzędzi AI VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - -- Rodzina Wnioskowania Phi - - [Wnioskowanie Phi na iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Wnioskowanie Phi na Android](./md/01.Introduction/03/Android_Inference.md) - - [Wnioskowanie Phi na Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Wnioskowanie Phi na AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Wnioskowanie Phi z wykorzystaniem Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Wnioskowanie Phi na serwerze lokalnym](./md/01.Introduction/03/Local_Server_Inference.md) - - [Wnioskowanie Phi na serwerze zdalnym przy użyciu AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Foundry lokalnie](./md/01.Introduction/02/07.FoundryLocal.md) + +- Wnioskowanie Phi Family + - [Wnioskowanie Phi w iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Wnioskowanie Phi w Android](./md/01.Introduction/03/Android_Inference.md) + - [Wnioskowanie Phi w Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Wnioskowanie Phi w AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Wnioskowanie Phi z użyciem Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Wnioskowanie Phi na lokalnym serwerze](./md/01.Introduction/03/Local_Server_Inference.md) + - [Wnioskowanie Phi na serwerze zdalnym za pomocą AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - [Wnioskowanie Phi z Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Wnioskowanie Phi--Vision lokalnie](./md/01.Introduction/03/Vision_Inference.md) + - [Wnioskowanie Phi-Vision lokalnie](./md/01.Introduction/03/Vision_Inference.md) - [Wnioskowanie Phi z Kaito AKS, Azure Containers (oficjalne wsparcie)](./md/01.Introduction/03/Kaito_Inference.md) -- [Kwantyzacja rodziny Phi](./md/01.Introduction/04/QuantifyingPhi.md) - - [Kwantyzacja Phi-3.5 / 4 za pomocą llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Kwantyzacja Phi-3.5 / 4 z rozszerzeniami generatywnego AI dla onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Kwantyzacja Phi-3.5 / 4 za pomocą Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Kwantyzacja Phi-3.5 / 4 za pomocą Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Ocena Phi - - [AI odpowiedzialne](./md/01.Introduction/05/ResponsibleAI.md) - - [Microsoft Foundry do oceny](./md/01.Introduction/05/AIFoundry.md) - - [Użycie Promptflow do oceny](./md/01.Introduction/05/Promptflow.md) +- [Quantifying Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) + - [Quantizing Phi-3.5 / 4 za pomocą llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Quantizing Phi-3.5 / 4 przy użyciu rozszerzeń Generative AI dla onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Quantizing Phi-3.5 / 4 za pomocą Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Quantizing Phi-3.5 / 4 z użyciem Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- Ewaluacja Phi + - [Odpowiedzialna AI](./md/01.Introduction/05/ResponsibleAI.md) + - [Microsoft Foundry do ewaluacji](./md/01.Introduction/05/AIFoundry.md) + - [Korzystanie z Promptflow do ewaluacji](./md/01.Introduction/05/Promptflow.md) - RAG z Azure AI Search - - [Jak używać Phi-4-mini i Phi-4-multimodal(RAG) z Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Jak używać Phi-4-mini i Phi-4-multimodal (RAG) z Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Chmura Lokalny Hybrid RAG z SQLite FTS5 i phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Przykłady tworzenia aplikacji Phi - - Aplikacje tekstowe i czatowe +- Przykłady rozwoju aplikacji Phi + - Aplikacje Tekstowe i Czat - Przykłady Phi-4 - [📓] [Czat z modelem Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Czat z lokalnym modelem Phi-4 ONNX w .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Konsolowa aplikacja czatu .NET z Phi-4 ONNX używająca Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Przykłady Phi-3 / 3.5 - - [Lokalny chatbot w przeglądarce korzystający z Phi3, ONNX Runtime Web i WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Czat z lokalnym modelem Phi-4 ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Czat w aplikacji konsolowej .NET z Phi-4 ONNX używając Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Przykłady + - [Lokalny chatbot w przeglądarce używający Phi3, ONNX Runtime Web i WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [Multi Model - Interaktywny Phi-3-mini i OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Tworzenie wrappera i używanie Phi-3 z MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Optymalizacja Modelu - Jak zoptymalizować model Phi-3-min na ONNX Runtime Web z Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Optymalizacja modelu - Jak zoptymalizować model Phi-3-mini dla ONNX Runtime Web za pomocą Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [Aplikacja WinUI3 z Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Multi Model AI Powered Notes App Sample](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Dostrajanie i integracja niestandardowych modeli Phi-3 z Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Dostrajanie i integracja niestandardowych modeli Phi-3 z Prompt flow w Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Ocena dostrojonego modelu Phi-3 / Phi-3.5 w Microsoft Foundry z naciskiem na zasady odpowiedzialnej AI Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Przykład przewidywania języka Phi-3.5-mini-instruct (chiński/angielski)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + -[Przykład aplikacji WinUI3 Multi Model Notes zasilane SI](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Dostrajanie i integracja własnych modeli Phi-3 z Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Dostrajanie i integracja własnych modeli Phi-3 z Prompt flow w Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Ocena dostrojonego modelu Phi-3 / Phi-3.5 w Microsoft Foundry z naciskiem na zasady Odpowiedzialnej Sztucznej Inteligencji Microsoftu](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Przykład predykcji języka Phi-3.5-mini-instruct (chiński/angielski)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Używanie GPU Windows do tworzenia rozwiązania Prompt flow z Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Używanie Microsoft Phi-3.5 tflite do tworzenia aplikacji na Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Wykorzystanie GPU Windows do stworzenia rozwiązania Prompt flow z Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Używanie Microsoft Phi-3.5 tflite do stworzenia aplikacji na Androida](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - [Przykład Q&A .NET używający lokalnego modelu ONNX Phi-3 z Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Konsolowa aplikacja chat .NET z Semantic Kernel i Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Konsolowa aplikacja czatu .NET z Semantic Kernel i Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Przykłady kodu na bazie Azure AI Inference SDK - - Przykłady Phi-4 - - [📓] [Generowanie kodu projektu używając Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Przykłady SDK Azure AI Inference oparte na kodzie + - Przykłady Phi-4 + - [📓] [Generowanie kodu projektu przy użyciu Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Przykłady Phi-3 / 3.5 - - [Zbuduj własny Visual Studio Code GitHub Copilot Chat z rodziną Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Stwórz własnego agenta chat Copilot dla Visual Studio Code z Phi-3.5 za pomocą modeli GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Stwórz własnego Visual Studio Code GitHub Copilot Chat z Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Stwórz własnego agenta Visual Studio Code Chat Copilot z Phi-3.5 korzystając z modeli GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - Zaawansowane przykłady rozumowania - - Przykłady Phi-4 - - [📓] [Przykłady Phi-4-mini-reasoning lub Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - Przykłady zaawansowanego rozumowania + - Przykłady Phi-4 + - [📓] [Phi-4-mini-reasoning lub Phi-4-reasoning Przykłady](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - [📓] [Dostrajanie Phi-4-mini-reasoning z Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Dostrajanie Phi-4-mini-reasoning z Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning z modelami z GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-reasoning z modelami GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-reasoning z modelami Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Dema - - [Dema Phi-4-mini hostowane na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Dema Phi-4-multimodal hostowane na Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Przykłady dla wizji - - Przykłady Phi-4 - - [📓] [Użyj Phi-4-multimodal do odczytu obrazów i generowania kodu](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [Phi-4-mini demo hostowane na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal demo hostowane na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Przykłady wizji + - Przykłady Phi-4 + - [📓] [Użycie Phi-4-multimodal do czytania obrazów i generowania kodu](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Przykłady Phi-3 / 3.5 - - [📓][Phi-3-vision-Obraz tekst na tekst](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-Obraz tekst do tekstu](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [DEMO: Phi-3 Recycling](https://github.com/jennifermarsman/PhiRecycling/) - [Phi-3-vision - wizualny asystent językowy - z Phi3-Vision i OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision próba wielu ramek lub wielu obrazów](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Lokalny model Phi-3 Vision ONNX z Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menu do lokalnego modelu Phi-3 Vision ONNX z Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Phi-3.5 Vision multi-frame lub multi-image sample](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision Lokalny model ONNX z użyciem Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Menu oparte Phi-3 Vision Lokalny model ONNX z użyciem Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - Przykłady Reasoning-Vision - - Phi-4-Reasoning-Vision-15B - - [📓] [Użycie Phi-4-Reasoning-Vision-15B do wykrywania przechodzenia na czerwonym](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - Phi-4-Reasoning-Vision-15B + - [📓] [Użycie Phi-4-Reasoning-Vision-15B do wykrywania nieprawidłowego przechodzenia przez ulicę](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [Użycie Phi-4-Reasoning-Vision-15B do matematyki](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - [📓] [Użycie Phi-4-Reasoning-Vision-15B do wykrywania UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Przykłady matematyczne - - Przykłady Phi-4-Mini-Flash-Reasoning-Instruct [Demo matematyczne z Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Przykłady Phi-4-Mini-Flash-Reasoning-Instruct [Demo matematyczne z Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - Przykłady audio - - Przykłady Phi-4 - - [📓] [Wyodrębnianie transkryptów audio z Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - Przykłady Phi-4 + - [📓] [Ekstrakcja transkryptów audio przy użyciu Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Przykład audio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Przykład tłumaczenia mowy Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [Konsolowa aplikacja .NET używająca Phi-4-multimodal Audio do analizy pliku audio i generowania transkryptu](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [Konsolowa aplikacja .NET używająca Phi-4-multimodal audio do analizowania pliku audio i generowania transkryptu](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - Przykłady MOE - Przykłady Phi-3 / 3.5 - - [📓] [Modele mieszaniny ekspertów Phi-3.5 (MoEs) - przykład dla mediów społecznościowych](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Phi-3.5 Modele Mieszanki Ekspertów (MoEs) przykład mediów społecznościowych](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - [📓] [Budowanie pipeline Retrieval-Augmented Generation (RAG) z NVIDIA NIM Phi-3 MOE, Azure AI Search i LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Przykłady wywoływania funkcji + - Przykłady wywołań funkcji - Przykłady Phi-4 🆕 - - [📓] [Użycie wywoływania funkcji z Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Użycie wywoływania funkcji do tworzenia multi-agentów z Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Użycie wywoływania funkcji z Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Użycie wywoływania funkcji z ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - [📓] [Użycie wywołań funkcji z Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Użycie wywołań funkcji do tworzenia multi-agentów z Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Użycie wywołań funkcji z Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Użycie wywołań funkcji z ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - Przykłady mieszania multimodalnego - Przykłady Phi-4 🆕 - [📓] [Użycie Phi-4-multimodal jako dziennikarza technologicznego](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) @@ -193,7 +195,7 @@ Wykonaj poniższe kroki, aby rozpocząć korzystanie z tych zasobów: - Dostrajanie próbek Phi - [Scenariusze dostrajania](./md/03.FineTuning/FineTuning_Scenarios.md) - [Dostrajanie vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Dostrajanie: Niech Phi-3 stanie się ekspertem branżowym](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Dostrajanie pozwalające Phi-3 zostać ekspertem branżowym](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - [Dostrajanie Phi-3 z AI Toolkit dla VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [Dostrajanie Phi-3 z Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - [Dostrajanie Phi-3 z Lora](./md/03.FineTuning/FineTuning_Lora.md) @@ -203,22 +205,23 @@ Wykonaj poniższe kroki, aby rozpocząć korzystanie z tych zasobów: - [Dostrajanie z Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Dostrajanie z Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - [Dostrajanie Phi-3-vision z Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Dostrajanie Phi-3 z Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Dostrajanie Phi-3-vision (oficjalne wsparcie)](./md/03.FineTuning/FineTuning_Vision.md) - - [Dostrajanie Phi-3 z Kaito AKS, Azure Containers (oficjalne wsparcie)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Dostrajanie Phi-3 i 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) -- Warsztaty praktyczne + - [Dostrojenie Phi-3 z wykorzystaniem Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Dostrojenie Phi-3-vision (oficjalne wsparcie)](./md/03.FineTuning/FineTuning_Vision.md) + - [Dostrojenie Phi-3 z Kaito AKS, Azure Containers (oficjalne wsparcie)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Dostrojenie Phi-3 i 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + +- Laboratorium praktyczne - [Eksploracja najnowocześniejszych modeli: LLM, SLM, lokalny rozwój i więcej](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Odblokowanie potencjału NLP: dostrajanie z Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Odkrywanie potencjału NLP: dostrajanie z Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Prace badawcze i publikacje akademickie +- Publikacje naukowe i artykuły - [Textbooks Are All You Need II: raport techniczny phi-1.5](https://arxiv.org/abs/2309.05463) - - [Raport techniczny Phi-3: wysoce zaawansowany model językowy lokalnie na Twoim telefonie](https://arxiv.org/abs/2404.14219) + - [Raport techniczny Phi-3: Wysoce zdolny model językowy lokalnie na twoim telefonie](https://arxiv.org/abs/2404.14219) - [Raport techniczny Phi-4](https://arxiv.org/abs/2412.08905) - - [Raport techniczny Phi-4-Mini: kompaktowe, lecz potężne multimodalne modele językowe z użyciem Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Raport techniczny Phi-4-Mini: kompaktowe lecz potężne multimodalne modele językowe dzięki mieszance LoRAs](https://arxiv.org/abs/2503.01743) - [Optymalizacja małych modeli językowych do wywoływania funkcji w pojeździe](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Dostrajanie PHI-3 do odpowiadania na pytania wielokrotnego wyboru: metodologia, wyniki i wyzwania](https://arxiv.org/abs/2501.01588) + - [(WhyPHI) Dostrajanie PHI-3 do pytań wielokrotnego wyboru: metodologia, wyniki i wyzwania](https://arxiv.org/abs/2501.01588) - [Raport techniczny Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Raport techniczny Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) @@ -226,28 +229,28 @@ Wykonaj poniższe kroki, aby rozpocząć korzystanie z tych zasobów: ### Phi na Microsoft Foundry -Możesz nauczyć się, jak korzystać z Microsoft Phi i jak budować kompleksowe rozwiązania na różnych urządzeniach sprzętowych. Aby osobiście doświadczyć Phi, zacznij od zabawy z modelami i dostosowywania Phi do swoich scenariuszy za pomocą [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Więcej informacji znajdziesz w przewodniku Rozpoczęcie pracy z [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md). +Możesz nauczyć się, jak korzystać z Microsoft Phi oraz jak budować rozwiązania E2E na różnych urządzeniach sprzętowych. Aby samodzielnie doświadczyć Phi, zacznij od eksperymentowania z modelami i dostosowywania Phi do swoich scenariuszy za pomocą [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Więcej dowiesz się w sekcji Rozpoczęcie pracy z [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Plac zabaw** +**Plac zabaw** Każdy model ma dedykowany plac zabaw do testowania modelu [Azure AI Playground](https://aka.ms/try-phi3). ### Phi na modelach GitHub -Możesz nauczyć się, jak korzystać z Microsoft Phi i jak budować kompleksowe rozwiązania na różnych urządzeniach sprzętowych. Aby osobiście doświadczyć Phi, zacznij od zabawy z modelem i dostosowywania Phi do swoich scenariuszy za pomocą [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Więcej informacji znajdziesz w przewodniku Rozpoczęcie pracy z [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md). +Możesz nauczyć się, jak korzystać z Microsoft Phi oraz jak budować rozwiązania E2E na różnych urządzeniach sprzętowych. Aby samodzielnie doświadczyć Phi, zacznij od eksperymentowania z modelem i dostosowywania Phi do swoich scenariuszy za pomocą [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Więcej dowiesz się w sekcji Rozpoczęcie pracy z [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Plac zabaw** +**Plac zabaw** Każdy model ma dedykowany [plac zabaw do testowania modelu](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi na Hugging Face -Model jest również dostępny na [Hugging Face](https://huggingface.co/microsoft). +Możesz także znaleźć model na [Hugging Face](https://huggingface.co/microsoft) -**Plac zabaw** -[Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Plac zabaw** + [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Inne kursy + ## 🎒 Inne kursy -Nasz zespół tworzy też inne kursy! Sprawdź: +Nasz zespół tworzy inne kursy! Sprawdź: ### LangChain @@ -260,59 +263,61 @@ Nasz zespół tworzy też inne kursy! Sprawdź: [![AZD dla początkujących](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI dla początkujących](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP dla początkujących](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Agenci AI dla początkujących](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agenci dla początkujących](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Seria Generative AI + +### Seria Generatywnej AI [![Generatywna AI dla początkujących](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Generatywna AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) [![Generatywna AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) [![Generatywna AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Podstawowa nauka + +### Podstawowe nauki [![ML dla początkujących](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Data Science dla początkujących](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI dla początkujących](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cyberbezpieczeństwo dla początkujących](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Cybersecurity dla początkujących](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Web Dev dla początkujących](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![IoT dla początkujących](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![Rozwój XR dla początkujących](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![XR Development dla początkujących](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Seria Copilot -[![Copilot dla programowania we współpracy z AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot dla AI Pair Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot dla C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Odpowiedzialne AI +## Odpowiedzialna AI + +Microsoft angażuje się w pomoc klientom w odpowiedzialnym korzystaniu z naszych produktów AI, dzielenie się naszymi doświadczeniami oraz budowanie partnerstw opartych na zaufaniu za pomocą narzędzi takich jak Transparency Notes i Impact Assessments. Wiele z tych zasobów można znaleźć na [https://aka.ms/RAI](https://aka.ms/RAI). +Podejście Microsoft do odpowiedzialnej AI opiera się na naszych zasadach AI: uczciwość, niezawodność i bezpieczeństwo, prywatność i ochrona, inkluzywność, przejrzystość oraz odpowiedzialność. + +Duże modele przetwarzania języka naturalnego, obrazu i mowy – takie jak użyte w tym przykładzie – mogą potencjalnie zachowywać się w sposób nieuczciwy, zawodny lub obraźliwy, co może prowadzić do szkód. Zapoznaj się z [notatką transparentności usługi Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), aby być świadomym ryzyk i ograniczeń. -Microsoft zobowiązuje się do pomocy klientom w odpowiedzialnym korzystaniu z naszych produktów AI, dzielenia się naszymi doświadczeniami oraz budowania partnerstw opartych na zaufaniu przy pomocy narzędzi takich jak Transparency Notes i Impact Assessments. Wiele z tych zasobów można znaleźć pod adresem [https://aka.ms/RAI](https://aka.ms/RAI). -Podejście Microsoft do odpowiedzialnej AI opiera się na naszych zasadach AI: uczciwości, niezawodności i bezpieczeństwie, prywatności i ochronie, integracyjności, przejrzystości oraz odpowiedzialności. -Duże modele językowe, obrazowe i mowy - takie jak te używane w tym przykładzie - mogą potencjalnie zachowywać się w sposób nieuczciwy, niespójny lub obraźliwy, co może powodować szkody. Prosimy o zapoznanie się z [Notatką transparentności usługi Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), aby zostać poinformowanym o ryzykach i ograniczeniach. -Zalecanym podejściem do łagodzenia tych ryzyk jest uwzględnienie w architekturze systemu bezpieczeństwa, który potrafi wykrywać i zapobiegać szkodliwemu zachowaniu. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) zapewnia niezależną warstwę ochrony, zdolną do wykrywania szkodliwych treści generowanych przez użytkowników i AI w aplikacjach i usługach. Azure AI Content Safety obejmuje API do analizy tekstu i obrazów, które pozwalają wykrywać materiały szkodliwe. W ramach Microsoft Foundry, usługa Content Safety umożliwia przeglądanie, eksplorowanie oraz wypróbowywanie przykładowego kodu do wykrywania szkodliwych treści w różnych modalnościach. Następująca [dokumentacja szybkiego startu](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) przeprowadzi Cię przez proces wysyłania zapytań do usługi. +Zalecanym podejściem do łagodzenia tych ryzyk jest uwzględnienie w architekturze systemu bezpieczeństwa, który może wykrywać i zapobiegać szkodliwemu zachowaniu. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) zapewnia niezależną warstwę ochrony, zdolną do wykrywania szkodliwych treści generowanych przez użytkowników i AI w aplikacjach i usługach. Azure AI Content Safety obejmuje API tekstu i obrazu, które pozwalają wykrywać materiały szkodliwe. W Microsoft Foundry usługa Content Safety pozwala na przeglądanie, eksplorowanie i wypróbowywanie przykładowego kodu do wykrywania szkodliwych treści w różnych trybach. Następująca [dokumentacja szybkiego startu](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) przeprowadzi Cię przez proces wykonywania żądań do usługi. -Innym aspektem, który należy uwzględnić, jest ogólna wydajność aplikacji. W przypadku aplikacji multimodalnych i wielomodelowych wydajność oznacza, że system działa zgodnie z oczekiwaniami twoimi i twoich użytkowników, w tym nie generuje szkodliwych wyników. Ważne jest, aby ocenić wydajność całej aplikacji za pomocą [evaluatorów wydajności, jakości oraz ryzyka i bezpieczeństwa](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Masz również możliwość tworzenia i oceniania za pomocą [evaluatorów niestandardowych](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Kolejnym aspektem, który należy wziąć pod uwagę, jest ogólna wydajność aplikacji. W przypadku aplikacji multimodalnych i z wieloma modelami, wydajność rozumiemy jako to, że system działa zgodnie z oczekiwaniami Twoimi i Twoich użytkowników, włączając w to niegenerowanie szkodliwych wyników. Ważne jest, aby ocenić wydajność całej aplikacji za pomocą [ewaluatorów Wydajności i Jakości oraz Ryzyka i Bezpieczeństwa](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Masz także możliwość tworzenia i oceny za pomocą [niestandardowych ewaluatorów](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Możesz ocenić swoją aplikację AI w środowisku deweloperskim za pomocą [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Mając zestaw danych testowych lub cel, generacje Twojej aplikacji generatywnej AI są mierzone ilościowo za pomocą wbudowanych evaluatorów lub wybranych evaluatorów niestandardowych. Aby zacząć korzystać z azure ai evaluation sdk do oceny Twojego systemu, możesz skorzystać z [przewodnika szybkiego startu](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Po wykonaniu uruchomienia oceny możesz [wizualizować wyniki w Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Możesz ocenić swoją aplikację AI w środowisku programistycznym, używając [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Na podstawie zestawu testowego lub celu, generacje Twojej aplikacji generatywnej AI są ilościowo mierzone za pomocą wbudowanych lub niestandardowych ewaluatorów według wyboru. Aby rozpocząć korzystanie z azure ai evaluation sdk do oceny Twojego systemu, możesz postępować zgodnie z [przewodnikiem szybkiego startu](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Po wykonaniu oceny możesz [wizualizować wyniki w Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Znaki towarowe -Projekt ten może zawierać znaki towarowe lub logotypy projektów, produktów lub usług. Autoryzowane użycie znaków towarowych lub logotypów Microsoft musi podlegać i przestrzegać [Wytycznych dotyczących znaków towarowych i marek Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Użycie znaków towarowych lub logotypów Microsoft w zmodyfikowanych wersjach tego projektu nie może powodować nieporozumień ani sugerować sponsorowania przez Microsoft. Każde użycie znaków towarowych lub logotypów stron trzecich podlega zasadom tych stron trzecich. +Ten projekt może zawierać znaki towarowe lub logotypy projektów, produktów lub usług. Autoryzowane użycie znaków towarowych lub logotypów Microsoft podlega i musi przestrzegać [Zasad dotyczących znaków towarowych i marki Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Użycie znaków towarowych lub logotypów Microsoft w zmodyfikowanych wersjach tego projektu nie może powodować zamieszania ani sugerować sponsorowania przez Microsoft. Każde użycie znaków towarowych lub logotypów stron trzecich podlega politykom tych stron trzecich. ## Uzyskanie pomocy -Jeśli napotkasz problemy lub masz pytania dotyczące tworzenia aplikacji AI, dołącz do: +Jeśli utkniesz lub masz pytania dotyczące tworzenia aplikacji AI, dołącz do: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Jeśli masz opinie o produkcie lub zgłaszasz błędy podczas tworzenia, odwiedź: +Jeśli masz opinie o produkcie lub błędy podczas tworzenia, odwiedź: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) @@ -320,5 +325,5 @@ Jeśli masz opinie o produkcie lub zgłaszasz błędy podczas tworzenia, odwied **Zastrzeżenie**: -Niniejszy dokument został przetłumaczony za pomocą usługi tłumaczenia AI [Co-op Translator](https://github.com/Azure/co-op-translator). Chociaż dokładamy starań, aby tłumaczenie było jak najbardziej precyzyjne, prosimy mieć na uwadze, że automatyczne tłumaczenia mogą zawierać błędy lub niedokładności. Oryginalny dokument w języku źródłowym powinien być uznawany za wiążące źródło informacji. W przypadku krytycznych informacji zalecane jest korzystanie z profesjonalnego tłumaczenia wykonanego przez człowieka. Nie ponosimy odpowiedzialności za jakiekolwiek nieporozumienia lub błędną interpretację wynikające z korzystania z tego tłumaczenia. +Niniejszy dokument został przetłumaczony za pomocą usługi tłumaczenia AI [Co-op Translator](https://github.com/Azure/co-op-translator). Choć dążymy do dokładności, prosimy pamiętać, że automatyczne tłumaczenia mogą zawierać błędy lub niedokładności. Oryginalny dokument w jego języku źródłowym należy uznawać za autorytatywne źródło. W przypadku informacji krytycznych zalecane jest skorzystanie z profesjonalnego tłumaczenia wykonanego przez człowieka. Nie ponosimy odpowiedzialności za jakiekolwiek nieporozumienia lub błędne interpretacje wynikające z użycia tego tłumaczenia. \ No newline at end of file diff --git a/translations/pl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/pl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..5bad87790 --- /dev/null +++ b/translations/pl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrydowy RAG z Microsoft phi-4-mini i SQLite FTS5 (Zero-Chmura SLM)\n", + "\n", + "> **Autor:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Tematyka:** Małe Modele Językowe (SLM), SQLite FTS5 BM25, Gęste Wektory, Fuzja Wzajemnych Rang (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motywacja: Dylemat Odzyskiwania Słów Kluczowych w Lokalnych SLM\n", + "Standardowe architektury RAG opierające się wyłącznie na gęstych wektorach często zawodzą przy odzyskiwaniu dokładnych tokenów liczbowych (np. `2,340,000 TL`, kody kontraktów, numery kont). \n", + "Z kolei rzadkie wyszukiwanie leksykalne (BM25) pomija semantyczne synonimy i pytania parafrazowane.\n", + "\n", + "Ten przewodnik pokazuje, jak zaimplementować **wysokowydajny, pamięciowy silnik wyszukiwania hybrydowego**, łączący:\n", + "1. **Gęste wektory** (podobieństwo kosinusowe)\n", + "2. **Rzadkie wyszukiwanie leksykalne** (SQLite FTS5 BM25)\n", + "3. **Fuzję wzajemnych rang (RRF, $k=60$)**\n", + "4. **Generowanie cytowań osadzonych (`[1]`, `[2]`)** z Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Podwójny schemat SQLite (Gęste wektory + Wirtualna tabela FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Benchmark przechwytywania i wykonania próbek\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Tworzenie Opartych na Faktach Promptów dla Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Zastrzeżenie**:\nNiniejszy dokument został przetłumaczony za pomocą usługi tłumaczenia AI [Co-op Translator](https://github.com/Azure/co-op-translator). Choć dążymy do dokładności, prosimy pamiętać, że automatyczne tłumaczenia mogą zawierać błędy lub niedokładności. Oryginalny dokument w jego języku źródłowym należy uznawać za autorytatywne źródło. W przypadku informacji krytycznych zalecane jest skorzystanie z profesjonalnego tłumaczenia wykonanego przez człowieka. Nie ponosimy odpowiedzialności za jakiekolwiek nieporozumienia lub błędne interpretacje wynikające z użycia tego tłumaczenia.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/pt-BR/.co-op-translator.json b/translations/pt-BR/.co-op-translator.json index c44a6d2a7..b0623dbb5 100644 --- a/translations/pt-BR/.co-op-translator.json +++ b/translations/pt-BR/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "pt-BR" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:44:07+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:18:10+00:00", "source_file": "README.md", "language_code": "pt-BR" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "pt-BR" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:19:02+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "pt-BR" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:40+00:00", diff --git a/translations/pt-BR/README.md b/translations/pt-BR/README.md index 314fec4c1..6d28b4241 100644 --- a/translations/pt-BR/README.md +++ b/translations/pt-BR/README.md @@ -1,42 +1,42 @@ -# Phi Cookbook: Exemplos Práticos com os Modelos Phi da Microsoft +# Livro de Receitas Phi: Exemplos Práticos com os Modelos Phi da Microsoft [![Abra e use os exemplos no GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Abra em Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Abrir em Contêineres de Desenvolvimento](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![Contribuidores GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![Issues GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Pull-requests GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Bem-vindos](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Contribuidores do GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![Issues no GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![Pull-requests no GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Bem-Vindos](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![Observadores GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![Forks GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![Estrelas GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![Observadores do GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![Forks no GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![Estrelas no GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) -[![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) +[![Discord Microsoft Foundry](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) Phi é uma série de modelos de IA open source desenvolvidos pela Microsoft. -Phi é atualmente o modelo de linguagem pequena (SLM) mais poderoso e econômico, com benchmarks muito bons em múltiplas línguas, raciocínio, geração de texto/chat, codificação, imagens, áudio e outros cenários. +Phi é atualmente o modelo de linguagem pequena (SLM) mais poderoso e econômico, com benchmarks muito bons em multi-idiomas, raciocínio, geração de texto/chat, codificação, imagens, áudio e outros cenários. -Você pode implantar Phi na nuvem ou em dispositivos na ponta, e pode facilmente construir aplicações de IA generativa com poder computacional limitado. +Você pode implantar Phi na nuvem ou em dispositivos de borda, e pode facilmente construir aplicações de IA generativa com poder computacional limitado. -Siga estes passos para começar a usar este recurso: -1. **Faça um Fork do Repositório**: Clique [![Forks GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +Siga estes passos para começar a usar esses recursos: +1. **Faça um Fork do Repositório**: Clique [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Clone o Repositório**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Participe da Comunidade AI Microsoft no Discord e conheça especialistas e outros desenvolvedores**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Junte-se à Comunidade de IA Microsoft Discord e conheça especialistas e outros desenvolvedores**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/pt-BR/cover.eb18d1b9605d754b.webp) ### 🌐 Suporte Multilíngue -#### Suportado via Ação do GitHub (Automatizado & Sempre Atualizado) +#### Suportado via GitHub Action (Automatizado & Sempre Atualizado) -[Árabe](../ar/README.md) | [Bengali](../bn/README.md) | [Búlgaro](../bg/README.md) | [Burmês (Myanmar)](../my/README.md) | [Chinês (Simplificado)](../zh-CN/README.md) | [Chinês (Tradicional, Hong Kong)](../zh-HK/README.md) | [Chinês (Tradicional, Macau)](../zh-MO/README.md) | [Chinês (Tradicional, Taiwan)](../zh-TW/README.md) | [Croata](../hr/README.md) | [Tcheco](../cs/README.md) | [Dinamarquês](../da/README.md) | [Holandês](../nl/README.md) | [Estoniano](../et/README.md) | [Finlandês](../fi/README.md) | [Francês](../fr/README.md) | [Alemão](../de/README.md) | [Grego](../el/README.md) | [Hebraico](../he/README.md) | [Hindi](../hi/README.md) | [Húngaro](../hu/README.md) | [Indonésio](../id/README.md) | [Italiano](../it/README.md) | [Japonês](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Coreano](../ko/README.md) | [Lituano](../lt/README.md) | [Malaio](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Pidgin Nigeriano](../pcm/README.md) | [Norueguês](../no/README.md) | [Persa (Farsi)](../fa/README.md) | [Polonês](../pl/README.md) | [Português (Brasil)](./README.md) | [Português (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romeno](../ro/README.md) | [Russo](../ru/README.md) | [Sérvio (Cirílico)](../sr/README.md) | [Eslovaco](../sk/README.md) | [Esloveno](../sl/README.md) | [Espanhol](../es/README.md) | [Suaíli](../sw/README.md) | [Sueco](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tâmil](../ta/README.md) | [Telugu](../te/README.md) | [Tailandês](../th/README.md) | [Turco](../tr/README.md) | [Ucraniano](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamita](../vi/README.md) +[Árabe](../ar/README.md) | [Bengali](../bn/README.md) | [Búlgaro](../bg/README.md) | [Birmanês (Myanmar)](../my/README.md) | [Chinês (Simplificado)](../zh-CN/README.md) | [Chinês (Tradicional, Hong Kong)](../zh-HK/README.md) | [Chinês (Tradicional, Macau)](../zh-MO/README.md) | [Chinês (Tradicional, Taiwan)](../zh-TW/README.md) | [Croata](../hr/README.md) | [Tcheco](../cs/README.md) | [Dinamarquês](../da/README.md) | [Holandês](../nl/README.md) | [Estoniano](../et/README.md) | [Finlandês](../fi/README.md) | [Francês](../fr/README.md) | [Alemão](../de/README.md) | [Grego](../el/README.md) | [Hebraico](../he/README.md) | [Hindi](../hi/README.md) | [Húngaro](../hu/README.md) | [Indonésio](../id/README.md) | [Italiano](../it/README.md) | [Japonês](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Coreano](../ko/README.md) | [Lituano](../lt/README.md) | [Malaio](../ms/README.md) | [Malaiala](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Pidgin Nigeriano](../pcm/README.md) | [Norueguês](../no/README.md) | [Persa (Farsi)](../fa/README.md) | [Polonês](../pl/README.md) | [Português (Brasil)](./README.md) | [Português (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romeno](../ro/README.md) | [Russo](../ru/README.md) | [Sérvio (Cirílico)](../sr/README.md) | [Eslovaco](../sk/README.md) | [Esloveno](../sl/README.md) | [Espanhol](../es/README.md) | [Suaíli](../sw/README.md) | [Sueco](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tâmil](../ta/README.md) | [Telugu](../te/README.md) | [Tailandês](../th/README.md) | [Turco](../tr/README.md) | [Ucraniano](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamita](../vi/README.md) > **Prefere Clonar Localmente?** > -> Este repositório inclui traduções em mais de 50 idiomas, o que aumenta significativamente o tamanho do download. Para clonar sem as traduções, use o checkout esparso: +> Este repositório inclui traduções para mais de 50 idiomas, o que aumenta significativamente o tamanho do download. Para clonar sem traduções, use o sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,73 +52,75 @@ Siga estes passos para começar a usar este recurso: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Isso fornece tudo o que você precisa para completar o curso com um download muito mais rápido. +> Isso lhe dá tudo o que precisa para completar o curso com um download muito mais rápido. -## Índice +## Sumário - Introdução - [Bem-vindo à Família Phi](./md/01.Introduction/01/01.PhiFamily.md) - [Configurando seu ambiente](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Compreendendo Tecnologias-Chave](./md/01.Introduction/01/01.Understandingtech.md) - - [Segurança em IA para os Modelos Phi](./md/01.Introduction/01/01.AISafety.md) + - [Segurança em IA para Modelos Phi](./md/01.Introduction/01/01.AISafety.md) - [Suporte de Hardware Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Modelos Phi & Disponibilidade nas plataformas](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Usando guidance-ai e Phi](./md/01.Introduction/01/01.Guidance.md) - - [Modelos no Marketplace do GitHub](https://github.com/marketplace/models) - - [Catálogo de Modelos AI do Azure](https://ai.azure.com) + - [Modelos Phi & Disponibilidade em diversas plataformas](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Usando Guidance-ai e Phi](./md/01.Introduction/01/01.Guidance.md) + - [Modelos no GitHub Marketplace](https://github.com/marketplace/models) + - [Catálogo de Modelos de IA do Azure](https://ai.azure.com) - Inferência Phi em diferentes ambientes - [Hugging face](./md/01.Introduction/02/01.HF.md) - [Modelos GitHub](./md/01.Introduction/02/02.GitHubModel.md) - - [Catálogo de Modelos Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Catálogo Microsoft Foundry Model](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - Inferência Família Phi - - [Inferência Phi em iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Inferência Phi em Android](./md/01.Introduction/03/Android_Inference.md) - - [Inferência Phi em Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Inferência Phi em AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Inferência Phi no iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Inferência Phi no Android](./md/01.Introduction/03/Android_Inference.md) + - [Inferência Phi no Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Inferência Phi no PC de IA](./md/01.Introduction/03/AIPC_Inference.md) - [Inferência Phi com Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Inferência Phi em Servidor Local](./md/01.Introduction/03/Local_Server_Inference.md) - - [Inferência Phi em Servidor Remoto usando AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Inferência Phi no Servidor Local](./md/01.Introduction/03/Local_Server_Inference.md) + - [Inferência Phi no Servidor Remoto usando AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - [Inferência Phi com Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inferência Phi--Visão em Local](./md/01.Introduction/03/Vision_Inference.md) + - [Inferência Phi--Visão Local](./md/01.Introduction/03/Vision_Inference.md) - [Inferência Phi com Kaito AKS, Azure Containers (suporte oficial)](./md/01.Introduction/03/Kaito_Inference.md) -- [Quantizando Família Phi](./md/01.Introduction/04/QuantifyingPhi.md) +- [Quantificando Família Phi](./md/01.Introduction/04/QuantifyingPhi.md) - [Quantizando Phi-3.5 / 4 usando llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Quantizando Phi-3.5 / 4 usando extensões Generative AI para onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Quantizando Phi-3.5 / 4 usando extensões de IA generativa para onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [Quantizando Phi-3.5 / 4 usando Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - [Quantizando Phi-3.5 / 4 usando Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- Avaliação Phi +- Avaliação Phi - [IA Responsável](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry para Avaliação](./md/01.Introduction/05/AIFoundry.md) - [Usando Promptflow para Avaliação](./md/01.Introduction/05/Promptflow.md) - RAG com Azure AI Search - - [Como usar Phi-4-mini e Phi-4-multimodal(RAG) com Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Como usar Phi-4-mini e Phi-4-multimodal (RAG) com Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG híbrido local Zero-Cloud com SQLite FTS5 e phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Exemplos de desenvolvimento de aplicações Phi - Aplicações de Texto & Chat - - Exemplos Phi-4 + - Exemplos Phi-4 - [📓] [Chat com Modelo Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Chat com Modelo Phi-4 local ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Chat .NET Console App com Phi-4 ONNX usando Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Exemplos Phi-3 / 3.5 + - [Aplicação Console Chat .NET com Phi-4 ONNX usando Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Exemplos - [Chatbot Local no navegador usando Phi3, ONNX Runtime Web e WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - Phi-3-mini interativo e OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [Multi Modelo - Phi-3-mini Interativo e OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Construindo um wrapper e usando Phi-3 com MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [Otimização de Modelo - Como otimizar o modelo Phi-3-min para ONNX Runtime Web com Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [App WinUI3 com Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Exemplo de App de Notas com IA Multi Model no WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Ajuste fino e integração de modelos Phi-3 personalizados com Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Ajuste fino e integração de modelos Phi-3 personalizados com Prompt flow no Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Avalie o modelo Phi-3 / Phi-3.5 ajustado fino no Microsoft Foundry com foco nos Princípios de IA Responsável da Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [Exemplo de App Notas AI Potencializado Multi Modelo WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Fine-tune e Integre modelos Phi-3 personalizados com Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Fine-tune e Integre modelos Phi-3 personalizados com Prompt flow no Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Avalie o Modelo Phi-3 / Phi-3.5 Fine-tuned no Microsoft Foundry com foco nos Princípios de IA Responsável da Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Exemplo de predição de linguagem Phi-3.5-mini-instruct (Chinês/Inglês)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Usando GPU do Windows para criar solução Prompt flow com Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) @@ -126,74 +128,74 @@ Siga estes passos para começar a usar este recurso: - [Exemplo Q&A .NET usando modelo ONNX Phi-3 local com Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - [App de chat console .NET com Semantic Kernel e Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Exemplos de Código Baseados em Azure AI Inference SDK - - Exemplos Phi-4 + - Exemplos de Código do Azure AI Inference SDK + - Exemplos Phi-4 - [📓] [Gerar código de projeto usando Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Exemplos Phi-3 / 3.5 - - [Construa seu próprio GitHub Copilot Chat no Visual Studio Code com Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Crie seu próprio Chat Copilot Agent no Visual Studio Code com Phi-3.5 pelos Modelos do GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Construa seu próprio Visual Studio Code GitHub Copilot Chat com Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Crie seu próprio Agente Copilot de Chat para Visual Studio Code com Phi-3.5 por GitHub Models](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Exemplos de Raciocínio Avançado - - Exemplos Phi-4 - - [📓] [Exemplos Phi-4-mini-reasoning ou Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Ajuste fino Phi-4-mini-reasoning com Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Ajuste fino Phi-4-mini-reasoning com Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning com Modelos GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning com Modelos Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Exemplos Phi-4 + - [📓] [Exemplos Phi-4-mini-raciocínio ou Phi-4-raciocínio](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Fine-tuning Phi-4-mini-raciocínio com Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Fine-tuning Phi-4-mini-raciocínio com Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-raciocínio com GitHub Models](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-raciocínio com Microsoft Foundry Models](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demos - [Demos Phi-4-mini hospedados no Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - [Demos Phi-4-multimodal hospedados no Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - Exemplos de Visão - - Exemplos Phi-4 - - [📓] [Use Phi-4-multimodal para ler imagens e gerar código](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Exemplos Phi-4 + - [📓] [Use Phi-4-multimodal para ler imagens e gerar código](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Exemplos Phi-3 / 3.5 - - [📓][Phi-3-visão Texto de imagem para texto](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-visão - texto de imagem para texto](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-visão CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Reciclagem Phi-3](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-visão - Assistente visual de linguagem - com Phi3-Visão e OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [📓][Phi-3-visão Embedding CLIP](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [DEMO: Phi-3 Reciclagem](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-visão - Assistente de linguagem visual - com Phi3-Vision e OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Visão Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Visão OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Exemplo Phi-3.5 Visão multi-frame ou multi-imagem](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Modelo ONNX Local Phi-3 Visão usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Modelo ONNX Local Phi-3 Visão baseado em menu usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Phi-3.5 Visão multi-quadro ou amostra multi-imagem](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Visão Modelo Local ONNX usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Modelo Local ONNX Phi-3 Visão baseado em Menu usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - Exemplos de Raciocínio-Visão - - Phi-4-Raciocínio-Visão-15B + - Phi-4-Raciocínio-Visão-15B - [📓] [Usando Phi-4-Raciocínio-Visão-15B para detectar atravessamento fora da faixa](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [Usando Phi-4-Raciocínio-Visão-15B para matemática](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - [📓] [Usando Phi-4-Raciocínio-Visão-15B para detectar UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Exemplos de Matemática - - Exemplos Phi-4-Mini-Flash-Reasoning-Instruct [Demo Matemática com Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Exemplos Phi-4-Mini-Flash-Raciocínio-Instrutivo [Demonstração de Matemática com Phi-4-Mini-Flash-Raciocínio-Instrutivo](./md/02.Application/09.Math/MathDemo.ipynb) - Exemplos de Áudio - - Exemplos Phi-4 - - [📓] [Extração de transcrições de áudio usando Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Exemplo de Áudio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Exemplo de Tradução de Fala Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - Exemplos Phi-4 + - [📓] [Extraindo transcrições de áudio usando Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal Exemplo de Áudio](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal Exemplo de Tradução de Fala](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - [Aplicação console .NET usando Phi-4-multimodal Áudio para analisar arquivo de áudio e gerar transcrição](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - Exemplos MOE - Exemplos Phi-3 / 3.5 - - [📓] [Modelos Mixture of Experts Phi-3.5 (MoEs) Exemplo em Mídias Sociais](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Construindo um pipeline de Geração Aumentada por Recuperação (RAG) com NVIDIA NIM Phi-3 MOE, Azure AI Search, e LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Modelos de Mistura de Especialistas (MoEs) Phi-3.5 - Exemplo de Mídias Sociais](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Construindo um Pipeline de Geração com Recuperação (RAG) com NVIDIA NIM Phi-3 MOE, Azure AI Search e LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Exemplos de Chamada de Função + - Exemplos de Function Calling - Exemplos Phi-4 🆕 - - [📓] [Usando Chamada de Função com Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Usando Chamada de Função para criar multi-agentes com Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Usando Chamada de Função com Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Usando Chamada de Função com ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Exemplos de Mistura Multimodal + - [📓] [Usando Function Calling com Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Usando Function Calling para criar multi-agentes com Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Usando Function Calling com Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Usando Function Calling com ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Exemplos de Multimodal Mixing - Exemplos Phi-4 🆕 - [📓] [Usando Phi-4-multimodal como jornalista de tecnologia](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [Aplicação console .NET usando Phi-4-multimodal para analisar imagens](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Exemplos de Fine-tuning Phi +- Fine-tuning Phi Exemplos - [Cenários de Fine-tuning](./md/03.FineTuning/FineTuning_Scenarios.md) - [Fine-tuning vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Deixe Phi-3 se tornar um especialista da indústria com Fine-tuning](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Fine-tuning Deixe Phi-3 se tornar um especialista da indústria](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - [Fine-tuning Phi-3 com AI Toolkit para VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [Fine-tuning Phi-3 com Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - [Fine-tuning Phi-3 com Lora](./md/03.FineTuning/FineTuning_Lora.md) @@ -201,39 +203,40 @@ Siga estes passos para começar a usar este recurso: - [Fine-tuning Phi-3 com Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - [Fine-tuning Phi-3 com Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - [Fine-tuning com Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Laboratório prático de Fine-tuning com Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) + - [Fine-tuning com Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - [Fine-tuning Phi-3-visão com Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Fine-tuning Phi-3 com Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Fine-tuning Phi-3-visão (suporte oficial)](./md/03.FineTuning/FineTuning_Vision.md) - - [Ajuste Fino do Phi-3 com Kaito AKS, Contêineres Azure (Suporte Oficial)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Ajuste Fino do Phi-3 e 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) -- Laboratório Prático + - [Ajuste fino do Phi-3 com o Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Ajuste fino do Phi-3-vision (suporte oficial)](./md/03.FineTuning/FineTuning_Vision.md) + - [Ajuste fino do Phi-3 com Kaito AKS , Contêineres Azure (suporte oficial)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Ajuste fino do Phi-3 e 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + +- Laboratório prático - [Explorando modelos de ponta: LLMs, SLMs, desenvolvimento local e mais](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Desbloqueando o Potencial de NLP: Ajuste Fino com Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Desbloqueando o potencial do NLP: ajuste fino com o Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - Artigos e Publicações Acadêmicas - [Textbooks Are All You Need II: relatório técnico phi-1.5](https://arxiv.org/abs/2309.05463) - - [Relatório Técnico Phi-3: Um Modelo de Linguagem Altamente Capaz Localmente no Seu Celular](https://arxiv.org/abs/2404.14219) + - [Relatório Técnico Phi-3: Um Modelo de Linguagem Altamente Capaz Localmente no Seu Telefone](https://arxiv.org/abs/2404.14219) - [Relatório Técnico Phi-4](https://arxiv.org/abs/2412.08905) - - [Relatório Técnico Phi-4-Mini: Modelos de Linguagem Multimodal Compactos porém Poderosos via Mistura de LoRAs](https://arxiv.org/abs/2503.01743) - - [Otimização de Pequenos Modelos de Linguagem para Chamada de Funções em Veículos](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Ajuste Fino do PHI-3 para Respostas de Perguntas de Múltipla Escolha: Metodologia, Resultados e Desafios](https://arxiv.org/abs/2501.01588) - - [Relatório Técnico Phi-4-raciocínio](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Relatório Técnico Phi-4-mini-raciocínio](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Relatório Técnico Phi-4-Mini: Modelos de Linguagem Multimodais Compactos e Poderosos via Mistura de LoRAs](https://arxiv.org/abs/2503.01743) + - [Otimização de Pequenos Modelos de Linguagem para Chamadas de Função em Veículos](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Ajuste Fino do PHI-3 para Respostas a Perguntas de Múltipla Escolha: Metodologia, Resultados e Desafios](https://arxiv.org/abs/2501.01588) + - [Relatório Técnico Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Relatório Técnico Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Usando Modelos Phi ### Phi no Microsoft Foundry -Você pode aprender como usar o Microsoft Phi e como construir soluções E2E em seus diferentes dispositivos de hardware. Para experimentar o Phi por conta própria, comece brincando com os modelos e personalizando o Phi para seus cenários usando o [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Você pode aprender mais em Começando com [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Você pode aprender como usar o Microsoft Phi e como construir soluções E2E em seus diferentes dispositivos de hardware. Para experimentar o Phi por conta própria, comece explorando os modelos e personalizando o Phi para seus cenários usando o [Catálogo de Modelos AI do Microsoft Foundry Azure](https://aka.ms/phi3-azure-ai). Você pode aprender mais em Introdução ao [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) **Playground** Cada modelo tem um playground dedicado para testar o modelo [Azure AI Playground](https://aka.ms/try-phi3). ### Phi nos Modelos do GitHub -Você pode aprender como usar o Microsoft Phi e como construir soluções E2E em seus diferentes dispositivos de hardware. Para experimentar o Phi por conta própria, comece brincando com o modelo e personalizando o Phi para seus cenários usando o [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Você pode aprender mais em Começando com [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Você pode aprender como usar o Microsoft Phi e como construir soluções E2E em seus diferentes dispositivos de hardware. Para experimentar o Phi por conta própria, comece explorando o modelo e personalizando o Phi para seus cenários usando o [Catálogo de Modelos do GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Você pode aprender mais em Introdução ao [Catálogo de Modelos do GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) **Playground** Cada modelo tem um [playground dedicado para testar o modelo](/md/02.QuickStart/GitHubModel_QuickStart.md). @@ -243,9 +246,9 @@ Cada modelo tem um [playground dedicado para testar o modelo](/md/02.QuickStart/ Você também pode encontrar o modelo no [Hugging Face](https://huggingface.co/microsoft) **Playground** - [Playground do Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) + [Playground Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Outros Cursos + ## 🎒 Outros Cursos Nossa equipe produz outros cursos! Confira: @@ -272,11 +275,11 @@ Nossa equipe produz outros cursos! Confira: --- -### Aprendizado Central +### Aprendizado Básico [![ML para Iniciantes](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Ciência de Dados para Iniciantes](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![IA para Iniciantes](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Segurança Cibernética para Iniciantes](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Cibersegurança para Iniciantes](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Desenvolvimento Web para Iniciantes](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![IoT para Iniciantes](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![Desenvolvimento XR para Iniciantes](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) @@ -284,41 +287,43 @@ Nossa equipe produz outros cursos! Confira: --- ### Série Copilot -[![Copilot para Programação Assistida por IA](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot para Programação Emparelhada com IA](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot para C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Aventura Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## IA Responsável +## IA Responsável A Microsoft está comprometida em ajudar nossos clientes a usar nossos produtos de IA de forma responsável, compartilhando nossos aprendizados e construindo parcerias baseadas em confiança por meio de ferramentas como Notas de Transparência e Avaliações de Impacto. Muitos desses recursos podem ser encontrados em [https://aka.ms/RAI](https://aka.ms/RAI). A abordagem da Microsoft para IA responsável é fundamentada em nossos princípios de IA de justiça, confiabilidade e segurança, privacidade e segurança, inclusão, transparência e responsabilidade. -Modelos em grande escala de linguagem natural, imagem e voz - como os usados neste exemplo - podem potencialmente se comportar de maneiras injustas, não confiáveis ou ofensivas, causando danos. Por favor, consulte a [nota de transparência do serviço Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) para estar informado sobre riscos e limitações. -A abordagem recomendada para mitigar esses riscos é incluir um sistema de segurança em sua arquitetura que possa detectar e prevenir comportamentos prejudiciais. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) fornece uma camada independente de proteção, capaz de detectar conteúdo prejudicial gerado por usuários e IA em aplicações e serviços. O Azure AI Content Safety inclui APIs de texto e imagem que permitem detectar material prejudicial. Dentro do Microsoft Foundry, o serviço Content Safety permite visualizar, explorar e testar códigos de exemplo para detectar conteúdo prejudicial em diferentes modalidades. A seguinte [documentação de início rápido](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) orienta você a fazer solicitações ao serviço. +Modelos de linguagem natural, imagem e fala em larga escala – como os usados neste exemplo – podem potencialmente se comportar de maneiras injustas, não confiáveis ou ofensivas, causando danos. Por favor, consulte a [nota de transparência do serviço Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) para se informar sobre riscos e limitações. + + +A abordagem recomendada para mitigar esses riscos é incluir um sistema de segurança em sua arquitetura que possa detectar e prevenir comportamentos nocivos. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) fornece uma camada independente de proteção, capaz de detectar conteúdo nocivo gerado por usuários e por IA em aplicações e serviços. O Azure AI Content Safety inclui APIs de texto e imagem que permitem detectar material que é prejudicial. Dentro do Microsoft Foundry, o serviço Content Safety permite que você visualize, explore e experimente código de exemplo para detectar conteúdo nocivo em diferentes modalidades. A seguinte [documentação de introdução rápida](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) orienta você a fazer solicitações ao serviço. -Outro aspecto a ser considerado é o desempenho geral do aplicativo. Com aplicações multimodais e multimodelos, consideramos desempenho como o fato do sistema funcionar conforme você e seus usuários esperam, incluindo não gerar saídas prejudiciais. É importante avaliar o desempenho do seu aplicativo geral usando os [avaliadores de Desempenho, Qualidade, Risco e Segurança](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Você também tem a capacidade de criar e avaliar com [avaliadores personalizados](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Outro aspecto a ser considerado é o desempenho geral da aplicação. Com aplicações multimodais e multimodelos, consideramos desempenho como o sistema funcionando conforme você e seus usuários esperam, inclusive não gerando saídas prejudiciais. É importante avaliar o desempenho da sua aplicação geral usando os [avaliadores de Desempenho e Qualidade e Risco e Segurança](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Você também tem a capacidade de criar e avaliar com [avaliadores personalizados](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Você pode avaliar sua aplicação de IA em seu ambiente de desenvolvimento usando o [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Dado um conjunto de dados de teste ou um alvo, as gerações da sua aplicação de IA generativa são medidas quantitativamente com avaliadores internos ou avaliadores personalizados de sua escolha. Para começar a usar o azure ai evaluation sdk para avaliar seu sistema, você pode seguir o [guia de início rápido](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Depois de executar uma avaliação, você pode [visualizar os resultados no Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Você pode avaliar sua aplicação de IA em seu ambiente de desenvolvimento usando o [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Dado um conjunto de dados de teste ou um alvo, as gerações da sua aplicação de IA generativa são medidas quantitativamente com avaliadores embutidos ou avaliadores personalizados de sua escolha. Para começar com o azure ai evaluation sdk para avaliar seu sistema, você pode seguir o [guia de introdução rápida](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Uma vez que você execute uma avaliação, você pode [visualizar os resultados no Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Marcas Registradas -Este projeto pode conter marcas registradas ou logotipos de projetos, produtos ou serviços. O uso autorizado das marcas ou logotipos da Microsoft está sujeito e deve seguir as [Diretrizes de Marcas e Identidade Visual da Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -O uso de marcas ou logotipos da Microsoft em versões modificadas deste projeto não deve causar confusão ou implicar patrocínio da Microsoft. Qualquer uso de marcas ou logotipos de terceiros está sujeito às políticas desses terceiros. +Este projeto pode conter marcas registradas ou logotipos de projetos, produtos ou serviços. O uso autorizado de marcas registradas ou logotipos da Microsoft está sujeito e deve seguir as [Diretrizes de Marca e Marcas Registradas da Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +O uso de marcas registradas ou logotipos da Microsoft em versões modificadas deste projeto não deve causar confusão ou implicar patrocínio da Microsoft. Qualquer uso de marcas registradas ou logotipos de terceiros está sujeito às políticas desses terceiros. -## Obter Ajuda +## Obtenha Ajuda -Se você ficar travado ou tiver alguma dúvida sobre como construir aplicativos de IA, junte-se: +Se ficar com dúvidas ou tiver perguntas sobre como construir aplicativos de IA, participe: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Se você tiver feedback sobre o produto ou encontrar erros durante a construção, visite: +Se você tiver feedback sobre o produto ou erros enquanto constrói, visite: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Aviso Legal**: -Este documento foi traduzido utilizando o serviço de tradução por IA [Co-op Translator](https://github.com/Azure/co-op-translator). Embora nos esforcemos para garantir a precisão, esteja ciente de que traduções automatizadas podem conter erros ou imprecisões. O documento original em seu idioma nativo deve ser considerado a fonte autorizada. Para informações críticas, recomenda-se a tradução profissional humana. Não nos responsabilizamos por quaisquer mal-entendidos ou interpretações errôneas decorrentes do uso desta tradução. +**Aviso Legal**: +Este documento foi traduzido usando o serviço de tradução por IA [Co-op Translator](https://github.com/Azure/co-op-translator). Embora nos esforcemos pela precisão, por favor, esteja ciente de que traduções automatizadas podem conter erros ou imprecisões. O documento original em seu idioma nativo deve ser considerado a fonte autorizada. Para informações críticas, recomenda-se tradução profissional humana. Não nos responsabilizamos por quaisquer mal-entendidos ou interpretações incorretas decorrentes do uso desta tradução. \ No newline at end of file diff --git a/translations/pt-BR/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/pt-BR/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..e2ad29a55 --- /dev/null +++ b/translations/pt-BR/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 RAG Híbrido com Microsoft phi-4-mini & SQLite FTS5 (SLM Zero-Cloud)\n", + "\n", + "> **Autor:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Foco:** Pequenos Modelos de Linguagem (SLMs), SQLite FTS5 BM25, Embeddings Densos, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivação: O Dilema da Recuperação por Palavra-chave em SLMs Locais\n", + "Arquiteturas RAG padrão baseadas puramente em embeddings densos de vetores frequentemente falham ao recuperar tokens numéricos exatos (ex.: `2,340,000 TL`, códigos de contratos, números de contas). \n", + "Por outro lado, a busca lexical esparsa (BM25) perde sinônimos semânticos e perguntas parafraseadas.\n", + "\n", + "Este cookbook demonstra como implementar um **motor de Recuperação Híbrido em memória de alta velocidade** combinando:\n", + "1. **Vetores Densos** (Similaridade Cosseno)\n", + "2. **Busca Lexical Esparsa** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Geração de Citação Fundamentada (`[1]`, `[2]`)** com Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Esquema SQLite Duplo (Vetores Densos + Tabela Virtual FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Benchmark de Ingestão e Execução de Amostras\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Formulação de Prompt Fundamentada para Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Aviso Legal**:\nEste documento foi traduzido usando o serviço de tradução por IA [Co-op Translator](https://github.com/Azure/co-op-translator). Embora nos esforcemos pela precisão, por favor, esteja ciente de que traduções automatizadas podem conter erros ou imprecisões. O documento original em seu idioma nativo deve ser considerado a fonte autorizada. Para informações críticas, recomenda-se tradução profissional humana. Não nos responsabilizamos por quaisquer mal-entendidos ou interpretações incorretas decorrentes do uso desta tradução.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/pt-PT/.co-op-translator.json b/translations/pt-PT/.co-op-translator.json index 96eada321..1c7832b62 100644 --- a/translations/pt-PT/.co-op-translator.json +++ b/translations/pt-PT/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "pt-PT" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:42:01+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:15:07+00:00", "source_file": "README.md", "language_code": "pt-PT" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "pt-PT" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:18:48+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "pt-PT" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:37+00:00", diff --git a/translations/pt-PT/README.md b/translations/pt-PT/README.md index f3ac74ec5..fb3be9b07 100644 --- a/translations/pt-PT/README.md +++ b/translations/pt-PT/README.md @@ -1,42 +1,42 @@ # Phi Cookbook: Exemplos Práticos com os Modelos Phi da Microsoft -[![Open and use the samples in GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Open in Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Abrir e usar os exemplos no GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Abrir em Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub contributors](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub issues](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub pull-requests](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Contribuidores GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![Problemas GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![Pedidos de Pull GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Bem-vindos](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub watchers](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub stars](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![Observadores GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![Forks GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![Estrelas GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi é uma série de modelos de IA de código aberto desenvolvidos pela Microsoft. +Phi é uma série de modelos de IA open source desenvolvidos pela Microsoft. -Phi é atualmente o modelo de linguagem pequena (SLM) mais poderoso e económico, com benchmarks muito bons em múltiplas línguas, raciocínio, geração de texto/chat, programação, imagens, áudio e outros cenários. +Phi é atualmente o modelo de linguagem pequeno (SLM) mais poderoso e económico, com benchmarks muito bons em multi-linguagem, raciocínio, geração de texto/chat, codificação, imagens, áudio e outros cenários. -Pode implementar o Phi na cloud ou em dispositivos edge e pode facilmente construir aplicações de IA generativa com capacidade computacional limitada. +Pode implantar o Phi na cloud ou em dispositivos edge, e pode construir facilmente aplicações de IA generativa com potência computacional limitada. -Siga estes passos para começar a utilizar estes recursos : -1. **Faça Fork do Repositório**: Clique em [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Clone o Repositório**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Junte-se à Comunidade Microsoft AI no Discord e conheça especialistas e outros desenvolvedores**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +Siga estes passos para começar a usar estes recursos: +1. **Fazer Fork do Repositório**: Clique em [![Forks GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Clonar o Repositório**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Junte-se à Comunidade Microsoft AI Discord e conheça especialistas e outros programadores**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/pt-PT/cover.eb18d1b9605d754b.webp) +![capa](../../translated_images/pt-PT/cover.eb18d1b9605d754b.webp) ### 🌐 Suporte Multilíngue #### Suportado via GitHub Action (Automatizado e Sempre Atualizado) -[Árabe](../ar/README.md) | [Bengali](../bn/README.md) | [Búlgaro](../bg/README.md) | [Birmanês (Myanmar)](../my/README.md) | [Chinês (Simplificado)](../zh-CN/README.md) | [Chinês (Tradicional, Hong Kong)](../zh-HK/README.md) | [Chinês (Tradicional, Macau)](../zh-MO/README.md) | [Chinês (Tradicional, Taiwan)](../zh-TW/README.md) | [Croata](../hr/README.md) | [Checo](../cs/README.md) | [Dinamarquês](../da/README.md) | [Holandês](../nl/README.md) | [Estónio](../et/README.md) | [Finlandês](../fi/README.md) | [Francês](../fr/README.md) | [Alemão](../de/README.md) | [Grego](../el/README.md) | [Hebraico](../he/README.md) | [Hindi](../hi/README.md) | [Húngaro](../hu/README.md) | [Indonésio](../id/README.md) | [Italiano](../it/README.md) | [Japonês](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Coreano](../ko/README.md) | [Lituano](../lt/README.md) | [Malaio](../ms/README.md) | [Malaiala](../ml/README.md) | [Marati](../mr/README.md) | [Nepali](../ne/README.md) | [Pidgin Nigeriano](../pcm/README.md) | [Norueguês](../no/README.md) | [Persa (Farsi)](../fa/README.md) | [Polaco](../pl/README.md) | [Português (Brasil)](../pt-BR/README.md) | [Português (Portugal)](./README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romeno](../ro/README.md) | [Russo](../ru/README.md) | [Sérvio (Cirílico)](../sr/README.md) | [Eslovaco](../sk/README.md) | [Esloveno](../sl/README.md) | [Espanhol](../es/README.md) | [Suaíli](../sw/README.md) | [Sueco](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugo](../te/README.md) | [Tailandês](../th/README.md) | [Turco](../tr/README.md) | [Ucraniano](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamita](../vi/README.md) +[Árabe](../ar/README.md) | [Bengali](../bn/README.md) | [Búlgaro](../bg/README.md) | [Birmanês (Myanmar)](../my/README.md) | [Chinês (Simplificado)](../zh-CN/README.md) | [Chinês (Tradicional, Hong Kong)](../zh-HK/README.md) | [Chinês (Tradicional, Macau)](../zh-MO/README.md) | [Chinês (Tradicional, Taiwan)](../zh-TW/README.md) | [Croata](../hr/README.md) | [Checo](../cs/README.md) | [Dinamarquês](../da/README.md) | [Holandês](../nl/README.md) | [Estónio](../et/README.md) | [Finlandês](../fi/README.md) | [Francês](../fr/README.md) | [Alemão](../de/README.md) | [Grego](../el/README.md) | [Hebraico](../he/README.md) | [Hindi](../hi/README.md) | [Húngaro](../hu/README.md) | [Indonésio](../id/README.md) | [Italiano](../it/README.md) | [Japonês](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Coreano](../ko/README.md) | [Lituano](../lt/README.md) | [Malaio](../ms/README.md) | [Malaiala](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalês](../ne/README.md) | [Pidgin Nigeriano](../pcm/README.md) | [Norueguês](../no/README.md) | [Persa (Farsi)](../fa/README.md) | [Polaco](../pl/README.md) | [Português (Brasil)](../pt-BR/README.md) | [Português (Portugal)](./README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romeno](../ro/README.md) | [Russo](../ru/README.md) | [Sérvio (Cirílico)](../sr/README.md) | [Eslovaco](../sk/README.md) | [Esloveno](../sl/README.md) | [Espanhol](../es/README.md) | [Suaíli](../sw/README.md) | [Sueco](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tâmil](../ta/README.md) | [Telugu](../te/README.md) | [Tailandês](../th/README.md) | [Turco](../tr/README.md) | [Ucraniano](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamita](../vi/README.md) > **Prefere Clonar Localmente?** > -> Este repositório inclui mais de 50 traduções de línguas, o que aumenta significativamente o tamanho do download. Para clonar sem traduções, use sparse checkout: +> Este repositório inclui mais de 50 traduções que aumentam significativamente o tamanho do download. Para clonar sem traduções, use o sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -60,10 +60,10 @@ Siga estes passos para começar a utilizar estes recursos : - Introdução - [Bem-vindo à Família Phi](./md/01.Introduction/01/01.PhiFamily.md) - [Configurar o seu ambiente](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Compreender Tecnologias Chave](./md/01.Introduction/01/01.Understandingtech.md) - - [Segurança em IA para Modelos Phi](./md/01.Introduction/01/01.AISafety.md) + - [Compreender as Tecnologias-Chave](./md/01.Introduction/01/01.Understandingtech.md) + - [Segurança de IA para Modelos Phi](./md/01.Introduction/01/01.AISafety.md) - [Suporte de Hardware Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Modelos Phi e Disponibilidade em várias plataformas](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Modelos Phi & Disponibilidade em várias plataformas](./md/01.Introduction/01/01.Edgeandcloud.md) - [Usar Guidance-ai e Phi](./md/01.Introduction/01/01.Guidance.md) - [Modelos do GitHub Marketplace](https://github.com/marketplace/models) - [Catálogo de Modelos Azure AI](https://ai.azure.com) @@ -77,7 +77,7 @@ Siga estes passos para começar a utilizar estes recursos : - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Inferência Phi Family +- Inferência Família Phi - [Inferência Phi em iOS](./md/01.Introduction/03/iOS_Inference.md) - [Inferência Phi em Android](./md/01.Introduction/03/Android_Inference.md) - [Inferência Phi em Jetson](./md/01.Introduction/03/Jetson_Inference.md) @@ -86,105 +86,107 @@ Siga estes passos para começar a utilizar estes recursos : - [Inferência Phi em Servidor Local](./md/01.Introduction/03/Local_Server_Inference.md) - [Inferência Phi em Servidor Remoto usando AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - [Inferência Phi com Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inferência Phi--Visão em Local](./md/01.Introduction/03/Vision_Inference.md) - - [Inferência Phi com Kaito AKS, Azure Containers (suporte oficial)](./md/01.Introduction/03/Kaito_Inference.md) -- [Quantificação da Família Phi](./md/01.Introduction/04/QuantifyingPhi.md) - - [Quantizar Phi-3.5 / 4 usando llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Quantizar Phi-3.5 / 4 usando extensões de IA Generativa para onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Quantizar Phi-3.5 / 4 usando Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Quantizar Phi-3.5 / 4 usando Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Avaliação Phi + - [Inferência Phi–Vision local](./md/01.Introduction/03/Vision_Inference.md) + - [Inferência Phi com Kaito AKS, Contêineres Azure (suporte oficial)](./md/01.Introduction/03/Kaito_Inference.md) +- [Quantificação Família Phi](./md/01.Introduction/04/QuantifyingPhi.md) + - [Quantização Phi-3.5 / 4 usando llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Quantização Phi-3.5 / 4 usando extensões de IA Generativa para onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Quantização Phi-3.5 / 4 usando Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Quantização Phi-3.5 / 4 usando Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- Avaliação Phi - [IA Responsável](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry para Avaliação](./md/01.Introduction/05/AIFoundry.md) - [Usar Promptflow para Avaliação](./md/01.Introduction/05/Promptflow.md) - RAG com Azure AI Search - [Como usar Phi-4-mini e Phi-4-multimodal (RAG) com Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG Híbrido Local Zero-Cloud com SQLite FTS5 e phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Exemplos de desenvolvimento de aplicações Phi - - Aplicações de Texto & Chat - - Exemplos Phi-4 - - [📓] [Chat com Modelo Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Chat com Modelo ONNX Phi-4 local .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [App Console Chat .NET com Phi-4 ONNX usando Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Exemplos Phi-3 / 3.5 - - [Chatbot Local no navegador usando Phi3, ONNX Runtime Web e WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - Aplicações de Texto e Chat + - Exemplos Phi-4 + - [📓] [Conversar com Modelo Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Conversar com Modelo Phi-4 local ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [App Console .NET para Chat com Phi-4 ONNX usando Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Exemplos + - [Chatbot local no browser usando Phi3, ONNX Runtime Web e WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Chat OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [Multi Modelo - Phi-3-mini interativo e OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Construir um wrapper e usar Phi-3 com MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [Otimização de Modelo - Como otimizar o modelo Phi-3-min para ONNX Runtime Web com Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [App WinUI3 com Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Exemplo de App de Notas com vários modelos alimentado por IA WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Ajuste fino e integração de modelos Phi-3 personalizados com Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Ajuste fino e integração de modelos Phi-3 personalizados com Prompt flow na Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Avaliação do modelo Phi-3 / Phi-3.5 ajustado na Microsoft Foundry com foco nos princípios de IA responsável da Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [Aplicação WinUI3 com Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[Exemplo de aplicação WinUI3 Multi Model AI Powered Notes](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Ajustar e integrar modelos Phi-3 personalizados com Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Ajustar e integrar modelos Phi-3 personalizados com Prompt flow na Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Avaliar o modelo Phi-3 / Phi-3.5 ajustado na Microsoft Foundry com foco nos princípios de IA responsável da Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Exemplo de predição de linguagem Phi-3.5-mini-instruct (Chinês/Inglês)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct Chatbot WebGPU RAG](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Usar GPU Windows para criar solução Prompt flow com Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Usar Microsoft Phi-3.5 tflite para criar app Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Chatbot Phi-3.5-Instruct WebGPU RAG](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Usar GPU do Windows para criar solução Prompt flow com Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Usar Microsoft Phi-3.5 tflite para criar aplicação Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - [Exemplo Q&A .NET usando modelo ONNX Phi-3 local com Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [App .NET consola de chat com Semantic Kernel e Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Aplicação chat consola .NET com Semantic Kernel e Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Exemplos Baseados em Código do Azure AI Inference SDK - - Exemplos Phi-4 + - Exemplos de Código do SDK de Inferência Azure AI + - Exemplos Phi-4 - [📓] [Gerar código de projeto usando Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Exemplos Phi-3 / 3.5 - - [Crie o seu próprio Visual Studio Code GitHub Copilot Chat com Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Crie o seu próprio agente Visual Studio Code Chat Copilot com Phi-3.5 usando modelos GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Construa o seu próprio VS Code GitHub Copilot Chat com a família Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Crie o seu próprio agente Copilot de chat no Visual Studio Code com Phi-3.5 pelos modelos do GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Exemplos de Raciocínio Avançado - - Exemplos Phi-4 - - [📓] [Exemplos Phi-4-mini-reasoning ou Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Ajuste fino de Phi-4-mini-reasoning com Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Ajuste fino de Phi-4-mini-reasoning com Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning com modelos GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning com modelos Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demonstrações - - [Demos Phi-4-mini alojados em Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Demos Phi-4-multimodal alojados em Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Exemplos Phi-4 + - [📓] [Exemplos Phi-4-mini-raciocínio ou Phi-4-raciocínio](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Ajuste fino Phi-4-mini-raciocínio com Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Ajuste fino Phi-4-mini-raciocínio com Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-raciocínio com modelos GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-raciocínio com modelos Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Demos + - [Demos Phi-4-mini hospedadas no Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Demos Phi-4-multimodal hospedadas no Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - Exemplos de Visão - - Exemplos Phi-4 - - [📓] [Usar Phi-4-multimodal para ler imagens e gerar código](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Exemplos Phi-4 + - [📓] [Usar Phi-4-multimodal para ler imagens e gerar código](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Exemplos Phi-3 / 3.5 - - [📓][Texto para texto de imagem Phi-3-vision](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 Recycling](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Assistente de linguagem visual - com Phi3-Vision e OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Exemplo Phi-3.5 Vision multi-frame ou multi-imagem](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Modelo Local ONNX Phi-3 Vision usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Modelo Local ONNX Phi-3 Vision baseado em menu usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Phi-3-visão - texto de imagem para texto](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-visão-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-visão CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [DEMO: Reciclagem Phi-3](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-visão - Assistente de linguagem visual - com Phi3-Vision e OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Visão Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Visão OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Visão multi-frame ou multi-imagem exemplo](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Visão Modelo ONNX Local usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Modelo ONNX Local baseado em menu Phi-3 Visão usando Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - Exemplos Raciocínio-Visão - - Phi-4-Reasoning-Vision-15B - - [📓] [Usar Phi-4-Reasoning-Vision-15B para detetar atravessamento ilegal](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Usar Phi-4-Reasoning-Vision-15B para matemática](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Usar Phi-4-Reasoning-Vision-15B para detetar UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - Phi-4-Raciocínio-Visão-15B + - [📓] [Usar Phi-4-Raciocínio-Visão-15B para detectar travessia indevida](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Usar Phi-4-Raciocínio-Visão-15B para matemática](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Usar Phi-4-Raciocínio-Visão-15B para detetar UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Exemplos de Matemática - - Exemplos Phi-4-Mini-Flash-Reasoning-Instruct [Demo Matemática com Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Exemplos Phi-4-Mini-Flash-Reasoning-Instruct [Demo Matemática com Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - Exemplos de Áudio - - Exemplos Phi-4 - - [📓] [Extrair transcrições de áudio usando Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Exemplo de Áudio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Exemplo de Tradução de Fala Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [Aplicação consola .NET usando Phi-4-multimodal Audio para analisar ficheiro áudio e gerar transcrição](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - Exemplos Áudio + - Exemplos Phi-4 + - [📓] [Extrair transcrições áudio usando Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Exemplo áudio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Exemplo de tradução de fala Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [Aplicação consola .NET usando Phi-4-multimodal áudio para analisar ficheiro áudio e gerar transcrição](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - Exemplos MOE - Exemplos Phi-3 / 3.5 - - [📓] [Exemplo Phi-3.5 Modelos de Mistura de Especialistas (MoEs) em Redes Sociais](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Construção de uma pipeline de geração aumentada por recuperação (RAG) com NVIDIA NIM Phi-3 MOE, Azure AI Search, e LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Modelos Phi-3.5 Mistura de Especialistas (MoEs) Exemplo Redes Sociais](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Construir Pipeline de Geração com Recuperação (RAG) com NVIDIA NIM Phi-3 MOE, Azure AI Search e LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - Exemplos de Chamada de Função - Exemplos Phi-4 🆕 - - [📓] [Usar Function Calling com Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Usar Function Calling para criar multi-agentes com Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Usar Function Calling com Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Usar Function Calling com ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - [📓] [Usar Chamada de Função com Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Usar Chamada de Função para criar multi-agentes com Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Usar Chamada de Função com Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Usar Chamada de Função com ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - Exemplos de Mistura Multimodal - Exemplos Phi-4 🆕 - [📓] [Usar Phi-4-multimodal como jornalista tecnológico](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) @@ -193,61 +195,62 @@ Siga estes passos para começar a utilizar estes recursos : - Ajuste fino de Exemplos Phi - [Cenários de ajuste fino](./md/03.FineTuning/FineTuning_Scenarios.md) - [Ajuste fino vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Ajuste fino para tornar Phi-3 um especialista na indústria](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Ajuste fino de Phi-3 com AI Toolkit para VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Ajuste fino de Phi-3 com Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Ajuste fino de Phi-3 com Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Ajuste fino de Phi-3 com QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Ajuste fino de Phi-3 com Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Ajuste fino de Phi-3 com Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Ajuste fino para tornar Phi-3 um perito da indústria](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Ajuste fino Phi-3 com AI Toolkit para VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Ajuste fino Phi-3 com Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Ajuste fino Phi-3 com Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Ajuste fino Phi-3 com QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Ajuste fino Phi-3 com Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Ajuste fino Phi-3 com Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - [Ajuste fino com Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Laboratório prático de ajuste fino com Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [Ajuste fino de Phi-3-vision com Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Ajuste fino de Phi-3 com Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Ajuste fino de Phi-3-vision (suporte oficial)](./md/03.FineTuning/FineTuning_Vision.md) - - [Ajuste Fino do Phi-3 com Kaito AKS, Azure Containers (Suporte Oficial)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Ajuste Fino do Phi-3 e Phi-3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + - [Ajuste fino Phi-3-vision com Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Ajuste fino do Phi-3 com o Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Ajuste fino do Phi-3-vision (suporte oficial)](./md/03.FineTuning/FineTuning_Vision.md) + - [Ajuste fino do Phi-3 com Kaito AKS, Azure Containers (suporte oficial)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Ajuste fino do Phi-3 e 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Laboratório Prático - - [Explorando modelos inovadores: LLMs, SLMs, desenvolvimento local e mais](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Desbloquear o Potencial do NLP: Ajuste Fino com Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Explorar modelos de ponta: LLMs, SLMs, desenvolvimento local e mais](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Desbloquear o potencial do NLP: Ajuste fino com Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Artigos e Publicações Académicas +- Artigos Académicos e Publicações - [Textbooks Are All You Need II: relatório técnico phi-1.5](https://arxiv.org/abs/2309.05463) - - [Relatório Técnico Phi-3: Um Modelo de Linguagem Altamente Capaz Localmente no Telemóvel](https://arxiv.org/abs/2404.14219) + - [Relatório Técnico Phi-3: Um Modelo de Linguagem Altamente Capaz Localmente no Seu Telemóvel](https://arxiv.org/abs/2404.14219) - [Relatório Técnico Phi-4](https://arxiv.org/abs/2412.08905) - - [Relatório Técnico Phi-4-Mini: Modelos de Linguagem Multimodais Compactos mas Poderosos via Mistura de LoRAs](https://arxiv.org/abs/2503.01743) - - [Otimização de Pequenos Modelos de Linguagem para Chamada de Funções em Veículos](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Ajuste Fino do PHI-3 para Respostas a Questões de Escolha Múltipla: Metodologia, Resultados e Desafios](https://arxiv.org/abs/2501.01588) + - [Relatório Técnico Phi-4-Mini: Modelos de Linguagem Multimodal Compactos mas Poderosos via Mistura-de-LoRAs](https://arxiv.org/abs/2503.01743) + - [Otimização de Pequenos Modelos de Linguagem para Chamada de Função In-Veículo](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Ajuste fino do PHI-3 para Resposta a Perguntas de Escolha Múltipla: Metodologia, Resultados e Desafios](https://arxiv.org/abs/2501.01588) - [Relatório Técnico Phi-4-raciocínio](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Relatório Técnico Phi-4-mini-raciocínio](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Utilizar Modelos Phi -### Phi na Microsoft Foundry +### Phi no Microsoft Foundry -Pode aprender a usar o Microsoft Phi e como construir soluções E2E nos seus diferentes dispositivos de hardware. Para experimentar o Phi por si próprio, comece a brincar com os modelos e a personalizar o Phi para os seus cenários usando o [Catálogo de Modelos Azure AI da Microsoft Foundry](https://aka.ms/phi3-azure-ai). Pode saber mais em Começar com a [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md). +Pode aprender a usar o Microsoft Phi e a construir soluções E2E nos seus diferentes dispositivos de hardware. Para experimentar o Phi, comece por brincar com os modelos e personalizar o Phi para os seus cenários usando o [Catálogo de Modelos Azure AI do Microsoft Foundry](https://aka.ms/phi3-azure-ai). Pode saber mais em Começar com [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Playground** -Cada modelo tem um playground dedicado para testar o modelo [Azure AI Playground](https://aka.ms/try-phi3). +**Área de Testes** +Cada modelo tem uma área de testes dedicada para experimentar o modelo [Azure AI Playground](https://aka.ms/try-phi3). -### Phi nos Modelos do GitHub +### Phi nos Modelos GitHub -Pode aprender como usar o Microsoft Phi e como construir soluções E2E nos seus diferentes dispositivos de hardware. Para experimentar o Phi por si próprio, comece por brincar com o modelo e personalizar o Phi para os seus cenários usando o [Catálogo de Modelos do GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Pode saber mais em Começar com o [Catálogo de Modelos do GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md). +Pode aprender a usar o Microsoft Phi e a construir soluções E2E nos seus diferentes dispositivos de hardware. Para experimentar o Phi, comece por brincar com o modelo e personalizar o Phi para os seus cenários usando o [Catálogo de Modelos GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Pode saber mais em Começar com [Catálogo de Modelos GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Playground** -Cada modelo tem um [playground dedicado para testar o modelo](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Área de Testes** +Cada modelo tem uma [área de testes dedicada para experimentar o modelo](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi no Hugging Face -Também pode encontrar o modelo no [Hugging Face](https://huggingface.co/microsoft). +Também pode encontrar o modelo no [Hugging Face](https://huggingface.co/microsoft) -**Playground** -[Playground Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Área de Testes** + [Área de testes Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Outros Cursos + ## 🎒 Outros Cursos -A nossa equipa produz outros cursos! Confira: +A nossa equipa produz outros cursos! Consulte: ### LangChain @@ -272,9 +275,9 @@ A nossa equipa produz outros cursos! Confira: --- -### Aprendizagem Base +### Aprendizagem Core [![ML para Iniciantes](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science para Iniciantes](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![Ciência de Dados para Iniciantes](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![IA para Iniciantes](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![Cibersegurança para Iniciantes](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Desenvolvimento Web para Iniciantes](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) @@ -284,41 +287,43 @@ A nossa equipa produz outros cursos! Confira: --- ### Série Copilot -[![Copilot para Programação Cooperativa de IA](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot para Programação em Par com IA](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot para C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Aventura Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## IA Responsável +## IA Responsável + +A Microsoft está comprometida em ajudar os nossos clientes a usar os nossos produtos de IA de forma responsável, partilhando as nossas aprendizagens e construindo parcerias baseadas na confiança através de ferramentas como Notas de Transparência e Avaliações de Impacto. Muitos destes recursos podem ser encontrados em [https://aka.ms/RAI](https://aka.ms/RAI). +A abordagem da Microsoft à IA responsável assenta nos nossos princípios de IA de justiça, fiabilidade e segurança, privacidade e segurança, inclusão, transparência e responsabilidade. + +Modelos de grande escala de linguagem natural, imagem e fala - como os usados neste exemplo - podem potencialmente comportar-se de maneiras injustas, não fiáveis ou ofensivas, causando assim prejuízos. Por favor consulte a [nota de Transparência do serviço Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) para estar informado sobre riscos e limitações. -A Microsoft compromete-se a ajudar os nossos clientes a usar os nossos produtos de IA de forma responsável, partilhando as nossas aprendizagens e construindo parcerias de confiança através de ferramentas como Notas de Transparência e Avaliações de Impacto. Muitos destes recursos podem ser encontrados em [https://aka.ms/RAI](https://aka.ms/RAI). -A abordagem da Microsoft à IA responsável baseia-se nos nossos princípios de IA de equidade, fiabilidade e segurança, privacidade e segurança, inclusão, transparência e responsabilidade. -Modelos de linguagem, imagem e voz de grande escala - como os usados neste exemplo - podem potencialmente comportar-se de formas que são injustas, pouco fiáveis ou ofensivas, causando assim danos. Por favor, consulte a [nota de transparência do serviço Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) para se informar sobre riscos e limitações. -A abordagem recomendada para mitigar estes riscos é incluir um sistema de segurança na sua arquitetura que possa detetar e prevenir comportamentos prejudiciais. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) fornece uma camada independente de proteção, capaz de detetar conteúdos prejudiciais gerados por utilizadores e por IA em aplicações e serviços. O Azure AI Content Safety inclui APIs de texto e imagem que permitem detetar material que é prejudicial. Dentro do Microsoft Foundry, o serviço Content Safety permite-lhe ver, explorar e experimentar código de exemplo para detetar conteúdos prejudiciais em diferentes modalidades. A seguinte [documentação quickstart](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) guia-o através da realização de pedidos ao serviço. +A abordagem recomendada para mitigar estes riscos é incluir um sistema de segurança na sua arquitetura que possa detetar e prevenir comportamentos nocivos. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) fornece uma camada independente de proteção, capaz de detetar conteúdos nocivos gerados por utilizadores e por IA em aplicações e serviços. O Azure AI Content Safety inclui APIs de texto e de imagem que lhe permitem detetar material nocivo. No Microsoft Foundry, o serviço Content Safety permite-lhe ver, explorar e experimentar código de exemplo para detetar conteúdo nocivo em diferentes modalidades. A seguinte documentação [quickstart](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) orienta-o na realização de pedidos ao serviço. -Outro aspeto a ter em conta é o desempenho geral da aplicação. Com aplicações multimodais e multimodelos, consideramos desempenho como o sistema a funcionar conforme espera, assim como os seus utilizadores, incluindo a não geração de resultados prejudiciais. É importante avaliar o desempenho da sua aplicação global utilizando os [avaliadores de Performance and Quality e Risk and Safety](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Também tem a capacidade de criar e avaliar com [avaliadores personalizados](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Outro aspeto a ter em conta é o desempenho geral da aplicação. Com aplicações multimodais e multimodelos, consideramos desempenho o facto do sistema funcionar como você e os seus utilizadores esperam, incluindo a não geração de saídas nocivas. É importante avaliar o desempenho da sua aplicação global utilizando os [avaliadores de Performance e Qualidade e de Risco e Segurança](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Também tem a possibilidade de criar e avaliar com [avaliadores personalizados](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Pode avaliar a sua aplicação de IA no seu ambiente de desenvolvimento usando o [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Dado um conjunto de dados de teste ou um objetivo, as gerações da sua aplicação de IA generativa são medidas quantitativamente com avaliadores incorporados ou avaliadores personalizados à sua escolha. Para começar com o azure ai evaluation sdk para avaliar o seu sistema, pode seguir o [guia quickstart](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Após executar uma avaliação, pode [visualizar os resultados no Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Pode avaliar a sua aplicação de IA no seu ambiente de desenvolvimento utilizando o [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Dado um conjunto de dados de teste ou um alvo, as gerações da sua aplicação de IA generativa são medidas quantitativamente com avaliadores integrados ou avaliadores personalizados à sua escolha. Para começar com o Azure AI Evaluation SDK para avaliar o seu sistema, pode seguir o [guia quickstart](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Depois de executar uma avaliação, pode [visualizar os resultados no Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## Marcas Comerciais +## Marcas Registadas -Este projeto pode conter marcas comerciais ou logótipos de projetos, produtos ou serviços. O uso autorizado das marcas comerciais ou logótipos da Microsoft está sujeito e deve seguir as [Diretrizes de Marca e Marca da Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -O uso das marcas comerciais ou logótipos da Microsoft em versões modificadas deste projeto não deve causar confusão nem implicar patrocínio da Microsoft. Qualquer uso de marcas comerciais ou logótipos de terceiros está sujeito às políticas desses terceiros. +Este projeto pode conter marcas registadas ou logótipos de projetos, produtos ou serviços. O uso autorizado das marcas registadas ou logótipos da Microsoft está sujeito e deve seguir as [Diretrizes de Marcas e Branding da Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +O uso das marcas registadas ou logótipos da Microsoft em versões modificadas deste projeto não deve causar confusão nem implicar patrocínio da Microsoft. Qualquer uso de marcas registadas ou logótipos de terceiros está sujeito às políticas desses terceiros. ## Obter Ajuda -Se ficar preso ou tiver alguma dúvida sobre como construir aplicações de IA, junte-se a: +Se ficar bloqueado ou tiver alguma dúvida sobre a criação de aplicações de IA, junte-se a: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Se tiver feedback sobre o produto ou encontrar erros durante o desenvolvimento, visite: +Se tiver feedback sobre produtos ou erros durante o desenvolvimento, visite: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Aviso Legal**: -Este documento foi traduzido utilizando o serviço de tradução por IA [Co-op Translator](https://github.com/Azure/co-op-translator). Embora nos esforcemos por garantir a precisão, tenha em atenção que traduções automáticas podem conter erros ou imprecisões. O documento original na sua língua nativa deve ser considerado a fonte autoritativa. Para informações críticas, recomenda-se a tradução profissional humana. Não nos responsabilizamos por quaisquer mal-entendidos ou interpretações incorretas decorrentes do uso desta tradução. +**Aviso Legal**: +Este documento foi traduzido utilizando o serviço de tradução automática [Co-op Translator](https://github.com/Azure/co-op-translator). Embora nos esforcemos pela precisão, esteja ciente de que traduções automáticas podem conter erros ou imprecisões. O documento original na sua língua nativa deve ser considerado a fonte autorizada. Para informações críticas, recomenda-se tradução profissional humana. Não nos responsabilizamos por quaisquer mal-entendidos ou interpretações incorretas resultantes da utilização desta tradução. \ No newline at end of file diff --git a/translations/pt-PT/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/pt-PT/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..595ec852e --- /dev/null +++ b/translations/pt-PT/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 RAG Híbrido com Microsoft phi-4-mini & SQLite FTS5 (SLM Zero-Cloud)\n", + "\n", + "> **Autor:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Foco:** Small Language Models (SLMs), SQLite FTS5 BM25, Embeddings Densos, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivação: O Dilema de Recuperação de Palavras-chave em SLMs Locais\n", + "Arquiteturas RAG padrão que dependem apenas de embeddings vetoriais densos frequentemente falham em recuperar tokens numéricos exatos (ex.: `2,340,000 TL`, códigos de contrato, números de conta). \n", + "Por outro lado, pesquisas lexicais esparsas (BM25) não identificam sinónimos semânticos e perguntas parafraseadas.\n", + "\n", + "Este cookbook demonstra como implementar um **motor de Recuperação Híbrido em memória de alta velocidade** combinando:\n", + "1. **Vetores Densos** (Similaridade Cosseno)\n", + "2. **Pesquisa Lexical Esparsa** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Geração de Citações Fundamentadas (`[1]`, `[2]`)** com Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Esquema SQLite Duplo (Vetores Densos + Tabela Virtual FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Exemplo de Ingestão e Benchmark de Execução\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Formulação de Prompt Fundamentada para Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Aviso Legal**:\nEste documento foi traduzido utilizando o serviço de tradução automática [Co-op Translator](https://github.com/Azure/co-op-translator). Embora nos esforcemos pela precisão, esteja ciente de que traduções automáticas podem conter erros ou imprecisões. O documento original na sua língua nativa deve ser considerado a fonte autorizada. Para informações críticas, recomenda-se tradução profissional humana. Não nos responsabilizamos por quaisquer mal-entendidos ou interpretações incorretas resultantes da utilização desta tradução.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/ro/.co-op-translator.json b/translations/ro/.co-op-translator.json index 8fc37b646..c228c354c 100644 --- a/translations/ro/.co-op-translator.json +++ b/translations/ro/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "ro" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T23:48:50+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:49:43+00:00", "source_file": "README.md", "language_code": "ro" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "ro" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:52:30+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "ro" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:32:06+00:00", diff --git a/translations/ro/README.md b/translations/ro/README.md index 790af7d71..bbbb57d3f 100644 --- a/translations/ro/README.md +++ b/translations/ro/README.md @@ -1,14 +1,14 @@ -# Carte de rețete Phi: Exemple practice cu modelele Phi de la Microsoft +# Manual Phi: Exemple Practice cu Modelele Phi de la Microsoft [![Deschide și folosește exemplele în GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Deschide în Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Deschide în Containere Dev](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![Contribuitori GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![Probleme GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Cereri de tragere GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PR-uri binevenite](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Cereri pull GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PR-uri Binevenite](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![Observatori GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![Urmăritori GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![Fork-uri GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![Stele GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) @@ -16,27 +16,27 @@ Phi este o serie de modele AI open source dezvoltate de Microsoft. -Phi este în prezent cel mai puternic și rentabil model mic de limbaj (SLM), cu repere foarte bune în multi-limbaj, raționament, generare text/chat, codare, imagini, audio și alte scenarii. +Phi este în prezent cel mai puternic și eficient model mic de limbaj (SLM), cu performanțe foarte bune în benchmark-uri pentru multi-limbaj, raționament, generare de text/chat, codare, imagini, audio și alte scenarii. -Poți implementa Phi în cloud sau pe dispozitive edge și poți construi ușor aplicații AI generative cu putere de calcul limitată. +Poți implementa Phi în cloud sau pe dispozitive edge și poți crea cu ușurință aplicații AI generative cu o putere de calcul limitată. Urmează acești pași pentru a începe să folosești aceste resurse: -1. **Fă un fork al Repozitoriului**: Click [![Fork-uri GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Clonează Repozitoriul**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Alătură-te Comunității Microsoft AI pe Discord și întâlnește experți și alți dezvoltatori**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +1. **Fă un fork al depozitului**: Click [![Fork-uri GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Clonează depozitul**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Alătură-te Comunității Microsoft AI pe Discord și întâlnește experți și dezvoltatori**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/ro/cover.eb18d1b9605d754b.webp) -### 🌐 Suport multilingv +### 🌐 Suport Multi-Limbaj -#### Susținut prin GitHub Action (automatizat și mereu actualizat) +#### Suportat prin GitHub Action (Automatizat & Întotdeauna Actualizat) -[Arabă](../ar/README.md) | [Bengaleză](../bn/README.md) | [Bulgară](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chineză (Simplificată)](../zh-CN/README.md) | [Chineză (Tradițională, Hong Kong)](../zh-HK/README.md) | [Chineză (Tradițională, Macau)](../zh-MO/README.md) | [Chineză (Tradițională, Taiwan)](../zh-TW/README.md) | [Croată](../hr/README.md) | [Cehă](../cs/README.md) | [Daneză](../da/README.md) | [Olandeză](../nl/README.md) | [Estonă](../et/README.md) | [Finlandeză](../fi/README.md) | [Franceză](../fr/README.md) | [Germană](../de/README.md) | [Greacă](../el/README.md) | [Ebraică](../he/README.md) | [Hindi](../hi/README.md) | [Maghiară](../hu/README.md) | [Indoneziană](../id/README.md) | [Italiană](../it/README.md) | [Japoneză](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Coreeană](../ko/README.md) | [Lituaniană](../lt/README.md) | [Malaysiană](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepaleză](../ne/README.md) | [Pidgin Nigerianc](../pcm/README.md) | [Norvegiană](../no/README.md) | [Persană (Farsi)](../fa/README.md) | [Poloneză](../pl/README.md) | [Portugheză (Brazilia)](../pt-BR/README.md) | [Portugheză (Portugalia)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Română](./README.md) | [Rusă](../ru/README.md) | [Sârbă (Chirilică)](../sr/README.md) | [Slovacă](../sk/README.md) | [Slovenă](../sl/README.md) | [Spaniolă](../es/README.md) | [Swahili](../sw/README.md) | [Suedeză](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thailandeză](../th/README.md) | [Turcă](../tr/README.md) | [Ucraineană](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnameză](../vi/README.md) +[Arabică](../ar/README.md) | [Bengaleză](../bn/README.md) | [Bulgară](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chineză (Simplificată)](../zh-CN/README.md) | [Chineză (Tradițională, Hong Kong)](../zh-HK/README.md) | [Chineză (Tradițională, Macau)](../zh-MO/README.md) | [Chineză (Tradițională, Taiwan)](../zh-TW/README.md) | [Croată](../hr/README.md) | [Cehă](../cs/README.md) | [Daneză](../da/README.md) | [Olandeză](../nl/README.md) | [Estoniană](../et/README.md) | [Finlandeză](../fi/README.md) | [Franceză](../fr/README.md) | [Germană](../de/README.md) | [Greacă](../el/README.md) | [Ebraică](../he/README.md) | [Hindi](../hi/README.md) | [Maghiară](../hu/README.md) | [Indoneziană](../id/README.md) | [Italiană](../it/README.md) | [Japoneză](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Coreeană](../ko/README.md) | [Lituaniană](../lt/README.md) | [Malaneză](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepaleză](../ne/README.md) | [Pidgin Nigeriana](../pcm/README.md) | [Norvegiană](../no/README.md) | [Persană (Farsi)](../fa/README.md) | [Poloneză](../pl/README.md) | [Portugheză (Brazilia)](../pt-BR/README.md) | [Portugheză (Portugalia)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Română](./README.md) | [Rusă](../ru/README.md) | [Sârbă (Chirilică)](../sr/README.md) | [Slovacă](../sk/README.md) | [Slovenă](../sl/README.md) | [Spaniolă](../es/README.md) | [Swahili](../sw/README.md) | [Suedeză](../sv/README.md) | [Tagalog (Filipineză)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thailandeză](../th/README.md) | [Turcă](../tr/README.md) | [Ucraineană](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnameză](../vi/README.md) > **Preferi să clonezi local?** > -> Acest depozit include peste 50 de traduceri în limbi diferite, ceea ce crește semnificativ dimensiunea descărcării. Pentru a clona fără traduceri, folosește sparse checkout: +> Acest depozit include peste 50 de traduceri de limbă care cresc semnificativ dimensiunea descărcării. Pentru a face clone fără traduceri, folosește sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,41 +52,41 @@ Urmează acești pași pentru a începe să folosești aceste resurse: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Aceasta îți oferă tot ce ai nevoie pentru a parcurge cursul cu o descărcare mult mai rapidă. +> Acest lucru îți oferă tot ce ai nevoie pentru a parcurge cursul cu o descărcare mult mai rapidă. ## Cuprins - Introducere - [Bine ai venit în familia Phi](./md/01.Introduction/01/01.PhiFamily.md) - - [Configurarea mediului](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Configurarea mediului tău](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Înțelegerea tehnologiilor cheie](./md/01.Introduction/01/01.Understandingtech.md) - [Siguranța AI pentru modelele Phi](./md/01.Introduction/01/01.AISafety.md) - [Suport hardware Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - [Modele Phi & Disponibilitate pe platforme](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Utilizarea Guidance-ai și Phi](./md/01.Introduction/01/01.Guidance.md) - - [Modele în GitHub Marketplace](https://github.com/marketplace/models) - - [Catalogul de modele AI Azure](https://ai.azure.com) + - [Folosirea Guidance-ai și Phi](./md/01.Introduction/01/01.Guidance.md) + - [Modele pe GitHub Marketplace](https://github.com/marketplace/models) + - [Catalogul modelelor AI Azure](https://ai.azure.com) - Inferență Phi în medii diferite - [Hugging face](./md/01.Introduction/02/01.HF.md) - [Modele GitHub](./md/01.Introduction/02/02.GitHubModel.md) - - [Catalogul modelelor Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Catalog Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [Kitul de unelte AI VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - Inferență Phi Family - - [Inferență Phi pe iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Inferență Phi pe Android](./md/01.Introduction/03/Android_Inference.md) - - [Inferență Phi pe Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Inferență Phi în iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Inferență Phi în Android](./md/01.Introduction/03/Android_Inference.md) + - [Inferență Phi în Jetson](./md/01.Introduction/03/Jetson_Inference.md) - [Inferență Phi pe AI PC](./md/01.Introduction/03/AIPC_Inference.md) - [Inferență Phi cu Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - [Inferență Phi pe server local](./md/01.Introduction/03/Local_Server_Inference.md) - [Inferență Phi pe server la distanță folosind AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - [Inferență Phi cu Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inferență Phi--Vision local](./md/01.Introduction/03/Vision_Inference.md) + - [Inferență Phi--Viziune locală](./md/01.Introduction/03/Vision_Inference.md) - [Inferență Phi cu Kaito AKS, Azure Containers (suport oficial)](./md/01.Introduction/03/Kaito_Inference.md) - [Cuantificarea Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) - [Cuantificarea Phi-3.5 / 4 folosind llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) @@ -97,214 +97,219 @@ Urmează acești pași pentru a începe să folosești aceste resurse: - Evaluarea Phi - [AI Responsabil](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry pentru evaluare](./md/01.Introduction/05/AIFoundry.md) - - [Utilizarea Promptflow pentru evaluare](./md/01.Introduction/05/Promptflow.md) + - [Folosirea Promptflow pentru evaluare](./md/01.Introduction/05/Promptflow.md) - RAG cu Azure AI Search - - [Cum să folosești Phi-4-mini și Phi-4-multimodal (RAG) cu Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Cum să folosești Phi-4-mini și Phi-4-multimodale (RAG) cu Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG hibrid Local Zero-Cloud cu SQLite FTS5 și phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Exemple de dezvoltare aplicații Phi - - Aplicații text și chat - - Mostre Phi-4 +- Exemple de dezvoltare a aplicațiilor Phi + - Aplicații Text & Chat + - Exemple Phi-4 - [📓] [Chat cu modelul Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Chat cu model local Phi-4 ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Aplicație de chat .NET Console cu Phi-4 ONNX folosind Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Mostre Phi-3 / 3.5 + - [Aplicație chat .NET Console cu Phi-4 ONNX folosind Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Exemple - [Chatbot local în browser folosind Phi3, ONNX Runtime Web și WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - Phi-3-mini interactiv și OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Construirea unui wrapper și utilizarea Phi-3 cu MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Optimizarea Modelului - Cum să optimizezi modelul Phi-3-min pentru ONNX Runtime Web cu Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Model multiplu - Phi-3-mini interactiv și OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Construirea unui wrapper și folosirea Phi-3 cu MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Optimizarea modelului - Cum să optimizezi modelul Phi-3-min pentru ONNX Runtime Web cu Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [Aplicație WinUI3 cu Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) -[Exemplu aplicație WinUI3 Multi Model AI Powered Notes](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Antrenare fină și integrare modele personalizate Phi-3 cu Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Antrenare fină și integrare modele personalizate Phi-3 cu Prompt flow în Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Evaluarea modelului Phi-3 / Phi-3.5 antrenat fin în Microsoft Foundry concentrându-se pe principiile AI responsabile de la Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Exemplu de predicție a limbajului Phi-3.5-mini-instruct (chineză/engleză)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Utilizarea GPU-ului Windows pentru a crea soluție Prompt flow cu Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Utilizarea Microsoft Phi-3.5 tflite pentru a crea aplicație Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Exemplu Q&A .NET folosind modelul local ONNX Phi-3 cu Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Aplicație console chat .NET cu Semantic Kernel și Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Exemple de cod bazate pe Azure AI Inference SDK - - Exemple Phi-4 - - [📓] [Generare cod proiect folosind Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [Fine-tuning și integrarea modelelor personalizate Phi-3 cu Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Fine-tuning și integrarea modelelor personalizate Phi-3 cu Prompt flow în Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Evaluarea modelului Phi-3 / Phi-3.5 finetuning în Microsoft Foundry cu focus pe principiile de AI responsabil ale Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Exemplu de predicție lingvistică Phi-3.5-mini-instruct (chineză/engleză)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Chatbot Phi-3.5-Instruct WebGPU RAG](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Folosirea GPU Windows pentru a crea o soluție Prompt flow cu Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Folosirea Microsoft Phi-3.5 tflite pentru a crea o aplicație Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Exemplu Q&A .NET folosind model local ONNX Phi-3 cu Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Aplicație chat consolă .NET cu Semantic Kernel și Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Exemple de cod bazate pe Azure AI Inference SDK + - Exemple Phi-4 + - [📓] [Generarea codului de proiect folosind Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Exemple Phi-3 / 3.5 - [Construiește-ți propriul Visual Studio Code GitHub Copilot Chat cu familia Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Creează-ți propriul agent Chat Copilot pentru Visual Studio Code cu Phi-3.5 de la modelele GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Creează-ți propriul agent Visual Studio Code Chat Copilot cu Phi-3.5 folosind modelele GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Exemple de raționament avansat - - Exemple Phi-4 + - Exemple Phi-4 - [📓] [Exemple Phi-4-mini-reasoning sau Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Antrenare fină Phi-4-mini-reasoning cu Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Antrenare fină Phi-4-mini-reasoning cu Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Fine-tuning Phi-4-mini-reasoning cu Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Fine-tuning Phi-4-mini-reasoning cu Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-reasoning cu modelele GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-reasoning cu modelele Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demo-uri - [Demo-uri Phi-4-mini găzduite pe Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Demo-uri Phi-4-multimodal găzduite pe Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [Demo-uri Phi-4-multimodal găzduite pe Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - Exemple Vision - - Exemple Phi-4 - - [📓] [Folosește Phi-4-multimodal pentru a citi imagini și a genera cod](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Exemple Phi-4 + - [📓] [Folosește Phi-4-multimodal pentru a citi imagini și a genera cod](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Exemple Phi-3 / 3.5 - [📓][Phi-3-vision-Text din imagine în text](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [DEMO: Phi-3 Reciclare](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Asistent vizual de limbaj - cu Phi3-Vision și OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3-vision - Asistent vizual lingvistic - cu Phi3-Vision și OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision multi-frame sau exemplu multi-imagine](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Model local ONNX Phi-3 Vision folosind Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Model local ONNX Phi-3 Vision bazat pe meniuri folosind Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Exemplu Phi-3.5 Vision cu mai multe cadre sau imagini](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision Model local ONNX folosind Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Model Phi-3 Vision local ONNX bazat pe meniuri folosind Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - Exemple Reasoning-Vision - - Phi-4-Reasoning-Vision-15B - - [📓] [Utilizarea Phi-4-Reasoning-Vision-15B pentru detectarea traversării neregulamentare](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - Phi-4-Reasoning-Vision-15B + - [📓] [Utilizarea Phi-4-Reasoning-Vision-15B pentru a detecta traversarea ilegală a străzii](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [Utilizarea Phi-4-Reasoning-Vision-15B pentru matematică](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Utilizarea Phi-4-Reasoning-Vision-15B pentru detectarea UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [Utilizarea Phi-4-Reasoning-Vision-15B pentru detectare UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Exemple matematică - Exemple Phi-4-Mini-Flash-Reasoning-Instruct [Demo matematică cu Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - Exemple audio - - Exemple Phi-4 + - Exemple Phi-4 - [📓] [Extrage transcrieri audio folosind Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Exemplu audio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Exemplu traducere vorbire Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [Aplicație console .NET folosind Phi-4-multimodal Audio pentru a analiza un fișier audio și a genera transcriere](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [Aplicație consolă .NET folosind Phi-4-multimodal Audio pentru a analiza un fișier audio și a genera transcriere](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - Exemple MOE - Exemple Phi-3 / 3.5 - - [📓] [Exemplu Social Media pentru Modelele de Mixtură a Experților (MoEs) Phi-3.5](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Construirea unui pipeline Retrieval-Augmented Generation (RAG) cu NVIDIA NIM Phi-3 MOE, Azure AI Search, și LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Modele Mixture of Experts (MoEs) Phi-3.5 exemplu rețele sociale](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Construirea unei pipeline Retrieval-Augmented Generation (RAG) cu NVIDIA NIM Phi-3 MOE, Azure AI Search și LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Exemple Apelare Funcții + - Exemple apel funcții - Exemple Phi-4 🆕 - - [📓] [Utilizarea Apelării Funcțiilor cu Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Utilizarea Apelării Funcțiilor pentru a crea agenți multipli cu Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Utilizarea Apelării Funcțiilor cu Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Utilizarea Apelării Funcțiilor cu ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Exemple amestec multimodal + - [📓] [Folosirea apelului funcției cu Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Folosirea apelului funcției pentru a crea multi-agenti cu Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Folosirea apelului funcției cu Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Folosirea apelului funcției cu ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Exemple de combinare multimodală - Exemple Phi-4 🆕 - - [📓] [Utilizarea Phi-4-multimodal ca jurnalist tehnologic](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [Aplicație console .NET folosind Phi-4-multimodal pentru analiza imaginilor](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Exemple de antrenare fină Phi - - [Scenarii de antrenare fină](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Antrenarea fină vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Antrenarea fină - Lasă Phi-3 să devină un expert în industrie](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Antrenarea fină Phi-3 cu AI Toolkit pentru VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Antrenarea fină Phi-3 cu Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Antrenarea fină Phi-3 cu Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Antrenarea fină Phi-3 cu QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Antrenarea fină Phi-3 cu Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Antrenarea fină Phi-3 cu Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Antrenarea fină cu Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Laborator practic pentru antrenarea fină cu Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [Antrenarea fină Phi-3-vision cu Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Antrenarea fină Phi-3 cu Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Antrenarea fină Phi-3-vision (suport oficial)](./md/03.FineTuning/FineTuning_Vision.md) - - [Fine-Tuning Phi-3 cu Kaito AKS , Containere Azure(Suport oficial)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Fine-Tuning Phi-3 și 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - -- Laborator Practic + - [📓] [Folosirea Phi-4-multimodal ca jurnalist tehnologic](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [Aplicație consolă .NET care folosește Phi-4-multimodal pentru a analiza imagini](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Fine-tuning exemple Phi + - [Scenarii Fine-tuning](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Fine-tuning vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Fine-tuning Lasă Phi-3 să devină expert în industrie](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Fine-tuning Phi-3 cu AI Toolkit pentru VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Fine-tuning Phi-3 cu Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Fine-tuning Phi-3 cu Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Fine-tuning Phi-3 cu QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Fine-tuning Phi-3 cu Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Fine-tuning Phi-3 cu Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Fine-tuning cu Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Fine-tuning cu Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) + - [Fine-tuning Phi-3-vision cu Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Ajustarea fină a lui Phi-3 cu Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Ajustarea fină a lui Phi-3-vision (suport oficial)](./md/03.FineTuning/FineTuning_Vision.md) + - [Ajustarea fină a lui Phi-3 cu Kaito AKS , Azure Containers (Suport oficial)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Ajustarea fină a lui Phi-3 și 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + +- Laborator practic - [Explorarea modelelor de ultimă generație: LLM-uri, SLM-uri, dezvoltare locală și altele](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Dezvăluirea potențialului NLP: Fine-Tuning cu Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Dezvăluirea potențialului NLP: Ajustare fină cu Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Lucrări academice și publicații - - [Manualele Sunt Tot Ce Ai Nevoie II: raport tehnic phi-1.5](https://arxiv.org/abs/2309.05463) - - [Raport tehnic Phi-3: Un model lingvistic foarte capabil local pe telefonul tău](https://arxiv.org/abs/2404.14219) +- Lucrări și publicații academice + - [Textbooks Are All You Need II: raport tehnic phi-1.5](https://arxiv.org/abs/2309.05463) + - [Raport tehnic Phi-3: Model lingvistic cu capacități ridicate local pe telefonul tău](https://arxiv.org/abs/2404.14219) - [Raport tehnic Phi-4](https://arxiv.org/abs/2412.08905) - - [Raport tehnic Phi-4-Mini: Modele lingvistice multimodale compacte dar puternice prin Mixtură de LoRA-uri](https://arxiv.org/abs/2503.01743) + - [Raport tehnic Phi-4-Mini: Modele multimodale compacte, dar puternice, prin Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - [Optimizarea modelelor lingvistice mici pentru apelarea funcțiilor în vehicul](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Fine-Tuning PHI-3 pentru răspunsuri la întrebări cu opțiuni multiple: Metodologie, rezultate și provocări](https://arxiv.org/abs/2501.01588) - - [Raport tehnic Phi-4-deducción](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Raport tehnic Phi-4-mini-deducción](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [(WhyPHI) Ajustarea fină PHI-3 pentru întrebări cu alegere multiplă: metodologie, rezultate și provocări](https://arxiv.org/abs/2501.01588) + - [Raport tehnic Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Raport tehnic Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Utilizarea modelelor Phi ### Phi pe Microsoft Foundry -Poți învăța cum să folosești Microsoft Phi și cum să construiești soluții end-to-end pe diverse dispozitive hardware. Pentru a experimenta Phi personal, începe prin a testa modelele și a personaliza Phi pentru scenariile tale folosind [Catalogul de modele AI Microsoft Foundry Azure](https://aka.ms/phi3-azure-ai). Poți afla mai multe de la Început cu [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Poți învăța cum să folosești Microsoft Phi și cum să construiești soluții E2E pe diferite dispozitive hardware. Pentru a experimenta Phi personal, începe să te joci cu modelele și să personalizezi Phi pentru scenariile tale folosind [Catalogul de modele Microsoft Foundry Azure AI](https://aka.ms/phi3-azure-ai). Poți afla mai multe la Începe cu [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Zona de testare** -Fiecare model are un spațiu dedicat pentru a testa modelul în [Azure AI Playground](https://aka.ms/try-phi3). +**Loc de joacă** +Fiecare model are un loc dedicat de joacă pentru a testa modelul [Azure AI Playground](https://aka.ms/try-phi3). ### Phi pe modelele GitHub -Poți învăța cum să folosești Microsoft Phi și cum să construiești soluții end-to-end pe diverse dispozitive hardware. Pentru a experimenta Phi personal, începe prin a testa modelul și a personaliza Phi pentru scenariile tale folosind [Catalogul de modele GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Poți afla mai multe de la Început cu [Catalogul de modele GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) +Poți învăța cum să folosești Microsoft Phi și cum să construiești soluții E2E pe diferite dispozitive hardware. Pentru a experimenta Phi personal, începe să te joci cu modelul și să personalizezi Phi pentru scenariile tale folosind [Catalogul modelelor GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Poți afla mai multe la Începe cu [Catalogul modelelor GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Zona de testare** -Fiecare model are un [spațiu dedicat pentru testare](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Loc de joacă** +Fiecare model are un [loc de joacă dedicat pentru a testa modelul](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi pe Hugging Face -Poți găsi modelul și pe [Hugging Face](https://huggingface.co/microsoft) +Poți găsi, de asemenea, modelul pe [Hugging Face](https://huggingface.co/microsoft) -**Zona de testare** -[Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Loc de joacă** + [Locul de joacă Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Alte cursuri + ## 🎒 Alte cursuri Echipa noastră produce și alte cursuri! Vezi: ### LangChain -[![LangChain4j pentru începători](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js pentru începători](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain pentru începători](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain4j pentru Începători](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js pentru Începători](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain pentru Începători](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- ### Azure / Edge / MCP / Agenți -[![AZD pentru începători](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI pentru începători](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP pentru începători](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Agenți AI pentru începători](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AZD pentru Începători](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI pentru Începători](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP pentru Începători](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Agenți AI pentru Începători](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Seria Generative AI -[![Generative AI pentru începători](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### Seria AI Generativă +[![AI Generativă pentru Începători](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Generativă (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![AI Generativă (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![AI Generativă (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- ### Învățare de bază -[![ML pentru începători](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science pentru începători](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI pentru începători](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Securitate cibernetică pentru începători](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Dezvoltare web pentru începători](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT pentru începători](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![Dezvoltare XR pentru începători](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![ML pentru Începători](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science pentru Începători](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI pentru Începători](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersecurity pentru Începători](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web Dev pentru Începători](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT pentru Începători](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![Dezvoltare XR pentru Începători](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- ### Seria Copilot -[![Copilot pentru programare AI asistată](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot pentru programare asistată AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot pentru C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Aventuri Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![Aventura Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## AI Responsabilă +## Inteligență artificială responsabilă -Microsoft este dedicat să ajute clienții noștri să utilizeze produsele noastre AI în mod responsabil, împărtășind învățăturile noastre și construind parteneriate bazate pe încredere prin instrumente cum ar fi Notele de Transparență și Evaluările de Impact. Multe dintre aceste resurse pot fi găsite la [https://aka.ms/RAI](https://aka.ms/RAI). +Microsoft este dedicată să ajute clienții noștri să utilizeze responsabil produsele noastre AI, împărtășind experiențele noastre și construind parteneriate bazate pe încredere prin instrumente precum Notele de Transparență și Evaluările Impactului. Multe dintre aceste resurse pot fi găsite la [https://aka.ms/RAI](https://aka.ms/RAI). Abordarea Microsoft pentru AI responsabilă se bazează pe principiile noastre AI de corectitudine, fiabilitate și siguranță, confidențialitate și securitate, incluziune, transparență și responsabilitate. -Modelele pe scară largă pentru limbaj natural, imagine și vorbire - precum cele folosite în acest exemplu - pot avea un comportament potențial nedrept, nesigur sau ofensator, cauzând astfel prejudicii. Vă rugăm să consultați [nota de transparență a serviciului Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) pentru a fi informat despre riscuri și limitări. -Abordarea recomandată pentru a reduce aceste riscuri este includerea unui sistem de siguranță în arhitectura dvs. care să poată detecta și preveni comportamentul dăunător. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) oferă un strat independent de protecție, capabil să detecteze conținut generat de utilizatori și de AI care este dăunător în aplicații și servicii. Azure AI Content Safety include API-uri pentru text și imagini care vă permit să detectați materialul dăunător. În cadrul Microsoft Foundry, serviciul Content Safety vă permite să vizualizați, să explorați și să testați coduri exemplu pentru detectarea conținutului dăunător în diferite modalități. Următoarea [documentație quickstart](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) vă ghidează prin efectuarea cererilor către serviciu. +Modelele la scară largă de limbaj natural, imagine și voce - precum cele folosite în acest exemplu - pot manifesta comportamente nedrepte, nesigure sau ofensatoare, cauzând daune. Vă rugăm să consultați [Nota de transparență a serviciului Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) pentru a fi informat despre riscuri și limitări. + + +Abordarea recomandată pentru a atenua aceste riscuri este includerea unui sistem de siguranță în arhitectura dvs. care poate detecta și preveni comportamentul dăunător. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) oferă un strat independent de protecție, capabil să detecteze conținut dăunător generat de utilizatori și de AI în aplicații și servicii. Azure AI Content Safety include API-uri pentru text și imagini care vă permit să detectați material dăunător. În cadrul Microsoft Foundry, serviciul Content Safety vă permite să vizualizați, să explorați și să încercați cod exemplu pentru detectarea conținutului dăunător în diferite modalități. Următoarea [documentație quickstart](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) vă ghidează prin efectuarea cererilor către serviciu. -Un alt aspect de luat în considerare este performanța generală a aplicației. În aplicațiile multi-modale și multi-modele, considerăm performanța ca însemnând că sistemul funcționează așa cum vă așteptați dvs. și utilizatorii dvs., inclusiv negenerarea de rezultate dăunătoare. Este important să evaluați performanța aplicației dvs. folosind [evaluatori de Performanță și Calitate și Risc și Siguranță](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Aveți, de asemenea, posibilitatea de a crea și evalua cu [evaluatori personalizați](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Un alt aspect de luat în considerare este performanța generală a aplicației. În cazul aplicațiilor multi-modale și multi-modele, considerăm că performanța înseamnă că sistemul funcționează așa cum vă așteptați dvs. și utilizatorii dvs., inclusiv faptul că nu generează rezultate dăunătoare. Este important să evaluați performanța aplicației dvs. globale folosind [evaluatori de Performanță și Calitate și de Risc și Siguranță](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). De asemenea, aveți posibilitatea de a crea și evalua utilizând [evaluatori personalizați](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Puteți evalua aplicația dvs. AI în mediul de dezvoltare folosind [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Având un set de date de test sau un obiectiv, generările aplicației dvs. AI generative sunt măsurate cantitativ cu evaluatori încorporați sau evaluatori personalizați la alegere. Pentru a începe cu azure ai evaluation sdk și a vă evalua sistemul, puteți urma [ghidul quickstart](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Odată ce executați o rulare de evaluare, puteți [vizualiza rezultatele în Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Puteți evalua aplicația dvs. AI în mediul dvs. de dezvoltare folosind [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Având fie un set de date de testare, fie un obiectiv, generațiile aplicației dvs. AI generative sunt măsurate cantitativ cu evaluatori încorporați sau evaluatori personalizați la alegere. Pentru a începe cu azure ai evaluation sdk pentru a evalua sistemul dvs., puteți urma [ghidul quickstart](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Odată ce rulați o evaluare, puteți [vizualiza rezultatele în Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Mărci comerciale -Acest proiect poate conține mărci comerciale sau logo-uri pentru proiecte, produse sau servicii. Utilizarea autorizată a mărcilor comerciale sau a logo-urilor Microsoft este supusă și trebuie să respecte [Ghidul de utilizare a mărcilor comerciale și brandurilor Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Utilizarea mărcilor comerciale sau a logo-urilor Microsoft în versiunile modificate ale acestui proiect nu trebuie să cauzeze confuzie sau să implice sponsorizarea Microsoft. Orice utilizare a mărcilor comerciale sau logo-urilor terțe este supusă politicilor acelor terțe părți. +Acest proiect poate conține mărci comerciale sau sigle pentru proiecte, produse sau servicii. Utilizarea autorizată a mărcilor comerciale sau siglelor Microsoft este supusă și trebuie să urmeze [Ghidul privind mărcile comerciale și brandul Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Utilizarea mărcilor comerciale sau siglelor Microsoft în versiuni modificate ale acestui proiect nu trebuie să cauzeze confuzie sau să implice sponsorizarea Microsoft. Orice utilizare a mărcilor comerciale sau siglelor unor terțe părți este supusă politelor acelor terțe părți. ## Obținerea ajutorului @@ -312,13 +317,13 @@ Dacă întâmpinați dificultăți sau aveți întrebări despre construirea apl [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Dacă aveți feedback despre produs sau erori în timpul construirii, vizitați: +Dacă aveți feedback despre produs sau erori în timp ce construiți, vizitați: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Avertisment**: -Acest document a fost tradus folosind serviciul de traducere AI [Co-op Translator](https://github.com/Azure/co-op-translator). Deși ne străduim pentru acuratețe, vă rugăm să rețineți că traducerile automate pot conține erori sau inexactități. Documentul original în limba sa nativă trebuie considerat sursa autorizată. Pentru informații critice, se recomandă traducerea profesională realizată de un specialist uman. Nu ne asumăm responsabilitatea pentru eventualele neînțelegeri sau interpretări greșite rezultate din utilizarea acestei traduceri. +**Declinare a responsabilității**: +Acest document a fost tradus folosind serviciul de traducere AI [Co-op Translator](https://github.com/Azure/co-op-translator). În timp ce ne străduim pentru acuratețe, vă rugăm să rețineți că traducerile automate pot conține erori sau inexactități. Documentul original în limba sa nativă trebuie considerat sursa autorizată. Pentru informații critice, se recomandă traducerea profesională realizată de un om. Nu ne asumăm responsabilitatea pentru eventualele neînțelegeri sau interpretări greșite care decurg din utilizarea acestei traduceri. \ No newline at end of file diff --git a/translations/ro/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/ro/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..75752bd51 --- /dev/null +++ b/translations/ro/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 RAG Hibrid cu Microsoft phi-4-mini și SQLite FTS5 (SLM fără cloud)\n", + "\n", + "> **Autor:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Concentrare:** Modele Lingvistice Mici (SLM-uri), SQLite FTS5 BM25, Încorporări Dense, Fuzionare Reciprocal Rank (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivație: Dilema Recunoașterii Cuvintelor-cheie în SLM-urile Locale\n", + "Arhitecturile RAG standard bazate exclusiv pe încorporări dense de vectori eșuează adesea să recupereze tokenuri numerice exacte (de ex., `2,340,000 TL`, coduri de contract, numere de cont). \n", + "În schimb, căutarea lexicală rară (BM25) ratează sinonimele semantice și întrebările parafrazate.\n", + "\n", + "Acest ghid arată cum să implementați un **motor de interogare hibrid, rapid, în memorie**, combinând:\n", + "1. **Vectori Densi** (similaritate cosinus)\n", + "2. **Căutare Lexicală Rară** (SQLite FTS5 BM25)\n", + "3. **Fuzionare Reciprocal Rank (RRF, $k=60$)**\n", + "4. **Generare de citări fundamentate (`[1]`, `[2]`)** cu Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Schema SQLite duală (Vectori densi + Tabel virtual FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Test de performanță pentru Ingestia și Executarea Probelor\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Formularea promptului bazat pe context pentru Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Declinare a responsabilității**:\nAcest document a fost tradus folosind serviciul de traducere AI [Co-op Translator](https://github.com/Azure/co-op-translator). În timp ce ne străduim pentru acuratețe, vă rugăm să rețineți că traducerile automate pot conține erori sau inexactități. Documentul original în limba sa nativă trebuie considerat sursa autorizată. Pentru informații critice, se recomandă traducerea profesională realizată de un om. Nu ne asumăm responsabilitatea pentru eventualele neînțelegeri sau interpretări greșite care decurg din utilizarea acestei traduceri.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/ru/.co-op-translator.json b/translations/ru/.co-op-translator.json index 22fe660a9..6e011e147 100644 --- a/translations/ru/.co-op-translator.json +++ b/translations/ru/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "ru" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:31:41+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:44:01+00:00", "source_file": "README.md", "language_code": "ru" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "ru" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:47:27+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "ru" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:29:32+00:00", diff --git a/translations/ru/README.md b/translations/ru/README.md index f1a1dcd34..a7da896fd 100644 --- a/translations/ru/README.md +++ b/translations/ru/README.md @@ -3,10 +3,10 @@ [![Открыть и использовать примеры в GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Открыть в Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![Контрибьюторы GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![Участники GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![Проблемы GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) [![Запросы на слияние GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![PRs приветствуются](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![Наблюдатели GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![Форки GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) @@ -14,29 +14,29 @@ [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi — это серия моделей искусственного интеллекта с открытым исходным кодом, разработанных Microsoft. +Phi — это серия моделей искусственного интеллекта с открытым исходным кодом, разработанных компанией Microsoft. -В настоящее время Phi является самой мощной и экономичной моделью малого языка (SLM), демонстрирующей очень хорошие показатели в многоязычии, логике, генерации текста/чата, программировании, изображениях, аудио и других сценариях. +В настоящее время Phi — самая мощная и экономичная малая языковая модель (SLM) с очень хорошими показателями в многоязычности, рассуждениях, генерации текста/чата, программировании, обработке изображений, аудио и других сценариях. -Вы можете развернуть Phi в облаке или на периферийных устройствах, а также легко создавать генеративные AI-приложения с ограниченной вычислительной мощностью. +Вы можете развернуть Phi в облаке или на пограничных устройствах, а также легко создавать генеративные AI-приложения с ограниченной вычислительной мощностью. -Выполните следующие шаги, чтобы начать использовать эти ресурсы: -1. **Сделайте форк репозитория**: Нажмите [![Форки GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +Следуйте этим шагам, чтобы начать работу с этими ресурсами: +1. **Форкните репозиторий**: Нажмите [![Форки GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Клонируйте репозиторий**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Присоединяйтесь к сообществу Microsoft AI Discord и знакомьтесь с экспертами и разработчиками**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Присоединитесь к сообществу Microsoft AI Discord и познакомьтесь с экспертами и разработчиками**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/ru/cover.eb18d1b9605d754b.webp) +![обложка](../../translated_images/ru/cover.eb18d1b9605d754b.webp) -### 🌐 Поддержка многоязычности +### 🌐 Многоязычная поддержка -#### Поддерживается через GitHub Action (автоматически и всегда актуально) +#### Поддерживается через GitHub Action (автоматически и всегда обновляется) -[Арабский](../ar/README.md) | [Бенгальский](../bn/README.md) | [Болгарский](../bg/README.md) | [Бирманский (Мьянма)](../my/README.md) | [Китайский (упрощенный)](../zh-CN/README.md) | [Китайский (традиционный, Гонконг)](../zh-HK/README.md) | [Китайский (традиционный, Макао)](../zh-MO/README.md) | [Китайский (традиционный, Тайвань)](../zh-TW/README.md) | [Хорватский](../hr/README.md) | [Чешский](../cs/README.md) | [Датский](../da/README.md) | [Нидерландский](../nl/README.md) | [Эстонский](../et/README.md) | [Финский](../fi/README.md) | [Французский](../fr/README.md) | [Немецкий](../de/README.md) | [Греческий](../el/README.md) | [Иврит](../he/README.md) | [Хинди](../hi/README.md) | [Венгерский](../hu/README.md) | [Индонезийский](../id/README.md) | [Итальянский](../it/README.md) | [Японский](../ja/README.md) | [Каннада](../kn/README.md) | [Кхмерский](../km/README.md) | [Корейский](../ko/README.md) | [Литовский](../lt/README.md) | [Малайский](../ms/README.md) | [Малаялам](../ml/README.md) | [Маратхи](../mr/README.md) | [Непальский](../ne/README.md) | [Нигерийский пиджин](../pcm/README.md) | [Норвежский](../no/README.md) | [Персидский (фарси)](../fa/README.md) | [Польский](../pl/README.md) | [Португальский (Бразилия)](../pt-BR/README.md) | [Португальский (Португалия)](../pt-PT/README.md) | [Пенджаби (гурмукхи)](../pa/README.md) | [Румынский](../ro/README.md) | [Русский](./README.md) | [Сербский (кириллица)](../sr/README.md) | [Словацкий](../sk/README.md) | [Словенский](../sl/README.md) | [Испанский](../es/README.md) | [Суахили](../sw/README.md) | [Шведский](../sv/README.md) | [Тагальский (филиппинский)](../tl/README.md) | [Тамильский](../ta/README.md) | [Телугу](../te/README.md) | [Тайский](../th/README.md) | [Турецкий](../tr/README.md) | [Украинский](../uk/README.md) | [Урду](../ur/README.md) | [Вьетнамский](../vi/README.md) +[Арабский](../ar/README.md) | [Бенгальский](../bn/README.md) | [Болгарский](../bg/README.md) | [Бирманский (Мьянма)](../my/README.md) | [Китайский (упрощенный)](../zh-CN/README.md) | [Китайский (традиционный, Гонконг)](../zh-HK/README.md) | [Китайский (традиционный, Макао)](../zh-MO/README.md) | [Китайский (традиционный, Тайвань)](../zh-TW/README.md) | [Хорватский](../hr/README.md) | [Чешский](../cs/README.md) | [Датский](../da/README.md) | [Нидерландский](../nl/README.md) | [Эстонский](../et/README.md) | [Финский](../fi/README.md) | [Французский](../fr/README.md) | [Немецкий](../de/README.md) | [Греческий](../el/README.md) | [Иврит](../he/README.md) | [Хинди](../hi/README.md) | [Венгерский](../hu/README.md) | [Индонезийский](../id/README.md) | [Итальянский](../it/README.md) | [Японский](../ja/README.md) | [Каннада](../kn/README.md) | [Кхмерский](../km/README.md) | [Корейский](../ko/README.md) | [Литовский](../lt/README.md) | [Малайский](../ms/README.md) | [Малаялам](../ml/README.md) | [Маратхи](../mr/README.md) | [Непальский](../ne/README.md) | [Нигерийский пиджин](../pcm/README.md) | [Норвежский](../no/README.md) | [Персидский (Фарси)](../fa/README.md) | [Польский](../pl/README.md) | [Португальский (Бразилия)](../pt-BR/README.md) | [Португальский (Португалия)](../pt-PT/README.md) | [Пенджаби (Гурмухи)](../pa/README.md) | [Румынский](../ro/README.md) | [Русский](./README.md) | [Сербский (кириллица)](../sr/README.md) | [Словацкий](../sk/README.md) | [Словенский](../sl/README.md) | [Испанский](../es/README.md) | [Суахили](../sw/README.md) | [Шведский](../sv/README.md) | [Тагальский (филиппинский)](../tl/README.md) | [Тамильский](../ta/README.md) | [Телугу](../te/README.md) | [Тайский](../th/README.md) | [Турецкий](../tr/README.md) | [Украинский](../uk/README.md) | [Урду](../ur/README.md) | [Вьетнамский](../vi/README.md) > **Предпочитаете клонировать локально?** > -> В этом репозитории более 50 языков переводов, что значительно увеличивает размер скачивания. Чтобы клонировать без переводов, используйте выборочную загрузку (sparse checkout): +> Этот репозиторий содержит более 50 переводов, что значительно увеличивает размер загрузки. Чтобы клонировать без переводов, используйте sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,201 +52,203 @@ Phi — это серия моделей искусственного интел > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Это даст вам всё необходимое для прохождения курса с гораздо более быстрой загрузкой. +> Это даст вам все необходимое для прохождения курса с гораздо более быстрой загрузкой. -## Оглавление +## Содержание - Введение - - [Добро пожаловать в семейство Phi](./md/01.Introduction/01/01.PhiFamily.md) + - [Добро пожаловать в семью Phi](./md/01.Introduction/01/01.PhiFamily.md) - [Настройка вашей среды](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Понимание ключевых технологий](./md/01.Introduction/01/01.Understandingtech.md) - [Безопасность ИИ для моделей Phi](./md/01.Introduction/01/01.AISafety.md) - [Поддержка оборудования Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - [Модели Phi и доступность на платформах](./md/01.Introduction/01/01.Edgeandcloud.md) - [Использование Guidance-ai и Phi](./md/01.Introduction/01/01.Guidance.md) - - [Модели на GitHub Marketplace](https://github.com/marketplace/models) + - [GitHub Marketplace Models](https://github.com/marketplace/models) - [Каталог моделей Azure AI](https://ai.azure.com) -- Инференс Phi в разных средах +- Вывод Phi в разных средах - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [Модели на GitHub](./md/01.Introduction/02/02.GitHubModel.md) + - [Модели GitHub](./md/01.Introduction/02/02.GitHubModel.md) - [Каталог моделей Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Инференс семейства Phi - - [Инференс Phi на iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Инференс Phi на Android](./md/01.Introduction/03/Android_Inference.md) - - [Инференс Phi на Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Инференс Phi на AI ПК](./md/01.Introduction/03/AIPC_Inference.md) - - [Инференс Phi с использованием Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Инференс Phi на локальном сервере](./md/01.Introduction/03/Local_Server_Inference.md) - - [Инференс Phi на удалённом сервере с использованием AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Инференс Phi с Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Инференс Phi--Vision локально](./md/01.Introduction/03/Vision_Inference.md) - - [Инференс Phi с Kaito AKS, Azure Containers (официальная поддержка)](./md/01.Introduction/03/Kaito_Inference.md) - -- [Квантование семейства Phi](./md/01.Introduction/04/QuantifyingPhi.md) - - [Квантование Phi-3.5 / 4 с помощью llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) +- Вывод Phi Family + - [Вывод Phi на iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Вывод Phi на Android](./md/01.Introduction/03/Android_Inference.md) + - [Вывод Phi на Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Вывод Phi на AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Вывод Phi с использованием Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Вывод Phi на локальном сервере](./md/01.Introduction/03/Local_Server_Inference.md) + - [Вывод Phi на удаленном сервере с использованием AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Вывод Phi с Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Вывод Phi–Vision локально](./md/01.Introduction/03/Vision_Inference.md) + - [Вывод Phi с Kaito AKS, Azure Containers (официальная поддержка)](./md/01.Introduction/03/Kaito_Inference.md) +- [Квантование Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) + - [Квантование Phi-3.5 / 4 с использованием llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - [Квантование Phi-3.5 / 4 с использованием расширений Generative AI для onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Квантование Phi-3.5 / 4 с помощью Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Квантование Phi-3.5 / 4 с использованием Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - [Квантование Phi-3.5 / 4 с использованием Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Оценка Phi - - [Ответственный ИИ](./md/01.Introduction/05/ResponsibleAI.md) + - [Ответственность ИИ](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry для оценки](./md/01.Introduction/05/AIFoundry.md) - [Использование Promptflow для оценки](./md/01.Introduction/05/Promptflow.md) - RAG с Azure AI Search - - [Как использовать Phi-4-mini и Phi-4-мультимодальный (RAG) с Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Как использовать Phi-4-mini и Phi-4-multimodal (RAG) с Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Безоблачный локальный гибридный RAG с SQLite FTS5 и phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Примеры разработки приложений Phi - Текстовые и чат-приложения - - Образцы Phi-4 + - Примеры Phi-4 - [📓] [Чат с моделью Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Чат с локальной моделью Phi-4 ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Чат .NET консольное приложение с Phi-4 ONNX с использованием Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Образцы Phi-3 / 3.5 - - [Локальный чатбот в браузере с использованием Phi3, ONNX Runtime Web и WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Чат .NET консольное приложение с Phi-4 ONNX, используя Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Примеры + - [Локальный чат-бот в браузере с использованием Phi3, ONNX Runtime Web и WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Мультимодель — интерактивный Phi-3-mini и OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow — создание оболочки и использование Phi-3 с MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Оптимизация модели — как оптимизировать модель Phi-3-min для ONNX Runtime Web с помощью Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [Приложение WinUI3 с Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Образец приложения WinUI3 с несколькими моделями AI Powered Notes](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Дообучение и интеграция пользовательских моделей Phi-3 с помощью Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Дообучение и интеграция пользовательских моделей Phi-3 с Prompt flow в Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Оценка дообученной модели Phi-3 / Phi-3.5 в Microsoft Foundry с акцентом на принципы ответственного ИИ Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Пример языкового предсказания Phi-3.5-mini-instruct (китайский/английский)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Веб-GPU чатбот Phi-3.5-Instruct RAG](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Использование GPU Windows для создания решения Prompt flow с Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Мульти-модель - Интерактивный Phi-3-mini и OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Создание обертки и использование Phi-3 с MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Оптимизация модели - Как оптимизировать модель Phi-3-min для ONNX Runtime Web с Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 приложение с Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[Образец приложения для заметок на WinUI3 с несколькими моделями на базе ИИ](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Тонкая настройка и интеграция пользовательских моделей Phi-3 с Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Тонкая настройка и интеграция пользовательских моделей Phi-3 с Prompt flow в Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Оценка тонко настроенной модели Phi-3 / Phi-3.5 в Microsoft Foundry с акцентом на принципы ответственного ИИ Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Пример предсказания языка Phi-3.5-mini-instruct (китайский/английский)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG чат-бот](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Использование Windows GPU для создания решения Prompt flow с Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Использование Microsoft Phi-3.5 tflite для создания Android-приложения](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Пример Q&A .NET с локальной моделью ONNX Phi-3 с использованием Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Консольное чат-приложение .NET с Semantic Kernel и Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Пример Q&A на .NET с использованием локальной ONNX модели Phi-3 через Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Консольное чат-приложение на .NET с Semantic Kernel и Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Образцы кода SDK Azure AI Inference - - Образцы Phi-4 + - Примеры на базе Azure AI Inference SDK + - Примеры Phi-4 - [📓] [Генерация кода проекта с использованием Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Образцы Phi-3 / 3.5 - - [Создайте собственный чат GitHub Copilot Visual Studio Code с Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Создайте собственного агента чата Visual Studio Code с Phi-3.5 по моделям GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - Примеры Phi-3 / 3.5 + - [Создайте собственного помощника GitHub Copilot Chat для Visual Studio Code с семейством Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Создайте собственного агента Visual Studio Code Chat Copilot с Phi-3.5 на моделях GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Примеры продвинутого рассуждения - - Образцы Phi-4 - - [📓] [Образцы Phi-4-mini-reasoning или Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Дообучение Phi-4-mini-reasoning с Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Дообучение Phi-4-mini-reasoning с Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - Примеры Phi-4 + - [📓] [Примеры Phi-4-mini-reasoning или Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Тонкая настройка Phi-4-mini-reasoning с Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Тонкая настройка Phi-4-mini-reasoning с Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-reasoning с моделями GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-reasoning с моделями Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Демонстрации - - [Демонстрации Phi-4-mini, размещённые на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Демонстрации Phi-4-multimodal, размещённые на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Образцы для Vision - - Образцы Phi-4 - - [📓] [Использование Phi-4-multimodal для чтения изображений и генерации кода](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Образцы Phi-3 / 3.5 - - [📓][Phi-3-vision — преобразование текста с изображения в текст](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Демонстрации Phi-4-mini размещены на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Демонстрации Phi-4-multimodal размещены на Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Примеры Vision + - Примеры Phi-4 + - [📓] [Используйте Phi-4-multimodal для чтения изображений и генерации кода](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Примеры Phi-3 / 3.5 + - [📓][Phi-3-vision-Image text to text](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [ДЕМО: утилизация Phi-3](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision — визуальный языковой помощник — с Phi3-Vision и OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [ДЕМОНСТРАЦИЯ: Phi-3 Recycling](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - Визуальный языковой ассистент - с Phi3-Vision и OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Пример мультифреймовой или мультиизображенческой модели Phi-3.5 Vision](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Локальная модель Phi-3 Vision ONNX с использованием Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Меню-ориентированная локальная модель Phi-3 Vision ONNX с использованием Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Phi-3.5 Vision multi-frame или multi-image пример](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision локальная ONNX модель с использованием Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Меню на базе Phi-3 Vision локальная ONNX модель с использованием Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - Образцы Reasoning-Vision - - Phi-4-Reasoning-Vision-15B - - [📓] [Использование Phi-4-Reasoning-Vision-15B для определения перехода в неположенном месте](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - Примеры Reasoning-Vision + - Phi-4-Reasoning-Vision-15B + - [📓] [Использование Phi-4-Reasoning-Vision-15B для обнаружения перехода в неположенном месте](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [Использование Phi-4-Reasoning-Vision-15B для математики](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Использование Phi-4-Reasoning-Vision-15B для определения интерфейса пользователя](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [Использование Phi-4-Reasoning-Vision-15B для обнаружения UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - Образцы для математических задач - - Образцы Phi-4-Mini-Flash-Reasoning-Instruct [Демонстрация математики с Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Примеры математики + - Примеры Phi-4-Mini-Flash-Reasoning-Instruct [Демо математики с Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - Образцы для аудио - - Образцы Phi-4 + - Примеры аудио + - Примеры Phi-4 - [📓] [Извлечение транскриптов аудио с использованием Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Аудиопример Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Пример перевода речи с помощью Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [Консольное .NET-приложение, использующее Phi-4-multimodal Audio для анализа аудиофайла и создания транскрипта](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [.NET консольное приложение, использующее Phi-4-multimodal для анализа аудиофайла и создания транскрипта](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - Образцы MOE - - Образцы Phi-3 / 3.5 - - [📓] [Модель смешивания экспертов Phi-3.5 (MoEs) — пример для социальных сетей](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - Примеры MOE + - Примеры Phi-3 / 3.5 + - [📓] [Пример модели Phi-3.5 Mixture of Experts (MoEs) для социальных сетей](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - [📓] [Создание конвейера Retrieval-Augmented Generation (RAG) с NVIDIA NIM Phi-3 MOE, Azure AI Search и LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Образцы вызова функций - - Образцы Phi-4 🆕 - - [📓] [Использование вызова функций с Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Использование вызова функций для создания мультиагентов с Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Использование вызова функций с Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Использование вызова функций с ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Образцы мультимодального смешивания - - Образцы Phi-4 🆕 + - Примеры Function Calling + - Примеры Phi-4 🆕 + - [📓] [Использование Function Calling с Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Использование Function Calling для создания мультиагентов с Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Использование Function Calling с Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Использование Function Calling с ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Примеры мультимодального смешивания + - Примеры Phi-4 🆕 - [📓] [Использование Phi-4-multimodal в роли технологического журналиста](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [Консольное .NET-приложение, использующее Phi-4-multimodal для анализа изображений](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Дообучение Phi - - [Сценарии дообучения](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Дообучение против RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Дообучение: пусть Phi-3 станет отраслевым экспертом](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Дообучение Phi-3 с AI Toolkit для VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Дообучение Phi-3 с помощью Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Дообучение Phi-3 с Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Дообучение Phi-3 с QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Дообучение Phi-3 с Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Дообучение Phi-3 с Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Дообучение с Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Практическая лаборатория по дообучению с Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [Дообучение Phi-3-vision с Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Дообучение Phi-3 с помощью Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Дообучение Phi-3-vision (официальная поддержка)](./md/03.FineTuning/FineTuning_Vision.md) - - [Тонкая настройка Phi-3 с Kaito AKS, Azure Containers(официальная поддержка)](./md/03.FineTuning/FineTuning_Kaito.md) + - [.NET консольное приложение с использованием Phi-4-multimodal для анализа изображений](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Тонкая настройка Phi Примеров + - [Сценарии тонкой настройки](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Тонкая настройка vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Тонкая настройка: Позвольте Phi-3 стать отраслевым экспертом](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Тонкая настройка Phi-3 с AI Toolkit для VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Тонкая настройка Phi-3 с Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Тонкая настройка Phi-3 с Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Тонкая настройка Phi-3 с QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Тонкая настройка Phi-3 с Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Тонкая настройка Phi-3 с Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Тонкая настройка с Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Практическая лаборатория по тонкой настройке с Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) + - [Тонкая настройка Phi-3-vision с Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Тонкая настройка Phi-3 с использованием Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Тонкая настройка Phi-3-vision (официальная поддержка)](./md/03.FineTuning/FineTuning_Vision.md) + - [Тонкая настройка Phi-3 с Kaito AKS, Azure Containers (официальная поддержка)](./md/03.FineTuning/FineTuning_Kaito.md) - [Тонкая настройка Phi-3 и 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Практическая лаборатория - [Изучение передовых моделей: LLM, SLM, локальная разработка и многое другое](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Раскрытие потенциала NLP: тонкая настройка с Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Раскрытие возможностей НЛП: тонкая настройка с Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Академические исследовательские статьи и публикации +- Научные статьи и публикации - [Textbooks Are All You Need II: технический отчет phi-1.5](https://arxiv.org/abs/2309.05463) - [Технический отчет Phi-3: мощная языковая модель локально на вашем телефоне](https://arxiv.org/abs/2404.14219) - [Технический отчет Phi-4](https://arxiv.org/abs/2412.08905) - - [Технический отчет Phi-4-Mini: компактные, но мощные мультимодальные языковые модели с помощью Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Технический отчет Phi-4-Mini: компактные, но мощные мультимодальные языковые модели с помощью смеси LoRAs](https://arxiv.org/abs/2503.01743) - [Оптимизация малых языковых моделей для вызова функций в автомобиле](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Тонкая настройка PHI-3 для многовариантного тестирования: методология, результаты и проблемы](https://arxiv.org/abs/2501.01588) - - [Технический отчет Phi-4 reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Технический отчет Phi-4-mini reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [(WhyPHI) Тонкая настройка PHI-3 для задачи с выбором нескольких вариантов ответа: методология, результаты и проблемы](https://arxiv.org/abs/2501.01588) + - [Технический отчет Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Технический отчет Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Использование моделей Phi ### Phi на Microsoft Foundry -Вы можете узнать, как использовать Microsoft Phi и как создавать сквозные решения на ваших различных аппаратных устройствах. Чтобы самостоятельно познакомиться с Phi, начните с работы с моделями и настройте Phi под ваши сценарии, используя [Каталог моделей Microsoft Foundry Azure AI](https://aka.ms/phi3-azure-ai). Подробнее вы узнаете в руководстве по началу работы с [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Вы можете научиться использовать Microsoft Phi и создавать сквозные решения для различных аппаратных устройств. Чтобы испытать Phi самостоятельно, начните с работы с моделями и настройки Phi под свои сценарии, используя [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai), подробности смотрите в разделе Быстрый старт с [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Песочница** -У каждой модели есть выделенная песочница для тестирования модели [Azure AI Playground](https://aka.ms/try-phi3). +**Площадка для экспериментов** +Для каждой модели предусмотрена отдельная тестовая площадка [Azure AI Playground](https://aka.ms/try-phi3). -### Phi на моделях GitHub +### Phi в репозитории GitHub моделей -Вы можете узнать, как использовать Microsoft Phi и как создавать сквозные решения на ваших различных аппаратных устройствах. Чтобы самостоятельно познакомиться с Phi, начните работу с моделью и настройте Phi под ваши сценарии с помощью [Каталога моделей GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Подробнее в руководстве по началу работы с [Каталогом моделей GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) +Вы можете научиться использовать Microsoft Phi и создавать сквозные решения для различных аппаратных устройств. Чтобы испытать Phi самостоятельно, начните с работы с моделью и настройки Phi под свои сценарии, используя [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo), подробности смотрите в разделе Быстрый старт с [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Песочница** -У каждой модели есть выделенная [песочница для тестирования модели](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Площадка для экспериментов** +Для каждой модели предусмотрена отдельная [площадка для тестирования модели](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi на Hugging Face Модель также доступна на [Hugging Face](https://huggingface.co/microsoft) -**Песочница** -[Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Площадка для экспериментов** + [Площадка Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Другие курсы + ## 🎒 Другие курсы Наша команда выпускает и другие курсы! Ознакомьтесь: @@ -264,61 +266,64 @@ Phi — это серия моделей искусственного интел [![AI агенты для начинающих](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Серия по генеративному ИИ + +### Серия "Генеративный ИИ" [![Генеративный ИИ для начинающих](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Генеративный ИИ (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) [![Генеративный ИИ (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) [![Генеративный ИИ (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Основное обучение + +### Основы обучения [![Машинное обучение для начинающих](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Наука о данных для начинающих](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![ИИ для начинающих](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![Кибербезопасность для начинающих](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Веб-разработка для начинающих](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![IoT для начинающих](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![Разработка XR для начинающих](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![XR-разработка для начинающих](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Серия Copilot -[![Copilot для AI парного программирования](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot для совместного программирования с ИИ](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot для C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Приключения Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![Приключения с Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## Ответственный ИИ -Microsoft стремится помочь нашим клиентам использовать наши продукты ИИ ответственно, делиться опытом и строить партнерства, основанные на доверии, с помощью таких инструментов, как заметки о прозрачности и оценки воздействия. Многие из этих ресурсов доступны по адресу [https://aka.ms/RAI](https://aka.ms/RAI). -Подход Microsoft к ответственному ИИ основан на наших принципах ИИ: справедливость, надежность и безопасность, конфиденциальность и безопасность, инклюзивность, прозрачность и подотчетность. +Microsoft стремится помогать нашим клиентам ответственно использовать наши ИИ-продукты, делиться нашим опытом и строить партнерские отношения на основе доверия с помощью таких инструментов, как Transparency Notes и Impact Assessments. Многие из этих ресурсов доступны по адресу [https://aka.ms/RAI](https://aka.ms/RAI). +Подход Microsoft к ответственному ИИ основан на наших принципах ИИ: справедливость, надежность и безопасность, конфиденциальность и безопасность, всеохватывающий подход, прозрачность и подотчетность. + +Крупномасштабные модели естественного языка, изображений и речи — как те, что используются в этом примере — могут вести себя несправедливо, ненадежно или оскорбительно, что может привести к вреду. Пожалуйста, ознакомьтесь с [заметкой о прозрачности службы Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), чтобы быть информированным о рисках и ограничениях. + -Модели для работы с естественным языком, изображениями и речью крупномасштабного уровня — такие, как используемые в этом примере — могут потенциально вести себя несправедливо, ненадежно или оскорбительно, что может привести к ущербу. Пожалуйста, ознакомьтесь с [Заметкой о прозрачности службы Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), чтобы быть в курсе рисков и ограничений. -Рекомендуемый подход к снижению этих рисков — включить в вашу архитектуру систему безопасности, которая может обнаруживать и предотвращать вредоносное поведение. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) предоставляет независимый уровень защиты, способный обнаруживать вредоносный контент, созданный пользователями и ИИ, в приложениях и сервисах. Azure AI Content Safety включает API для текста и изображений, которые позволяют обнаруживать вредоносные материалы. В Microsoft Foundry сервис Content Safety позволяет просматривать, изучать и пробовать пример кода для обнаружения вредоносного контента в различных модальностях. Следующая [документация по быстрому старту](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) проведет вас через процесс отправки запросов к сервису. +Рекомендуемый подход к снижению этих рисков заключается в включении в вашу архитектуру системы безопасности, которая может обнаруживать и предотвращать вредоносное поведение. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) предоставляет независимый уровень защиты, способный обнаруживать вредоносный контент, создаваемый пользователями и ИИ, в приложениях и сервисах. Azure AI Content Safety включает API для текста и изображений, которые позволяют обнаруживать вредоносный материал. В рамках Microsoft Foundry сервис Content Safety позволяет просматривать, исследовать и опробовать пример кода для обнаружения вредоносного контента в разных модальностях. Следующая [документация для быстрого старта](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) проведет вас через процесс выполнения запросов к сервису. -Другой аспект, который следует учитывать — общая производительность приложения. В многомодальных и многомодельных приложениях под производительностью понимается выполнение системы в соответствии с вашими и ожиданиями ваших пользователей, включая негенерацию вредоносных результатов. Важно оценить производительность вашего общего приложения с помощью [оценщиков производительности и качества, а также оценки рисков и безопасности](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). У вас также есть возможность создавать и оценивать с помощью [пользовательских оценщиков](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Другим аспектом, который следует учитывать, является общая производительность приложения. В много-модальных и много-модельных приложениях под производительностью понимается способность системы работать так, как вы и ваши пользователи ожидаете, включая отсутствие генерации вредоносных результатов. Важно оценивать производительность вашего общего приложения с помощью [оценщиков производительности, качества, рисков и безопасности](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). У вас также есть возможность создавать и проводить оценку с помощью [пользовательских оценщиков](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Вы можете оценивать ваше AI-приложение в среде разработки с помощью [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Используя тестовый набор данных или целевой показатель, ваши генерации в генеративном AI приложении количественно измеряются с помощью встроенных или пользовательских оценщиков по вашему выбору. Чтобы начать работу с azure ai evaluation sdk для оценки вашей системы, вы можете следовать [руководству по быстрому старту](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). После выполнения оценки вы можете [визуализировать результаты в Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Вы можете оценивать ваше AI-приложение в среде разработки с помощью [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Используя тестовый набор данных или целевой набор, генерации вашего генеративного AI-приложения количественно оцениваются с помощью встроенных или пользовательских оценщиков. Чтобы начать работу с azure ai evaluation sdk для оценки вашей системы, вы можете следовать [руководству для быстрого старта](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). После выполнения оценки вы можете [визуализировать результаты в Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## Торговые марки +## Товарные знаки -Этот проект может содержать торговые марки или логотипы проектов, продуктов или услуг. Авторизованное использование торговых марок или логотипов Microsoft подчиняется и должно соответствовать [Руководству Microsoft по торговым маркам и брендам](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). Использование торговых марок или логотипов Microsoft в модифицированных версиях этого проекта не должно вызывать путаницу или предполагать спонсорство Microsoft. Любое использование торговых марок или логотипов третьих сторон подчиняется политике этих третьих сторон. +Этот проект может содержать товарные знаки или логотипы проектов, продуктов или услуг. Разрешенное использование товарных знаков или логотипов Microsoft подчиняется и должно следовать [Руководству Microsoft по товарным знакам и брендам](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Использование товарных знаков или логотипов Microsoft в изменённых версиях этого проекта не должно вызывать путаницу или подразумевать спонсорство Microsoft. Любое использование товарных знаков или логотипов третьих сторон подчиняется политикам этих третьих сторон. ## Получение помощи -Если у вас возникли трудности или есть вопросы по созданию AI-приложений, присоединяйтесь к: +Если вы застряли или у вас есть вопросы по созданию AI-приложений, присоединяйтесь: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Если у вас есть отзывы о продукте или ошибки в процессе разработки, посетите: +Если у вас есть отзывы о продукте или ошибки при создании, посетите: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Отказ от ответственности**: -Этот документ был переведен с использованием сервиса автоматического перевода [Co-op Translator](https://github.com/Azure/co-op-translator). Несмотря на наши усилия обеспечить точность, имейте в виду, что автоматические переводы могут содержать ошибки или неточности. Оригинальный документ на его исходном языке следует считать авторитетным источником. Для критически важной информации рекомендуется профессиональный перевод человеком. Мы не несем ответственности за любые недоразумения или неправильные толкования, возникшие в результате использования данного перевода. +**Отказ от ответственности**: +Этот документ был переведен с использованием сервиса машинного перевода [Co-op Translator](https://github.com/Azure/co-op-translator). Несмотря на наши усилия по обеспечению точности, имейте в виду, что автоматический перевод может содержать ошибки или неточности. Оригинальный документ на его исходном языке следует считать авторитетным источником. Для получения критически важной информации рекомендуется обратиться к профессиональному человеческому переводу. Мы не несем ответственности за любые недоразумения или неправильные толкования, возникшие в результате использования этого перевода. \ No newline at end of file diff --git a/translations/ru/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/ru/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..0fe99b1e6 --- /dev/null +++ b/translations/ru/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Гибридный RAG с Microsoft phi-4-mini и SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Автор:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Фокус:** Малые языковые модели (SLM), SQLite FTS5 BM25, плотные эмбеддинги, объединение по обратному рангу (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Мотивация: дилемма точного поиска ключевых слов в локальных SLM\n", + "Стандартные архитектуры RAG, основанные исключительно на плотных векторных эмбеддингах, часто не могут извлечь точные числовые токены (например, `2,340,000 TL`, коды контрактов, номера счетов). \n", + "В то же время, разреженный лексический поиск (BM25) пропускает семантические синонимы и переформулированные вопросы.\n", + "\n", + "В этом кулинарном гайде показано, как реализовать **высокоскоростной гибридный движок поиска в памяти**, который комбинирует:\n", + "1. **Плотные векторы** (косинусное сходство)\n", + "2. **Разреженный лексический поиск** (SQLite FTS5 BM25)\n", + "3. **Объединение по обратному рангу (RRF, $k=60$)**\n", + "4. **Генерация обоснованных цитат (`[1]`, `[2]`)** с Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Двойная схема SQLite (Плотные векторы + виртуальная таблица FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Тестирование загрузки и выполнения образцов\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Формулировка обоснованных подсказок для Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Отказ от ответственности**:\nЭтот документ был переведен с использованием сервиса машинного перевода [Co-op Translator](https://github.com/Azure/co-op-translator). Несмотря на наши усилия по обеспечению точности, имейте в виду, что автоматический перевод может содержать ошибки или неточности. Оригинальный документ на его исходном языке следует считать авторитетным источником. Для получения критически важной информации рекомендуется обратиться к профессиональному человеческому переводу. Мы не несем ответственности за любые недоразумения или неправильные толкования, возникшие в результате использования этого перевода.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/sk/.co-op-translator.json b/translations/sk/.co-op-translator.json index ea66c6a1c..54f6efb78 100644 --- a/translations/sk/.co-op-translator.json +++ b/translations/sk/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "sk" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T23:47:11+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:47:23+00:00", "source_file": "README.md", "language_code": "sk" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "sk" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:52:10+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "sk" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:32:02+00:00", diff --git a/translations/sk/README.md b/translations/sk/README.md index dbed4afd6..c16ddc2bc 100644 --- a/translations/sk/README.md +++ b/translations/sk/README.md @@ -1,42 +1,42 @@ -# Phi Cookbook: Praktické príklady s modelmi Microsoft Phi +# Phi Kuchárka: Praktické príklady s Microsoft Phi modelmi -[![Otvorte a používajte príklady v GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Otvorte a používajte ukážky v GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Otvoriť v Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub prispievatelia](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub issues](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub pull-requests](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Vitamíny](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Prispievatelia GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![Problémy GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![Žiadosti o zmenu GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Vítané](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub sledovatelia](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub forky](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub hviezdy](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![Sledovatelia GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![Rozdvojenia GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![Hviezdy GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) Phi je séria open source AI modelov vyvinutých spoločnosťou Microsoft. -Phi je v súčasnosti najvýkonnejší a najnákladovo efektívnejší malý jazykový model (SLM), s veľmi dobrými výsledkami v mnohých jazykoch, uvažovaní, generovaní textu/chatov, kódovaní, obrazoch, zvuku a ďalších scenároch. +Phi je v súčasnosti najsilnejší a nákladovo najefektívnejší malý jazykový model (SLM), s veľmi dobrými výsledkami v oblasti viacjazyčnosti, uvažovania, generovania textu/rozhovorov, programovania, obrázkov, zvuku a iných scenárov. -Phi môžete nasadiť do cloudu alebo na edge zariadenia a ľahko vytvárať generatívne AI aplikácie s obmedzeným výpočtovým výkonom. +Môžete nasadiť Phi do cloudu alebo na okrajové zariadenia a ľahko vytvárať generatívne AI aplikácie s obmedzeným výpočtovým výkonom. Postupujte podľa týchto krokov, aby ste začali používať tieto zdroje: -1. **Vytvorte fork repozitára**: Kliknite [![GitHub forky](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Klonujte repozitár**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Pridajte sa do Microsoft AI Discord komunity a stretnite sa s odborníkmi a vývojármi**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +1. **Forknite repozitár**: Kliknite na [![Rozdvojenia GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Klónujte repozitár**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Pridajte sa do Microsoft AI Discord komunity a stretnite sa s odborníkmi a kolegami vývojármi**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/sk/cover.eb18d1b9605d754b.webp) +![obal](../../translated_images/sk/cover.eb18d1b9605d754b.webp) -### 🌐 Podpora viacerých jazykov +### 🌐 Podpora viacjazyčnosti -#### Podporované cez GitHub Action (automaticky a vždy aktuálne) +#### Podporované cez GitHub Action (Automatizované a vždy aktuálne) -[Arabčina](../ar/README.md) | [Bengálčina](../bn/README.md) | [Bulharčina](../bg/README.md) | [Barmčina (Myanmar)](../my/README.md) | [Čínština (zjednodušená)](../zh-CN/README.md) | [Čínština (tradičná, Hongkong)](../zh-HK/README.md) | [Čínština (tradičná, Macau)](../zh-MO/README.md) | [Čínština (tradičná, Taiwan)](../zh-TW/README.md) | [Chorvátčina](../hr/README.md) | [Čeština](../cs/README.md) | [Dánčina](../da/README.md) | [Holandčina](../nl/README.md) | [Estónčina](../et/README.md) | [Fínčina](../fi/README.md) | [Francúzština](../fr/README.md) | [Nemčina](../de/README.md) | [Gréčtina](../el/README.md) | [Hebrejčina](../he/README.md) | [Hindi](../hi/README.md) | [Maďarčina](../hu/README.md) | [Indonézština](../id/README.md) | [Taliančina](../it/README.md) | [Japončina](../ja/README.md) | [Kannadčina](../kn/README.md) | [Khmerčina](../km/README.md) | [Kórejčina](../ko/README.md) | [Litovčina](../lt/README.md) | [Malajčina](../ms/README.md) | [Malayálam](../ml/README.md) | [Maráthčina](../mr/README.md) | [Nepálčina](../ne/README.md) | [Nigerijský pidžin](../pcm/README.md) | [Nórčina](../no/README.md) | [Perzština (Farsi)](../fa/README.md) | [Poľština](../pl/README.md) | [Portugalčina (Brazília)](../pt-BR/README.md) | [Portugalčina (Portugalsko)](../pt-PT/README.md) | [Pandžábčina (Gurmukhi)](../pa/README.md) | [Rumunčina](../ro/README.md) | [Ruština](../ru/README.md) | [Srbčina (Cyrilika)](../sr/README.md) | [Slovenčina](./README.md) | [Slovinčina](../sl/README.md) | [Španielčina](../es/README.md) | [Swahilčina](../sw/README.md) | [Švédčina](../sv/README.md) | [Tagalog (Filipínčina)](../tl/README.md) | [Tamilčina](../ta/README.md) | [Telugčina](../te/README.md) | [Thajčina](../th/README.md) | [Turečtina](../tr/README.md) | [Ukrajinčina](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamčina](../vi/README.md) +[Arabčina](../ar/README.md) | [Bengálčina](../bn/README.md) | [Bulharčina](../bg/README.md) | [Barmčina (Myanmar)](../my/README.md) | [Čínština (zjednodušená)](../zh-CN/README.md) | [Čínština (tradičná, Hongkong)](../zh-HK/README.md) | [Čínština (tradičná, Macau)](../zh-MO/README.md) | [Čínština (tradičná, Taiwan)](../zh-TW/README.md) | [Chorvátčina](../hr/README.md) | [Čeština](../cs/README.md) | [Dánčina](../da/README.md) | [Holandčina](../nl/README.md) | [Estónčina](../et/README.md) | [Fínčina](../fi/README.md) | [Nemčina](../de/README.md) | [Gréčtina](../el/README.md) | [Hebrejčina](../he/README.md) | [Hindi](../hi/README.md) | [Maďarčina](../hu/README.md) | [Indonézština](../id/README.md) | [Taliančina](../it/README.md) | [Japančina](../ja/README.md) | [Kannada](../kn/README.md) | [Khmerčina](../km/README.md) | [Kórejčina](../ko/README.md) | [Litovčina](../lt/README.md) | [Malajčina](../ms/README.md) | [Malajalámčina](../ml/README.md) | [Maráthčina](../mr/README.md) | [Nepálčina](../ne/README.md) | [Nigérijský pidžin](../pcm/README.md) | [Nórčina](../no/README.md) | [Perzština (Farsi)](../fa/README.md) | [Poľština](../pl/README.md) | [Portugalčina (Brazília)](../pt-BR/README.md) | [Portugalčina (Portugalsko)](../pt-PT/README.md) | [Pandžábčina (Gurmukhi)](../pa/README.md) | [Rumunčina](../ro/README.md) | [Ruština](../ru/README.md) | [Srbochorvátčina (Cyrilika)](../sr/README.md) | [Slovenčina](./README.md) | [Slovinčina](../sl/README.md) | [Španielčina](../es/README.md) | [Svahilčina](../sw/README.md) | [Švédčina](../sv/README.md) | [Tagalog (Filipínčina)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thajčina](../th/README.md) | [Turečtina](../tr/README.md) | [Ukrajinčina](../uk/README.md) | [Urdčina](../ur/README.md) | [Vietnamčina](../vi/README.md) > **Preferujete klonovať lokálne?** > -> Tento repozitár obsahuje viac ako 50 jazykových prekladov, čo výrazne zväčšuje veľkosť sťahovania. Ak chcete klonovať bez prekladov, použite sparse checkout: +> Tento repozitár obsahuje viac ako 50 jazykových prekladov, čo výrazne zvyšuje veľkosť stiahnutia. Ak chcete klonovať bez prekladov, použite sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,47 +52,47 @@ Postupujte podľa týchto krokov, aby ste začali používať tieto zdroje: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> To vám poskytne všetko, čo potrebujete na dokončenie kurzu s oveľa rýchlejším stiahnutím. +> Toto vám poskytne všetko potrebné na dokončenie kurzu s oveľa rýchlejším sťahovaním. ## Obsah - Úvod - - [Vítame Vás v rodine Phi](./md/01.Introduction/01/01.PhiFamily.md) - - [Nastavenie vášho prostredia](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Vitajte v Phi rodine](./md/01.Introduction/01/01.PhiFamily.md) + - [Nastavenie prostredia](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Pochopenie kľúčových technológií](./md/01.Introduction/01/01.Understandingtech.md) - - [Bezpečnosť AI pre modely Phi](./md/01.Introduction/01/01.AISafety.md) + - [Bezpečnosť AI pre Phi modely](./md/01.Introduction/01/01.AISafety.md) - [Podpora hardvéru Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Modely Phi a dostupnosť na platformách](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Použitie Guidance-ai a Phi](./md/01.Introduction/01/01.Guidance.md) - - [GitHub Marketplace modely](https://github.com/marketplace/models) - - [Azure AI katalóg modelov](https://ai.azure.com) + - [Phi modely a dostupnosť na rôznych platformách](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Použitie Guidancia-ai a Phi](./md/01.Introduction/01/01.Guidance.md) + - [GitHub Marketplace Modely](https://github.com/marketplace/models) + - [Azure AI Katalóg modelov](https://ai.azure.com) -- Inference Phi v rôznych prostrediach +- Inferencia Phi v rôznych prostrediach - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [GitHub modely](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry katalóg modelov](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [GitHub Modely](./md/01.Introduction/02/02.GitHubModel.md) + - [Microsoft Foundry Katalóg modelov](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry lokálne](./md/01.Introduction/02/07.FoundryLocal.md) - -- Inference Phi Family - - [Inference Phi v iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Inference Phi v Android](./md/01.Introduction/03/Android_Inference.md) - - [Inference Phi v Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Inference Phi v AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Inference Phi s Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Inference Phi v lokálnom serveri](./md/01.Introduction/03/Local_Server_Inference.md) - - [Inference Phi na vzdialenom serveri pomocou AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Inference Phi s Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inference Phi—Vision lokálne](./md/01.Introduction/03/Vision_Inference.md) - - [Inference Phi s Kaito AKS, Azure Containers (oficiálna podpora)](./md/01.Introduction/03/Kaito_Inference.md) -- [Kvantifikácia Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) - - [Kvantifikácia Phi-3.5 / 4 pomocou llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Kvantifikácia Phi-3.5 / 4 pomocou rozšírení Generative AI pre onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Kvantifikácia Phi-3.5 / 4 pomocou Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Kvantifikácia Phi-3.5 / 4 pomocou Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Foundry Lokálne](./md/01.Introduction/02/07.FoundryLocal.md) + +- Inferencia Phi rodiny + - [Inferencia Phi v iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Inferencia Phi v Android](./md/01.Introduction/03/Android_Inference.md) + - [Inferencia Phi v Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Inferencia Phi v AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Inferencia Phi s Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Inferencia Phi na lokálnom serveri](./md/01.Introduction/03/Local_Server_Inference.md) + - [Inferencia Phi na vzdialenom serveri pomocou AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Inferencia Phi s Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Inferencia Phi--Vision lokálne](./md/01.Introduction/03/Vision_Inference.md) + - [Inferencia Phi s Kaito AKS, Azure Containers (oficiálna podpora)](./md/01.Introduction/03/Kaito_Inference.md) +- [Kvácovanie Phi rodiny](./md/01.Introduction/04/QuantifyingPhi.md) + - [Kvácovanie Phi-3.5 / 4 pomocou llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Kvácovanie Phi-3.5 / 4 pomocou Generative AI rozšírení pre onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Kvácovanie Phi-3.5 / 4 pomocou Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Kvácovanie Phi-3.5 / 4 pomocou Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Hodnotenie Phi - [Zodpovedné AI](./md/01.Introduction/05/ResponsibleAI.md) @@ -101,124 +101,127 @@ Postupujte podľa týchto krokov, aby ste začali používať tieto zdroje: - RAG s Azure AI Search - [Ako používať Phi-4-mini a Phi-4-multimodal (RAG) s Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Lokálne hybridné RAG Zero-Cloud s SQLite FTS5 a phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Ukážky vývoja aplikácií Phi - Textové a chatové aplikácie - - Phi-4 ukážky + - Ukážky Phi-4 - [📓] [Chat s Phi-4-mini ONNX modelom](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Chat s Phi-4 lokálnym ONNX modelom .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Chat .NET konzolová aplikácia s Phi-4 ONNX využívajúc Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 ukážky + - [Chat s miestnym ONNX modelom Phi-4 .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Chat .NET konzolová aplikácia s Phi-4 ONNX pomocou Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Príklady - [Lokálny chatbot v prehliadači pomocou Phi3, ONNX Runtime Web a WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - Interaktívny Phi-3-mini a OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [Viac modelov - Interaktívny Phi-3-mini a OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - Vytváranie obalu a používanie Phi-3 s MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Optimalizácia modelu - Ako optimalizovať model Phi-3-min pre ONNX Runtime Web s Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 app s Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Ukážka aplikácie WinUI3 Multi Model AI Powered Notes](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Doladenie a integrácia vlastných modelov Phi-3 s Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Doladenie a integrácia vlastných modelov Phi-3 s Prompt flow v Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Vyhodnotenie doladeného modelu Phi-3 / Phi-3.5 v Microsoft Foundry so zameraním na zásady zodpovednej AI od Microsoftu](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Ukážka jazykovej predikcie Phi-3.5-mini-instruct (čínske/anglické)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Optimalizácia modelu - Ako optimalizovať model Phi-3-min pre ONNX Runtime Web pomocou Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 aplikácia s Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 Multimodelová aplikácia poznámok poháňaná umelou inteligenciou](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Doladenie a integrácia vlastných Phi-3 modelov s Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Doladenie a integrácia vlastných Phi-3 modelov s Prompt flow vo Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Vyhodnotenie doladeného modelu Phi-3 / Phi-3.5 vo Microsoft Foundry so zameraním na zodpovedné princípy AI od Microsoftu](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct príklad predikcie jazyka (čínska/anglická)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Použitie Windows GPU na vytvorenie riešenia Prompt flow s Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Použitie Microsoft Phi-3.5 tflite na vytvorenie aplikácie pre Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET ukážka používania lokálneho ONNX modelu Phi-3 pomocou Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Použitie Microsoft Phi-3.5 tflite na vytvorenie Android aplikácie](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Q&A .NET príklad používania lokálneho ONNX Phi-3 modelu s Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - [Konzolová chat .NET aplikácia so Semantic Kernel a Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK kódové ukážky - - Phi-4 ukážky + - Ukážky Azure AI Inference SDK založené na kóde + - Príklady Phi-4 - [📓] [Generovanie kódu projektu pomocou Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 ukážky - - [Vybudujte si vlastného Visual Studio Code GitHub Copilot Chat s Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Vytvorte si vlastného Visual Studio Code Chat Copilot agenta s Phi-3.5 pomocou GitHub modelov](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - Pokročilé ukážky uvažovania - - Phi-4 ukážky - - [📓] [Phi-4-mini-uvažovanie alebo Phi-4-uvažovanie ukážky](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Doladenie Phi-4-mini-uvažovanie s Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Doladenie Phi-4-mini-uvažovanie s Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-uvažovanie s GitHub modelmi](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-uvažovanie s Microsoft Foundry modelmi](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demos - - [Phi-4-mini demo hosťované na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demo hosťované na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Vizuálne ukážky - - Phi-4 ukážky - - [📓] [Použitie Phi-4-multimodal na čítanie obrázkov a generovanie kódu](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 ukážky + - Príklady Phi-3 / 3.5 + - [Vytvorte si vlastný Visual Studio Code GitHub Copilot Chat s Microsoft Phi-3 rodinou](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Vytvorte si vlastného chat Copilot agenta v Visual Studio Code s Phi-3.5 pomocou GitHub modelov](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - Pokročilé príklady uvažovania + - Príklady Phi-4 + - [📓] [Phi-4-mini-reasoning alebo Phi-4-reasoning príklady](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Doladenie Phi-4-mini-reasoning s Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Doladenie Phi-4-mini-reasoning s Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-reasoning s GitHub modelmi](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-reasoning s Microsoft Foundry modelmi](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Ukážky + - [Phi-4-mini ukážky hosťované na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal ukážky hosťované na Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Príklady Vision + - Príklady Phi-4 + - [📓] [Použitie Phi-4-multimodal na čítanie obrázkov a generovanie kódu](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Príklady Phi-3 / 3.5 - [📓][Phi-3-vision-Obrázok text na text](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 recyklovanie](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - vizuálny jazykový asistent - s Phi3-Vision a OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [DEMO: Phi-3 Recyklácia](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - Vizuálny jazykový asistent - s Phi3-Vision a OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision viaczáberový alebo viacobrázkový príklad](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision lokálny ONNX model pomocou Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menu založený Phi-3 Vision lokálny ONNX model pomocou Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Ukážky uvažovania a videnia - - Phi-4-Uvažovanie-Videnie-15B - - [📓] [Použitie Phi-4-Uvažovanie-Videnie-15B na detekciu nezákonného prechádzania cez cestu](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Použitie Phi-4-Uvažovanie-Videnie-15B na matematiku](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Použitie Phi-4-Uvažovanie-Videnie-15B na detekciu UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - Matematické ukážky - - Phi-4-Mini-Flash-Uvažovanie-Inštrukt ukážky [Matematická ukážka s Phi-4-Mini-Flash-Uvažovanie-Inštrukt](./md/02.Application/09.Math/MathDemo.ipynb) - - - Audio ukážky - - Phi-4 ukážky - - [📓] [Vytváranie prepisu zvuku pomocou Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Ukážka audio Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Ukážka prekladu reči Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET konzolová aplikácia používajúca Phi-4-multimodal Audio na analýzu zvukového súboru a generovanie prepisu](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - MOE ukážky - - Phi-3 / 3.5 ukážky - - [📓] [Phi-3.5 Mix expertov modelov (MoEs) ukážka sociálnych médií](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Stavanie Retrieval-Augmented Generation (RAG) pipeline s NVIDIA NIM Phi-3 MOE, Azure AI Search a LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓][Phi-3.5 Vision multi-frame alebo multi-obrázkový príklad](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision Lokálny ONNX model pomocou Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Ponuka založená Phi-3 Vision Lokálny ONNX model pomocou Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Príklady Reasoning-Vision + - Phi-4-Reasoning-Vision-15B + - [📓] [Použitie Phi-4-Reasoning-Vision-15B na detekciu prechádzania mimo priechod](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Použitie Phi-4-Reasoning-Vision-15B na matematiku](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Použitie Phi-4-Reasoning-Vision-15B na detekciu UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - Príklady Matematiky + - Phi-4-Mini-Flash-Reasoning-Instruct príklady [Matematická demo s Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + + - Príklady Audio + - Phi-4 Príklady + - [📓] [Extrahovanie audio prepisov pomocou Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal Audio príklad](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal príklad pretlmočenia reči](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET konzolová aplikácia používajúca Phi-4-multimodal Audio na analýzu audio súboru a generovanie prepisu](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE Príklady + - Phi-3 / 3.5 Príklady + - [📓] [Phi-3.5 Modely Mixture of Experts (MoEs) Social Media príklad](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Vytváranie Retrieval-Augmented Generation (RAG) pipeline s NVIDIA NIM Phi-3 MOE, Azure AI Search a LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Ukážky volania funkcií - - Phi-4 ukážky 🆕 + - Príklady volania funkcií + - Phi-4 Príklady 🆕 - [📓] [Použitie volania funkcií s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - [📓] [Použitie volania funkcií na vytvorenie multi-agentov s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Použitie volania funkcií s Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [Použitie volania funkcií s ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Ukážky multimodálneho mixovania - - Phi-4 ukážky 🆕 + - Príklady multimodálneho mixovania + - Phi-4 Príklady 🆕 - [📓] [Použitie Phi-4-multimodal ako technologický novinár](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [.NET konzolová aplikácia používajúca Phi-4-multimodal na analýzu obrázkov](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Doladenie Phi ukážok +- Doladenie Phi príkladov - [Scenáre doladenia](./md/03.FineTuning/FineTuning_Scenarios.md) - [Doladenie vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - [Doladenie: Nechajte Phi-3 stať sa priemyselným expertom](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Doladenie Phi-3 s AI Toolkit pre VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Doladenie Phi-3 s Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Doladenie Phi-3 s Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Doladenie Phi-3 s QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Doladenie Phi-3 s Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Doladenie Phi-3 s Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Doladenie s Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Doladenie s Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - - [Doladenie Phi-3-vision s Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Doladenie Phi-3 pomocou AI Toolkit pre VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Doladenie Phi-3 pomocou Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Doladenie Phi-3 pomocou Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Doladenie Phi-3 pomocou QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Doladenie Phi-3 pomocou Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Doladenie Phi-3 pomocou Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Doladenie pomocou Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Doladenie pomocou Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) + - [Doladenie Phi-3-vision pomocou Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Doladenie Phi-3 s Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - [Doladenie Phi-3-vision (oficiálna podpora)](./md/03.FineTuning/FineTuning_Vision.md) - - [Doladenie Phi-3 s Kaito AKS, Azure kontajnery (oficiálna podpora)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Doladenie Phi-3 s Kaito AKS, Azure Containers (oficiálna podpora)](./md/03.FineTuning/FineTuning_Kaito.md) - [Doladenie Phi-3 a 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Praktický workshop - - [Preskúmanie najmodernejších modelov: LLM, SLM, lokálny vývoj a ďalšie](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Preskúmanie najmodernejších modelov: LLM, SLM, lokálny vývoj a viac](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - [Odomknutie potenciálu NLP: Doladenie s Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - Akademické výskumné práce a publikácie - [Textbooks Are All You Need II: technická správa phi-1.5](https://arxiv.org/abs/2309.05463) - - [Technická správa Phi-3: Vyspelý jazykový model lokálne na vašom telefóne](https://arxiv.org/abs/2404.14219) + - [Technická správa Phi-3: vysoko schopný jazykový model lokálne na vašom telefóne](https://arxiv.org/abs/2404.14219) - [Technická správa Phi-4](https://arxiv.org/abs/2412.08905) - - [Technická správa Phi-4-Mini: Kompaktné, no výkonné multimodálne jazykové modely cez Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Optimalizácia malých jazykových modelov pre volanie funkcií v vozidle](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Doladenie PHI-3 pre odpovede na otázky s viacerými možnosťami: metódy, výsledky a výzvy](https://arxiv.org/abs/2501.01588) + - [Technická správa Phi-4-Mini: Kompaktné a výkonné multimodálne jazykové modely cez zmes LoRA](https://arxiv.org/abs/2503.01743) + - [Optimalizácia malých jazykových modelov pre volanie funkcií vo vozidle](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Doladenie PHI-3 pre odpovedanie na otázky s viacerými možnosťami: metodológia, výsledky a výzvy](https://arxiv.org/abs/2501.01588) - [Technická správa Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Technická správa Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) @@ -226,28 +229,28 @@ Postupujte podľa týchto krokov, aby ste začali používať tieto zdroje: ### Phi na Microsoft Foundry -Môžete sa naučiť, ako používať Microsoft Phi a ako vytvárať E2E riešenia vo vašich rôznych hardvérových zariadeniach. Ak chcete Phi sami skúsiť, začnite hraním sa s modelmi a prispôsobením Phi pre vaše scenáre pomocou [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Viac sa dozviete v návode Začíname s [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Môžete sa naučiť, ako používať Microsoft Phi a ako vytvárať E2E riešenia vo vašich rôznych hardvérových zariadeniach. Na vyskúšanie Phi začnite hraním sa s modelmi a prispôsobovaním Phi pre vaše scenáre pomocou [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Viac sa dozviete v sekcii Začíname s [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Hracia plocha** -Každý model má vyhradenú hraciu plochu na testovanie modelu [Azure AI Playground](https://aka.ms/try-phi3). +**Hracia plocha** +Každý model má venovanú hraciu plochu na testovanie modelu [Azure AI Playground](https://aka.ms/try-phi3). ### Phi na GitHub modeloch -Môžete sa naučiť, ako používať Microsoft Phi a ako vytvárať E2E riešenia vo vašich rôznych hardvérových zariadeniach. Ak chcete Phi sami skúsiť, začnite hraním sa s modelom a prispôsobením Phi pre vaše scenáre pomocou [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Viac sa dozviete v návode Začíname s [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Môžete sa naučiť, ako používať Microsoft Phi a ako vytvárať E2E riešenia vo vašich rôznych hardvérových zariadeniach. Na vyskúšanie Phi začnite hraním sa s modelom a prispôsobovaním Phi pre vaše scenáre pomocou [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Viac sa dozviete v sekcii Začíname s [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Hracia plocha** -Každý model má vyhradenú [hraciu plochu na testovanie modelu](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Hracia plocha** +Každý model má venovanú [hraciu plochu na testovanie modelu](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi na Hugging Face -Model môžete nájsť aj na [Hugging Face](https://huggingface.co/microsoft) +Model nájdete aj na [Hugging Face](https://huggingface.co/microsoft) -**Hracia plocha** -[Hugging Chat hracia plocha](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Hracia plocha** + [Hugging Chat hracia plocha](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Ďalšie kurzy + ## 🎒 Ostatné kurzy -Náš tím pripravuje aj ďalšie kurzy! Pozrite si: +Náš tím produkuje ďalšie kurzy! Pozrite si: ### LangChain @@ -256,68 +259,71 @@ Náš tím pripravuje aj ďalšie kurzy! Pozrite si: [![LangChain pre začiatočníkov](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agenti +### Azure / Edge / MCP / Agentov [![AZD pre začiatočníkov](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI pre začiatočníkov](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP pre začiatočníkov](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI agenti pre začiatočníkov](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agentov pre začiatočníkov](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Generatívna AI séria + +### Séria Generatívnej AI [![Generatívna AI pre začiatočníkov](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Generatívna AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) [![Generatívna AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) [![Generatívna AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Základné učenie -[![ML pre začiatočníkov](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science pre začiatočníkov](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) + +### Základné vzdelávanie +[![Strojové učenie pre začiatočníkov](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Dáta veda pre začiatočníkov](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI pre začiatočníkov](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Kybernetická bezpečnosť pre začiatočníkov](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Webový vývoj pre začiatočníkov](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![Kyberbezpečnosť pre začiatočníkov](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web vývoj pre začiatočníkov](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![IoT pre začiatočníkov](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR vývoj pre začiatočníkov](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Vývoj XR pre začiatočníkov](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Copilot séria -[![Copilot pre AI párované programovanie](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) + +### Séria Copilot +[![Copilot pre AI párové programovanie](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot pre C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot dobrodružstvo](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![Copilot Dobrodružstvo](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Zodpovedná AI +## Zodpovedná AI + +Microsoft sa zaviazal pomáhať svojim zákazníkom využívať naše AI produkty zodpovedne, zdielať naše skúsenosti a budovať dôverné partnerstvá prostredníctvom nástrojov ako Transparency Notes a Impact Assessments. Mnohé z týchto zdrojov nájdete na [https://aka.ms/RAI](https://aka.ms/RAI). +Prístup Microsoftu k zodpovednej AI je založený na našich zásadách AI, ktorými sú spravodlivosť, spoľahlivosť a bezpečnosť, súkromie a bezpečnosť, inkluzívnosť, transparentnosť a zodpovednosť. + +Veľké modely spracovania prirodzeného jazyka, obrazu a reči – ako tie použité v tomto príklade – môžu potenciálne konať nespravodlivo, nespoľahlivo alebo urážlivo, čo môže viesť k škodám. Prosím, preštudujte si [oznam o transparentnosti Azure OpenAI služby](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), aby ste sa oboznámili s rizikami a obmedzeniami. -Microsoft je odhodlaný pomáhať našim zákazníkom používať naše AI produkty zodpovedne, zdieľať naše poznatky a budovať partnerstvá založené na dôvere prostredníctvom nástrojov ako Transparency Notes a Impact Assessments. Mnohé z týchto zdrojov nájdete na [https://aka.ms/RAI](https://aka.ms/RAI). -Prístup Microsoftu k zodpovednej AI je založený na našich princípoch AI: spravodlivosť, spoľahlivosť a bezpečnosť, ochrana súkromia a bezpečnosť, inkluzívnosť, transparentnosť a zodpovednosť. -Veľké modely prírodného jazyka, obrázkov a reči – ako tie používané v tejto ukážke – môžu potenciálne prejavovať správanie, ktoré je nespravodlivé, nespoľahlivé alebo urážlivé, čo môže viesť k škodám. Prosím, konzultujte [Transparency note služby Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), aby ste sa oboznámili s rizikami a obmedzeniami. -Odporúčaným prístupom k zmierneniu týchto rizík je zahrnúť do vašej architektúry bezpečnostný systém, ktorý dokáže detegovať a predchádzať škodlivému správaniu. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) poskytuje nezávislú ochrannú vrstvu, schopnú detegovať škodlivý obsah vytvorený používateľmi aj AI v aplikáciách a službách. Azure AI Content Safety zahŕňa textové a obrazové API, ktoré umožňujú detekciu škodlivého materiálu. V Microsoft Foundry služba Content Safety umožňuje prezerať, skúmať a vyskúšať ukážkový kód na detekciu škodlivého obsahu cez rôzne modality. Nasledujúca [dokumentácia rýchleho štartu](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) vás prevedie procesom vytvárania požiadaviek na službu. +Odporúčaným prístupom na zmiernenie týchto rizík je zahrnúť do vašej architektúry bezpečnostný systém, ktorý dokáže detekovať a zabrániť škodlivému správaniu. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) poskytuje nezávislú vrstvu ochrany, schopnú detekovať škodlivý obsah vytvorený používateľmi aj AI v aplikáciách a službách. Azure AI Content Safety obsahuje textové a obrazové API, ktoré vám umožnia detekovať škodlivý materiál. V rámci Microsoft Foundry vám služba Content Safety umožňuje prezerať, skúmať a vyskúšať ukážkový kód na detekciu škodlivého obsahu v rôznych modalitách. Nasledujúca [dokumentácia Quickstart](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) vás prevedie vytváraním požiadaviek na službu. -Ďalším aspektom, ktorý treba zohľadniť, je celkový výkon aplikácie. Pri multimodálnych a multimodelových aplikáciách rozumieme výkonnosťou to, že systém pracuje podľa očakávaní vás a vašich používateľov vrátane neprodukcie škodlivých výstupov. Je dôležité posúdiť výkon celej vašej aplikácie pomocou [hodnotiteľov výkonnosti, kvality, rizika a bezpečnosti](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Tiež máte možnosť vytvárať a vyhodnocovať pomocou [vlastných hodnotiteľov](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Ďalším aspektom, ktorý treba brať do úvahy, je celkový výkon aplikácie. Pri viacmodálnych a viacmodelových aplikáciách sa výkon považuje za to, že systém pracuje tak, ako vy a vaši používatelia očakávate, vrátane toho, že nevytvára škodlivé výstupy. Je dôležité zhodnotiť výkon vašej celkovej aplikácie pomocou [evaluátorov výkonu, kvality, rizika a bezpečnosti](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Tiež máte možnosť vytvárať a hodnotiť pomocou [vlastných evaluátorov](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Svoju AI aplikáciu môžete vyhodnocovať vo vašom vývojovom prostredí pomocou [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Na základe testovacej dátovej množiny alebo cieľa sú generácie vašej generatívnej AI aplikácie kvantitatívne merané vstavanými hodnotiteľmi alebo vlastnými hodnotiteľmi podľa vášho výberu. Pre začiatok práce so azure ai evaluation sdk na vyhodnotenie vášho systému môžete sledovať [sprievodcu rýchleho štartu](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Po spustení vyhodnotenia môžete [vizualizovať výsledky v Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Svoju AI aplikáciu môžete hodnotiť vo svojom vývojovom prostredí pomocou [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Na základe testovacej dátovej množiny alebo cieľa sa generované výstupy vašej generatívnej AI aplikácie kvantitatívne merajú pomocou vstavaných alebo vlastných evaluátorov podľa vášho výberu. Ak chcete začať s azure ai evaluation sdk na hodnotenie vášho systému, môžete postupovať podľa [príručky Quickstart](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Po vykonaní hodnotenia môžete [vizualizovať výsledky v Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Ochranné známky -Tento projekt môže obsahovať ochranné známky alebo logá projektov, produktov alebo služieb. Autorizované používanie ochranných známok alebo log Microsoftu podlieha a musí dodržiavať [Pravidlá používania ochranných známok a značiek Microsoftu](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). Používanie ochranných známok alebo log Microsoftu v upravených verziách tohto projektu nesmie spôsobovať zmätok ani naznačovať sponzorstvo zo strany Microsoftu. Akékoľvek použitie ochranných známok alebo log tretích strán podlieha politikám týchto tretích strán. +Tento projekt môže obsahovať ochranné známky alebo logá projektov, produktov alebo služieb. Autorizované používanie ochranných známok alebo log Microsoftu musí dodržiavať a riadiť sa [Smernicami pre ochranné známky a značky spoločnosti Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Používanie ochranných známok alebo log Microsoftu v upravených verziách tohto projektu nesmie spôsobovať zmätok ani naznačovať sponzorstvo Microsoftom. Akékoľvek použitie ochranných známok alebo log tretích strán podlieha pravidlám týchto tretích strán. -## Získanie pomoci +## Dostať pomoc -Ak máte problémy alebo otázky o tvorbe AI aplikácií, pripojte sa k: +Ak uviaznete alebo máte otázky týkajúce sa tvorby AI aplikácií, pripojte sa k: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Ak máte spätnú väzbu k produktu alebo chyby počas vývoja, navštívte: +Ak máte spätnú väzbu o produkte alebo sa stretnete s chybami pri tvorbe, navštívte: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Zrieknutie sa zodpovednosti**: -Tento dokument bol preložený pomocou AI prekladateľskej služby [Co-op Translator](https://github.com/Azure/co-op-translator). Aj keď sa snažíme o presnosť, vezmite prosím na vedomie, že automatizované preklady môžu obsahovať chyby alebo nepresnosti. Originálny dokument v jeho pôvodnom jazyku by mal byť považovaný za autoritatívny zdroj. Pre kritické informácie sa odporúča profesionálny ľudský preklad. Nezodpovedáme za akékoľvek nedorozumenia alebo nesprávne interpretácie vyplývajúce z použitia tohto prekladu. +**Vyhlásenie o zodpovednosti**: +Tento dokument bol preložený pomocou AI prekladateľskej služby [Co-op Translator](https://github.com/Azure/co-op-translator). Hoci sa snažíme o presnosť, vezmite prosím na vedomie, že automatické preklady môžu obsahovať chyby alebo nepresnosti. Pôvodný dokument v jeho natívnom jazyku by mal byť považovaný za autoritatívny zdroj. Pre kritické informácie sa odporúča profesionálny ľudský preklad. Nie sme zodpovední za žiadne nedorozumenia alebo nesprávne interpretácie vyplývajúce z použitia tohto prekladu. \ No newline at end of file diff --git a/translations/sk/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/sk/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..8c9b3f36d --- /dev/null +++ b/translations/sk/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybridný RAG s Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Autor:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Zameranie:** Malé jazykové modely (SLM), SQLite FTS5 BM25, husté vektory, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivácia: Dilema vyhľadávania kľúčových slov v lokálnych SLM\n", + "Štandardné RAG architektúry, ktoré sa spoliehajú výlučne na husté vektorové vkladače, často nezvládnu vyhľadať presné číselné tokeny (napr. `2,340,000 TL`, kódy zmlúv, čísla účtov). \n", + "Naopak, riedke lexikálne vyhľadávanie (BM25) prehliada semantické synonymá a parafrázované otázky.\n", + "\n", + "Tento návod ukazuje, ako implementovať **vysokorýchlostný hybridný vyhľadávací mechanizmus v pamäti**, ktorý kombinuje:\n", + "1. **Husté vektory** (Kosínová podobnosť)\n", + "2. **Riedke lexikálne vyhľadávanie** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Generovanie podložených citácií (`[1]`, `[2]`)** s Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Dvojitá SQLite schéma (Husté vektory + virtuálna FTS5 BM25 tabuľka)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Test vstupu vzorky a vykonávania\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Formulovanie zakotvených podnetov pre Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Vyhlásenie o zodpovednosti**:\nTento dokument bol preložený pomocou AI prekladateľskej služby [Co-op Translator](https://github.com/Azure/co-op-translator). Hoci sa snažíme o presnosť, vezmite prosím na vedomie, že automatické preklady môžu obsahovať chyby alebo nepresnosti. Pôvodný dokument v jeho natívnom jazyku by mal byť považovaný za autoritatívny zdroj. Pre kritické informácie sa odporúča profesionálny ľudský preklad. Nie sme zodpovední za žiadne nedorozumenia alebo nesprávne interpretácie vyplývajúce z použitia tohto prekladu.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/sl/.co-op-translator.json b/translations/sl/.co-op-translator.json index ebe079a39..f73546377 100644 --- a/translations/sl/.co-op-translator.json +++ b/translations/sl/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "sl" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:17:57+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:57:52+00:00", "source_file": "README.md", "language_code": "sl" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "sl" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:58:49+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "sl" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:32:24+00:00", diff --git a/translations/sl/README.md b/translations/sl/README.md index 66d52e08e..4f28396c8 100644 --- a/translations/sl/README.md +++ b/translations/sl/README.md @@ -1,43 +1,43 @@ -# Kuharska knjiga Phi: Praktični primeri z modeli Phi podjetja Microsoft +# Phi Kuharica: Praktični primeri z modeli Phi podjetja Microsoft -[![Odprite in uporabite primere v GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Odprite v Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Odpri in uporabi primere v GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Odpri v Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![Prispevki na GitHubu](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![Težave na GitHubu](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Zahteve za združitev na GitHubu](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PR dobrodošli](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub sodelavci](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub odprte težave](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub zahteve za poteg](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Dobrodošli](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![Opazovalci na GitHubu](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![Razvodi na GitHubu](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![Zvezde na GitHubu](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub opazovalci](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub vilice](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub zvezde](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi je serija odprtokodnih modelov umetne inteligence, ki jih je razvil Microsoft. +Phi je serija odprtokodnih AI modelov, ki jih razvija Microsoft. -Phi je trenutno najučinkovitejši in stroškovno najbolj učinkovit majhen jezikovni model (SLM), z zelo dobrimi rezultati v večjezičnosti, sklepanju, ustvarjanju besedil/klepetu, kodiranju, slikah, zvoku in drugih scenarijih. +Phi je trenutno najmočnejši in stroškovno najučinkovitejši majhen jezikovni model (SLM), z zelo dobrimi rezultati na področju večjezičnosti, sklepanja, generiranja besedila/klepeta, kodiranja, slik, zvoka in drugih scenarijev. -Phi lahko namestite v oblak ali na robne naprave in z omejenimi računalniškimi zmogljivostmi enostavno ustvarite generativne AI aplikacije. +Phi lahko namestite v oblak ali na edge naprave, in z malo računalniške moči lahko enostavno gradite generativne AI aplikacije. -Za začetek rabe teh virov sledite tem korakom: -1. **Razvezi repozitorij**: Kliknite [![Razvodi na GitHubu](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Kloniraj repozitorij**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Pridruži se Microsoft AI Discord skupnosti in spoznaj strokovnjake ter druge razvijalce**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +Sledite tem korakom, da začnete uporabljati te vire: +1. **Razvejite repozitorij**: Kliknite [![GitHub vilice](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Klonirajte repozitorij**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Pridružite se Microsoft AI Discord skupnosti in spoznajte strokovnjake ter druge razvijalce**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/sl/cover.eb18d1b9605d754b.webp) -### 🌐 Večjezična podpora +### 🌐 Podpora za več jezikov -#### Podprto preko GitHub Action (avtomatizirano in vedno posodobljeno) +#### Podprto preko GitHub Actions (avtomatizirano in vedno posodobljeno) -[Arabščina](../ar/README.md) | [Bengalščina](../bn/README.md) | [Bolgarščina](../bg/README.md) | [Burmanščina (Mjanmar)](../my/README.md) | [Kitajščina (poenostavljena)](../zh-CN/README.md) | [Kitajščina (tradicionalna, Hong Kong)](../zh-HK/README.md) | [Kitajščina (tradicionalna, Macao)](../zh-MO/README.md) | [Kitajščina (tradicionalna, Tajvan)](../zh-TW/README.md) | [Hrvaščina](../hr/README.md) | [Češčina](../cs/README.md) | [Danščina](../da/README.md) | [Nizozemščina](../nl/README.md) | [Estonščina](../et/README.md) | [Finščina](../fi/README.md) | [Francoščina](../fr/README.md) | [Nemščina](../de/README.md) | [Grščina](../el/README.md) | [Hebrejščina](../he/README.md) | [Hindujščina](../hi/README.md) | [Madžarščina](../hu/README.md) | [Indonezijščina](../id/README.md) | [Italijanščina](../it/README.md) | [Japonščina](../ja/README.md) | [Kannada](../kn/README.md) | [Kmerski jezik](../km/README.md) | [Korejščina](../ko/README.md) | [Litovščina](../lt/README.md) | [Malajščina](../ms/README.md) | [Malajalščina](../ml/README.md) | [Maratščina](../mr/README.md) | [Nepalščina](../ne/README.md) | [Nigerijski pidžin](../pcm/README.md) | [Norveščina](../no/README.md) | [Perzijščina (Farsi)](../fa/README.md) | [Poljščina](../pl/README.md) | [Portugalščina (Brazilija)](../pt-BR/README.md) | [Portugalščina (Portugalska)](../pt-PT/README.md) | [Pandžabščina (Gurmukhi)](../pa/README.md) | [Romunščina](../ro/README.md) | [Ruščina](../ru/README.md) | [Srbščina (cirilica)](../sr/README.md) | [Slovaščina](../sk/README.md) | [Slovenščina](./README.md) | [Španščina](../es/README.md) | [Svahili](../sw/README.md) | [Švedščina](../sv/README.md) | [Tagalog (filipinski)](../tl/README.md) | [Tamilščina](../ta/README.md) | [Telugujščina](../te/README.md) | [Tajščina](../th/README.md) | [Turščina](../tr/README.md) | [Ukrajinščina](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamščina](../vi/README.md) +[Arabščina](../ar/README.md) | [Bengalščina](../bn/README.md) | [Bolgarščina](../bg/README.md) | [Burmanščina (Mjanmar)](../my/README.md) | [Kitajščina (poenostavljena)](../zh-CN/README.md) | [Kitajščina (tradicionalna, Hong Kong)](../zh-HK/README.md) | [Kitajščina (tradicionalna, Macao)](../zh-MO/README.md) | [Kitajščina (tradicionalna, Tajvan)](../zh-TW/README.md) | [Hrvaščina](../hr/README.md) | [Češčina](../cs/README.md) | [Danščina](../da/README.md) | [Nizozemščina](../nl/README.md) | [Estonščina](../et/README.md) | [Finščina](../fi/README.md) | [Francoščina](../fr/README.md) | [Nemščina](../de/README.md) | [Grščina](../el/README.md) | [Hebrejščina](../he/README.md) | [Hindijščina](../hi/README.md) | [Madžarščina](../hu/README.md) | [Indonezijščina](../id/README.md) | [Italijanščina](../it/README.md) | [Japonščina](../ja/README.md) | [Kannada](../kn/README.md) | [Khmerščina](../km/README.md) | [Korejščina](../ko/README.md) | [Litvanščina](../lt/README.md) | [Malajščina](../ms/README.md) | [Malajalščina](../ml/README.md) | [Maratščina](../mr/README.md) | [Nepalščina](../ne/README.md) | [Nigerijski pidgin](../pcm/README.md) | [Norveščina](../no/README.md) | [Perzijski (Farsi)](../fa/README.md) | [Poljščina](../pl/README.md) | [Portugalščina (Brazilija)](../pt-BR/README.md) | [Portugalščina (Portugalska)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romunščina](../ro/README.md) | [Ruščina](../ru/README.md) | [Srbščina (cirilica)](../sr/README.md) | [Slovaščina](../sk/README.md) | [Slovenščina](./README.md) | [Španščina](../es/README.md) | [Svahili](../sw/README.md) | [Švedščina](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamilščina](../ta/README.md) | [Telugu](../te/README.md) | [Tajščina](../th/README.md) | [Turščina](../tr/README.md) | [Ukrajinščina](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamščina](../vi/README.md) > **Raje klonirate lokalno?** -> -> Ta repozitorij vključuje prevode v več kot 50 jezikov, kar znatno poveča velikost prenosa. Če želite klonirati brez prevodov, uporabite sparse checkout: -> +> +> Ta repozitorij vključuje več kot 50 jezikovnih prevodov, kar znatno poveča velikost prenosa. Če želite klonirati brez prevodov, uporabite sparse checkout: +> > **Bash / macOS / Linux:** > ```bash > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git @@ -52,47 +52,47 @@ Za začetek rabe teh virov sledite tem korakom: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Tako dobite vse, kar potrebujete za dokončanje tečaja, z veliko hitrejšo prenosno hitrostjo. +> Tako boste dobili vse potrebno za dokončanje tečaja z veliko hitrejšim prenosom. -## Vsebina +## Kazalo vsebine - Uvod - - [Dobrodošli v družino Phi](./md/01.Introduction/01/01.PhiFamily.md) - - [Nastavitev okolja](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Dobrodošli v družini Phi](./md/01.Introduction/01/01.PhiFamily.md) + - [Nastavitev vašega okolja](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Razumevanje ključnih tehnologij](./md/01.Introduction/01/01.Understandingtech.md) - [Varnost AI za modele Phi](./md/01.Introduction/01/01.AISafety.md) - - [Podpora za strojno opremo Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Modeli Phi in razpoložljivost na platformah](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Podpora strojne opreme Phi](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Modeli Phi in razpoložljivost na različnih platformah](./md/01.Introduction/01/01.Edgeandcloud.md) - [Uporaba Guidance-ai in Phi](./md/01.Introduction/01/01.Guidance.md) - - [Modeli na GitHub Marketplace](https://github.com/marketplace/models) + - [Modeli GitHub Marketplace](https://github.com/marketplace/models) - [Katalog modelov Azure AI](https://ai.azure.com) -- Inferenca Phi v različnih okoljih +- Izvajanje Phi v različnih okoljih - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [GitHub modeli](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry katalog modelov](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Modeli GitHub](./md/01.Introduction/02/02.GitHubModel.md) + - [Katalog modelov Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry lokalno](./md/01.Introduction/02/07.FoundryLocal.md) - -- Inferenca Phi družine - - [Inferenca Phi v iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Inferenca Phi v Androidu](./md/01.Introduction/03/Android_Inference.md) - - [Inferenca Phi v Jetsonu](./md/01.Introduction/03/Jetson_Inference.md) - - [Inferenca Phi na AI PCju](./md/01.Introduction/03/AIPC_Inference.md) - - [Inferenca Phi z Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Inferenca Phi na lokalnem strežniku](./md/01.Introduction/03/Local_Server_Inference.md) - - [Inferenca Phi na oddaljenem strežniku z uporabo AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Inferenca Phi z Rustom](./md/01.Introduction/03/Rust_Inference.md) - - [Inferenca Phi–Vid pred lokalno uporabo](./md/01.Introduction/03/Vision_Inference.md) - - [Inferenca Phi z Kaito AKS, Azure Containers (uradna podpora)](./md/01.Introduction/03/Kaito_Inference.md) -- [Kvantifikacija družine Phi](./md/01.Introduction/04/QuantifyingPhi.md) - - [Kvantifikacija Phi-3.5 / 4 z uporabo llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Kvantifikacija Phi-3.5 / 4 z uporabo Generative AI razširitev za onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Kvantifikacija Phi-3.5 / 4 z uporabo Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Kvantifikacija Phi-3.5 / 4 z uporabo Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) + +- Izvajanje Phi družine + - [Izvajanje Phi v iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Izvajanje Phi v Androidu](./md/01.Introduction/03/Android_Inference.md) + - [Izvajanje Phi v Jetsonu](./md/01.Introduction/03/Jetson_Inference.md) + - [Izvajanje Phi na AI računalniku](./md/01.Introduction/03/AIPC_Inference.md) + - [Izvajanje Phi z Apple MLX ogrodjem](./md/01.Introduction/03/MLX_Inference.md) + - [Izvajanje Phi na lokalnem strežniku](./md/01.Introduction/03/Local_Server_Inference.md) + - [Izvajanje Phi na oddaljenem strežniku z AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Izvajanje Phi z Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Izvajanje Phi--Vision lokalno](./md/01.Introduction/03/Vision_Inference.md) + - [Izvajanje Phi z Kaito AKS, Azure Containers (uradna podpora)](./md/01.Introduction/03/Kaito_Inference.md) +- [Kvantiacija Phi družine](./md/01.Introduction/04/QuantifyingPhi.md) + - [Kvanticiranje Phi-3.5 / 4 z uporabo llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Kvanticiranje Phi-3.5 / 4 z uporabo razširitev Generative AI za onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Kvanticiranje Phi-3.5 / 4 z uporabo Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Kvanticiranje Phi-3.5 / 4 z uporabo Apple MLX frameworka](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Evalvacija Phi - [Odgovorna AI](./md/01.Introduction/05/ResponsibleAI.md) @@ -100,154 +100,157 @@ Za začetek rabe teh virov sledite tem korakom: - [Uporaba Promptflow za evalvacijo](./md/01.Introduction/05/Promptflow.md) - RAG z Azure AI Search - - [Kako uporabljati Phi-4-mini in Phi-4-multimodal (RAG) z Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) - -- Primeri razvoja aplikacij s Phi - - Besedilne in klepetalne aplikacije - - Phi-4 primeri - - [📓] [Klepet z modelom Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Klepet z lokalnim modelom Phi-4 ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Klepet konzolna aplikacija .NET z Phi-4 ONNX z uporabo Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 primeri + - [Kako uporabiti Phi-4-mini in Phi-4-multimodal (RAG) z Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud lokalni hibridni RAG s SQLite FTS5 in phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) + +- Primeri razvoja aplikacij Phi + - Besedilo in klepet aplikacije + - Phi-4 primeri + - [📓] [Klepet z Phi-4-mini ONNX modelom](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Klepet z lokalnim Phi-4 ONNX modelom .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Klepet .NET konzolna aplikacija s Phi-4 ONNX z uporabo Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Vzorci Phi-3 / 3.5 - [Lokalni klepetalni robot v brskalniku z uporabo Phi3, ONNX Runtime Web in WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Klepet](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Večmodelni - Interaktivni Phi-3-mini in OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Izdelava ovojnice in uporaba Phi-3 z MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [OpenVino klepet](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Večmodelni - interaktivni Phi-3-mini in OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Izdelava ovoja in uporaba Phi-3 z MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [Optimizacija modela - Kako optimizirati model Phi-3-min za ONNX Runtime Web z Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [WinUI3 aplikacija s Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Vzorec večmodelnih opomb z umetno inteligenco WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Natančno prilagodite in integrirajte lastne modele Phi-3 s Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Natančno prilagodite in integrirajte lastne modele Phi-3 s Prompt flow v Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Ocenjevanje natančno prilagojenega modela Phi-3 / Phi-3.5 v Microsoft Foundry s poudarkom na načelih odgovorne umetne inteligence podjetja Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Vzorec jezikovne napovedi Phi-3.5-mini-instruct (kitajščina/angleščina)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct spletni GPU RAG klepetalnik](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + -[Vzorec aplikacije za Windows Multi Model AI Powered Notes WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Prilagoditev in integracija lastnih modelov Phi-3 z Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Prilagoditev in integracija lastnih modelov Phi-3 z Prompt flow v Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Evaluacija prilagojenega modela Phi-3 / Phi-3.5 v Microsoft Foundry z osredotočenjem na načela odgovorne umetne inteligence Microsofta](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Vzorec napovedovanja jezika Phi-3.5-mini-instruct (kitajščina/angleščina)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG klepetalni robot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Uporaba Windows GPU za ustvarjanje rešitve Prompt flow s Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Uporaba Microsoft Phi-3.5 tflite za ustvarjanje Android aplikacije](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET primer uporabe lokalnega ONNX Phi-3 modela z Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Konzolna .NET aplikacija za klepet s Semantic Kernel in Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Q&A .NET primer z uporabo lokalnega ONNX Phi-3 modela z Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Konzolna aplikacija za klepet .NET s Semantic Kernel in Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK primeri na osnovi kode + - Vzorci programske opreme Azure AI Inference SDK - Vzorci Phi-4 - - [📓] [Generiranje kode projekta s Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [📓] [Generiranje kode projekta z uporabo Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Vzorci Phi-3 / 3.5 - - [Ustvarite svoj Visual Studio Code GitHub Copilot klepet z Microsoft Phi-3 družino](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Ustvarite svoj Visual Studio Code klepetalnega copilot agenta s Phi-3.5 po modelih GitHub](./md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Ustvarite svoj Visual Studio Code GitHub Copilot klepet s Microsoft Phi-3 družino](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Ustvarite svoj Visual Studio Code klepetalni Copilot agent z Phi-3.5 z GitHub modeli](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - Napredni vzorci sklepanja + - Vzorci naprednega sklepanja - Vzorci Phi-4 - - [📓] [Vzorec Phi-4-mini-sklepanje ali Phi-4-sklepanje](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Natančno prilagajanje Phi-4-mini-sklepanje z Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Natančno prilagajanje Phi-4-mini-sklepanje z Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-sklepanje z GitHub modeli](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-sklepanje z Microsoft Foundry modeli](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demos - - [Phi-4-mini demo na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demo na Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Vzorci za vid + - [📓] [Vzorec Phi-4-mini-reasoning ali Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Prilagoditev Phi-4-mini-reasoning s Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Prilagoditev Phi-4-mini-reasoning z Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-reasoning z GitHub modeli](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-reasoning z Microsoft Foundry modeli](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Demonstracije + - [Phi-4-mini demonstracije gostovane na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal demonstracije gostovane na Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Vzorci za Vizijo - Vzorci Phi-4 - - [📓] [Uporaba Phi-4-multimodal za branje slik in ustvarjanje kode](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [📓] [Uporaba Phi-4-multimodal za branje slik in generiranje kode](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Vzorci Phi-3 / 3.5 - - [📓][Phi-3-vid-slikovni tekst v tekst](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-vid-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vid CLIP embediranje](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 Recikliranje](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vid - Vizualni jezikovni pomočnik - s Phi3-Vision in OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Vid Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Vid OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vid večslojni ali več-slikovni vzorec](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vid lokalni ONNX model z uporabo Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Meni osnovan Phi-3 Vid lokalni ONNX model z uporabo Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Razsnovanje-Vid vzorci - - Phi-4-Razsnovanje-Vid-15B - - [📓] [Uporaba Phi-4-Razsnovanje-Vid-15B za zaznavanje prečkanja ceste na mestu, kjer ni prehoda](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Uporaba Phi-4-Razsnovanje-Vid-15B za matematiko](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Uporaba Phi-4-Razsnovanje-Vid-15B za zaznavanje uporabniškega vmesnika](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - Matematični vzorci - - Phi-4-Mini-Flash-Razsnovanje-Instrukt vzorci [Matematični demo s Phi-4-Mini-Flash-Razsnovanje-Instrukt](./md/02.Application/09.Math/MathDemo.ipynb) - - - Zvočni vzorci + - [📓][Phi-3-vision - pretvorba slike v besedilo](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vision CLIP vdelava](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [DEMOS: Phi-3 recikliranje](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - vizualni jezikovni pomočnik - s Phi3-Vision in OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision vzorec za več okvirjev ali več slik](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision lokalni ONNX model z Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Meni osnovan Phi-3 Vision lokalni ONNX model z Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Vzorci sklepanja in vizije + - Phi-4-sklapanje-vizije-15B + - [📓] [Uporaba Phi-4-sklapanje-vizije-15B za zaznavo prepovedane hoje čez cesto](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Uporaba Phi-4-sklapanje-vizije-15B za matematiko](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Uporaba Phi-4-sklapanje-vizije-15B za zaznavo uporabniškega vmesnika](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - Vzorci matematike + - Vzorci Phi-4-mini-flash-sklapanje-instruct [Matematični demo s Phi-4-mini-flash-sklapanje-instruct](./md/02.Application/09.Math/MathDemo.ipynb) + + - Vzorci zvoka - Vzorci Phi-4 - [📓] [Izvleček zvočnih prepisov z uporabo Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Vzorec zvočnega posnetka Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Vzorec prevajanja govora Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET konzolna aplikacija, ki uporablja Phi-4-multimodal Audio za analizo zvočne datoteke in generiranje prepisa](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [📓] [Phi-4-multimodal vzorec za zvočne primere](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal vzorec za prevajanje govora](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET konzolna aplikacija, ki uporablja Phi-4-multimodal za analiziranje zvočne datoteke in ustvarjanje prepisa](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - MOE Vzorci + - Vzorci MOE - Vzorci Phi-3 / 3.5 - - [📓] [Phi-3.5 Mešanica strokovnjakov (MoEs) vzorec na družbenih omrežjih](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Izdelava Retrieval-Augmented Generation (RAG) cevovoda z NVIDIA NIM Phi-3 MOE, Azure AI Search in LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 model mešanice strokovnjakov (MoEs) vzorec za družbena omrežja](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Izdelava Retrieval-Augmented Generation (RAG) poteka z NVIDIA NIM Phi-3 MOE, Azure AI Search in LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Vzorci klicev funkcij + - Vzorci klicanja funkcij - Vzorci Phi-4 🆕 - - [📓] [Uporaba klicev funkcij s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Uporaba klicev funkcij za ustvarjanje več agentov s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Uporaba klicev funkcij z Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Uporaba klicev funkcij z ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Vzorci multimodalnega mešanja + - [📓] [Uporaba klicanja funkcij s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Uporaba klicanja funkcij za ustvarjanje multi-agentov s Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Uporaba klicanja funkcij z Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Uporaba klicanja funkcij z ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Vzorci mešanja multimodalnih vsebin - Vzorci Phi-4 🆕 - - [📓] [Uporaba Phi-4-multimodal kot tehnološkega novinarja](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [📓] [Uporaba Phi-4-multimodal kot tehnološki novinar](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [.NET konzolna aplikacija, ki uporablja Phi-4-multimodal za analizo slik](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Natančno prilagajanje Phi vzorcev - - [Scenariji natančnega prilagajanja](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Natančno prilagajanje vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Natančno prilagajanje naj Phi-3 postane industrijski strokovnjak](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Natančno prilagajanje Phi-3 z AI orodji za VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Natančno prilagajanje Phi-3 z Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Natančno prilagajanje Phi-3 z Loro](./md/03.FineTuning/FineTuning_Lora.md) - - [Natančno prilagajanje Phi-3 z QLoro](./md/03.FineTuning/FineTuning_Qlora.md) - - [Natančno prilagajanje Phi-3 z Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Natančno prilagajanje Phi-3 z Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Natančno prilagajanje z Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Natančno prilagajanje z Microsoft Olive praksa v laboratoriju](./md/03.FineTuning/olive-lab/readme.md) - - [Natančno prilagajanje Phi-3-vision z Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Natančno prilagajanje Phi-3 z Apple MLX okvirom](./md/03.FineTuning/FineTuning_MLX.md) - - [Natančno prilagajanje Phi-3-vision (uradna podpora)](./md/03.FineTuning/FineTuning_Vision.md) - - [Fino nastavljanje Phi-3 z Kaito AKS, Azure Containers (uradna podpora)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Fino nastavljanje Phi-3 in 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) +- Prilagoditev vzorcev Phi + - [Scenariji prilagoditve](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Prilagoditev vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Naj Phi-3 postane industrijski strokovnjak](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Prilagoditev Phi-3 z AI Toolkit za VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Prilagoditev Phi-3 z Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Prilagoditev Phi-3 z Loro](./md/03.FineTuning/FineTuning_Lora.md) + - [Prilagoditev Phi-3 z QLoro](./md/03.FineTuning/FineTuning_Qlora.md) + - [Prilagoditev Phi-3 z Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Prilagoditev Phi-3 z Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Prilagoditev z Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Prilagoditev z Microsoft Olive praktični laboratorij](./md/03.FineTuning/olive-lab/readme.md) + - [Prilagoditev Phi-3-vision z Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Prilagajanje Phi-3 z Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Prilagajanje Phi-3-vision (uradna podpora)](./md/03.FineTuning/FineTuning_Vision.md) + - [Prilagajanje Phi-3 z Kaito AKS , Azure Containers (uradna podpora)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Prilagajanje Phi-3 in 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Praktična delavnica - - [Raziskovanje najnovejših modelov: LLM, SLM, lokalni razvoj in še več](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Odklepanje potenciala NLP: fino nastavljanje z Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Raziščite najsodobnejše modele: LLM, SLM, lokalni razvoj in več](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Odklepanje NLP potenciala: prilagajanje z Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - Akademske raziskovalne naloge in publikacije - - [Vse, kar potrebujete, II: tehnično poročilo phi-1.5](https://arxiv.org/abs/2309.05463) - - [Tehnično poročilo Phi-3: zelo zmogljiv jezikovni model lokalno na vašem telefonu](https://arxiv.org/abs/2404.14219) - - [Tehnično poročilo Phi-4](https://arxiv.org/abs/2412.08905) - - [Tehnično poročilo Phi-4-Mini: kompaktni, vendar zmogljivi multimodalni jezikovni modeli preko mešanice LoRA](https://arxiv.org/abs/2503.01743) - - [Optimizacija majhnih jezikovnih modelov za funkcijsko klicanje v vozilu](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Fino nastavljanje PHI-3 za odgovarjanje na vprašanja z več izbirami: metodologija, rezultati in izzivi](https://arxiv.org/abs/2501.01588) - - [Tehnično poročilo Phi-4 za sklepanje](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Tehnično poročilo Phi-4-mini za sklepanje](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Textbooks Are All You Need II: tehnično poročilo phi-1.5](https://arxiv.org/abs/2309.05463) + - [Phi-3 tehnično poročilo: zelo zmogljiv jezikovni model lokalno na vašem telefonu](https://arxiv.org/abs/2404.14219) + - [Phi-4 tehnično poročilo](https://arxiv.org/abs/2412.08905) + - [Phi-4-Mini tehnično poročilo: kompaktni a zmogljivi multimodalni jezikovni modeli prek mešanice LoRA](https://arxiv.org/abs/2503.01743) + - [Optimizacija malih jezikovnih modelov za poklicanje funkcij v vozilu](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Prilagajanje PHI-3 za odgovarjanje na vprašanja z več izbirami: metodologija, rezultati in izzivi](https://arxiv.org/abs/2501.01588) + - [Phi-4-reasoning tehnično poročilo](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-reasoning tehnično poročilo](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Uporaba Phi modelov +## Uporaba modelov Phi ### Phi na Microsoft Foundry -Lahko se naučite, kako uporabljati Microsoft Phi in kako graditi E2E rešitve na različnih strojnih napravah. Da bi doživeli Phi sami, začnite z igranjem z modeli in prilagajanjem Phi za vaše scenarije z uporabo [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Več lahko izveste v vodiču Začetek dela z [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md). +Naučite se, kako uporabljati Microsoft Phi in kako graditi E2E rešitve na različnih strojnih napravah. Za lastno izkušnjo Phi začnite z igranjem z modeli in prilagajanjem Phi za vaše scenarije z uporabo [Microsoft Foundry Azure AI kataloga modelov](https://aka.ms/phi3-azure-ai), več lahko izveste v Pričetek dela z [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) **Igralnica** -Vsak model ima svojo namenjeno igralnico za testiranje modela [Azure AI Playground](https://aka.ms/try-phi3). +Vsak model ima namensko igralnico za testiranje modela [Azure AI Playground](https://aka.ms/try-phi3). ### Phi na GitHub modelih -Lahko se naučite, kako uporabljati Microsoft Phi in kako graditi E2E rešitve na različnih strojnih napravah. Da bi doživeli Phi sami, začnite z igranjem z modelom in prilagajanjem Phi za vaše scenarije z uporabo [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Več lahko izveste v vodiču Začetek dela z [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md). +Naučite se, kako uporabljati Microsoft Phi in kako graditi E2E rešitve na različnih strojnih napravah. Za lastno izkušnjo Phi začnite z igranjem z modelom in prilagajanjem Phi za vaše scenarije z uporabo [GitHub kataloga modelov](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo), več lahko izveste v Pričetek dela z [GitHub kataloga modelov](/md/02.QuickStart/GitHubModel_QuickStart.md) **Igralnica** -Vsak model ima namenjeno [igralnico za testiranje modela](/md/02.QuickStart/GitHubModel_QuickStart.md). +Vsak model ima namensko [igralnico za testiranje modela](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi na Hugging Face -Model lahko prav tako najdete na [Hugging Face](https://huggingface.co/microsoft). +Model lahko najdete tudi na [Hugging Face](https://huggingface.co/microsoft) **Igralnica** -[Hugging Chat igralnica](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) + [Hugging Chat igralnica](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Drugi tečaji + ## 🎒 Drugi tečaji -Naša ekipa ustvarja tudi druge tečaje! Oglejte si: +Naša ekipa pripravlja tudi druge tečaje! Oglejte si: ### LangChain @@ -260,55 +263,57 @@ Naša ekipa ustvarja tudi druge tečaje! Oglejte si: [![AZD za začetnike](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI za začetnike](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP za začetnike](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agent za začetnike](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI agenti za začetnike](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Serija generativne umetne inteligence + +### Serija Generativne AI [![Generativna AI za začetnike](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Generativna AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) [![Generativna AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) [![Generativna AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Temeljno učenje -[![ML za začetnike](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) + +### Osnovno učenje +[![Strojno učenje za začetnike](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Podatkovna znanost za začetnike](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI za začetnike](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![Kibernetska varnost za začetnike](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Spletni razvoj za začetnike](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![IoT za začetnike](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR razvoj za začetnike](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Razvoj XR za začetnike](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Serija Copilot -[![Copilot za AI programsko parjenje](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot za AI parno programiranje](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot za C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot pustolovščina](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![Copilot Avantura](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## Odgovorna umetna inteligenca -Microsoft se zavezuje pomagati našim strankam, da uporabljajo naše AI izdelke odgovorno, deliti naša spoznanja in graditi partnerske odnose, ki temeljijo na zaupanju, preko orodij, kot so Beležke transparentnosti in Ocene vpliva. Veliko teh virov lahko najdete na [https://aka.ms/RAI](https://aka.ms/RAI). -Microsoftov pristop k odgovorni AI temelji na naših načelih AI: pravičnost, zanesljivost in varnost, zasebnost in varnost, vključenost, transparentnost in odgovornost. +Microsoft si prizadeva pomagati našim strankam, da uporabljajo naše izdelke umetne inteligence odgovorno, delijo naše izkušnje in gradijo partnerske odnose, temelječe na zaupanju, prek orodij, kot so Opombe o preglednosti in Vplivne ocene. Mnogo teh virov najdete na [https://aka.ms/RAI](https://aka.ms/RAI). +Pristop Microsofta k odgovorni umetni inteligenci temelji na naših AI načelih pravičnosti, zanesljivosti in varnosti, zasebnosti in varnosti, vključenosti, preglednosti in odgovornosti. + +Veliki naravni jezikovni, slikovni in govorni modeli – kot so tisti, uporabljeni v tem vzorcu – se lahko obnašajo na načine, ki niso pravični, zanesljivi ali so žaljivi, kar lahko povzroči škodo. Prosimo, preberite [Opombo o preglednosti storitve Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), da se seznanite z tveganji in omejitvami. + -Veliki naravni jezikovni, slikovni in glasovni modeli – kot tisti, uporabljeni v tem vzorcu – se lahko potencialno vedejo na načine, ki niso pravični, zanesljivi ali so žaljivi, kar lahko povzroči škodo. Za informacije o tveganjih in omejitvah si oglejte [Opombo o transparentnosti storitve Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text). -Priporočeni pristop za omilitev teh tveganj je vključitev varnostnega sistema v vašo arhitekturo, ki lahko zazna in prepreči škodljivo vedenje. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) zagotavlja neodvisno zaščitno plast, ki je sposobna zaznati škodljivo vsebino, ki jo ustvarijo uporabniki ali AI, v aplikacijah in storitvah. Azure AI Content Safety vključuje API-je za besedilo in slike, ki omogočajo zaznavanje škodljivega gradiva. V okviru Microsoft Foundry storitev Content Safety omogoča ogled, raziskovanje in preizkus vzorčne kode za zaznavanje škodljive vsebine v različnih oblikah. Naslednja [hitri začetek dokumentacije](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) vas vodi skozi pošiljanje zahtevkov tej storitvi. +Priporočeni pristop za ublažitev teh tveganj je vključitev varnostnega sistema v vašo arhitekturo, ki lahko zazna in prepreči škodljivo vedenje. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) zagotavlja neodvisno zaščitno plast, sposobno zaznati škodljivo uporabniško generirano in AI generirano vsebino v aplikacijah in storitvah. Azure AI Content Safety vključuje API-je za besedilo in slike, ki vam omogočajo zaznavanje škodljivega gradiva. Znotraj Microsoft Foundry storitev Content Safety omogoča ogled, raziskovanje in preizkušanje vzorčnih kod za zaznavanje škodljive vsebine v različnih modalitetah. Naslednja [dokumentacija za hitro začetek](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) vas vodi skozi pošiljanje zahtevkov storitvi. -Drugi vidik, ki ga je treba upoštevati, je splošna zmogljivost aplikacije. Pri večmodalnih in večmodelnih aplikacijah zmogljivost razumemo kot to, da sistem deluje tako, kot vi in vaši uporabniki pričakujete, vključno s tem, da ne ustvarja škodljivih izhodov. Pomembno je oceniti zmogljivost vaše celotne aplikacije z uporabo [ocenjevalcev zmogljivosti, kakovosti, tveganja in varnosti](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Prav tako imate možnost ustvarjati in ocenjevati z [lastnimi ocenjevalci](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Drug vidik, ki ga je treba upoštevati, je splošna zmogljivost aplikacije. Pri aplikacijah z večmodalnostjo in več modeli razumemo zmogljivost tako, da sistem deluje, kot vi in vaši uporabniki pričakujete, vključno s tem, da ne ustvarja škodljivih izhodov. Pomembno je oceniti zmogljivost vaše celotne aplikacije z uporabo [ocenjevalcev zmogljivosti in kakovosti ter tveganja in varnosti](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Prav tako imate možnost ustvariti in ocenjevati z [lastnimi ocenjevalci](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Svojo AI aplikacijo lahko ocenjujete v vašem razvojnem okolju z uporabo [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Glede na testni nabor podatkov ali ciljno vrednost so generacije vaše generativne AI aplikacije kvantitativno izmerjene s pomočjo vgrajenih ali vaših lastnih izbranih ocenjevalcev. Za začetek z azure ai evaluation sdk za ocenjevanje vašega sistema lahko sledite [hitremu vodniku](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Ko izvedete ocenjevalno izvedbo, lahko [vizualizirate rezultate v Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Vašo AI aplikacijo lahko ocenjujete v razvojni okolju z uporabo [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Glede na testni nabor podatkov ali ciljno vrednost se generacije vaše generativne AI aplikacije kvantitativno merijo z vgrajenimi ocenjevalci ali lastnimi ocenjevalci po vaši izbiri. Za začetek z azure ai evaluation sdk za ocenjevanje vašega sistema lahko sledite [vodniku za hitro začetek](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Ko izvedete zagon ocenjevanja, lahko [vizualizirate rezultate v Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Blagovne znamke -Ta projekt lahko vsebuje blagovne znamke ali logotipe za projekte, izdelke ali storitve. Dovoljena uporaba Microsoftovih blagovnih znamk ali logotipov je predmet in mora slediti [Microsoftovim smernicam za blagovne znamke in znamčenje](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Uporaba Microsoftovih blagovnih znamk ali logotipov v spremenjenih različicah tega projekta ne sme povzročiti zmede ali nakazovati sponzorstva s strani Microsofta. Vsaka uporaba blagovnih znamk ali logotipov tretjih oseb je predmet politik teh tretjih oseb. +Ta projekt lahko vsebuje blagovne znamke ali logotipe za projekte, izdelke ali storitve. Dovoljeno uporabo Microsoftovih blagovnih znamk ali logotipov urejajo in morajo slediti [Microsoftovim smernicam za blagovne znamke in znamčenje](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Uporaba Microsoftovih blagovnih znamk ali logotipov v spremenjenih različicah tega projekta ne sme povzročiti zmede ali namigovati na sponzorstvo Microsofta. Vsaka uporaba blagovnih znamk ali logotipov tretjih oseb je predmet politik teh tretjih oseb. ## Pomoč -Če naletite na težave ali imate vprašanja o ustvarjanju AI aplikacij, se pridružite: +Če se zataknete ali imate vprašanja glede ustvarjanja AI aplikacij, se pridružite: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) @@ -320,5 +325,5 @@ Uporaba Microsoftovih blagovnih znamk ali logotipov v spremenjenih različicah t **Omejitev odgovornosti**: -Ta dokument je bil preveden z uporabo storitve za prevajanje z umetno inteligenco [Co-op Translator](https://github.com/Azure/co-op-translator). Čeprav si prizadevamo za natančnost, upoštevajte, da lahko avtomatizirani prevodi vsebujejo napake ali netočnosti. Izvirni dokument v njegovem maternem jeziku se šteje za avtoritativni vir. Za ključne informacije priporočamo strokovni človeški prevod. Ne odgovarjamo za morebitna nesporazume ali napačne interpretacije, ki izhajajo iz uporabe tega prevoda. +Ta dokument je bil preveden z uporabo AI prevajalske storitve [Co-op Translator](https://github.com/Azure/co-op-translator). Čeprav si prizadevamo za natančnost, vas prosimo, da upoštevate, da avtomatizirani prevodi lahko vsebujejo napake ali netočnosti. Izvirni dokument v njegovem izvirnem jeziku je treba obravnavati kot avtoritativni vir. Za kritične informacije je priporočljiv strokovni človeški prevod. Ne odgovarjamo za morebitna nesporazume ali napačne interpretacije, ki izhajajo iz uporabe tega prevoda. \ No newline at end of file diff --git a/translations/sl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/sl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..8ace01995 --- /dev/null +++ b/translations/sl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hibridni RAG z Microsoft phi-4-mini in SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Avtor:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Poudarek:** Majhni jezikovni modeli (SLM), SQLite FTS5 BM25, Gosta vdelava, Recipročno združevanje rangov (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivacija: Dilema priklica ključnih besed v lokalnih SLM-jih\n", + "Standardne RAG arhitekture, ki temeljijo zgolj na gostih vektorjih, pogosto ne zmorejo priklicati natančnih številskih tokenov (npr. `2,340,000 TL`, kode pogodb, številke računov). \n", + "Nasprotno pa pri redkem leksikalnem iskanju (BM25) primanjkuje semantičnih sopomenk in parafraziranih vprašanj.\n", + "\n", + "Ta priročnik prikazuje, kako implementirati **hitri hibridni iskalni mehanizem v pomnilniku**, ki združuje:\n", + "1. **Goste vektorje** (Kosinusna podobnost)\n", + "2. **Redko leksikalno iskanje** (SQLite FTS5 BM25)\n", + "3. **Reciprocno združevanje rangov (RRF, $k=60$)**\n", + "4. **Generiranje utemeljenih citatov (`[1]`, `[2]`)** z Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Dvojna shema SQLite (goste vektorje + virtualno FTS5 BM25 tabelo)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Referenčni preizkus za zajemanje vzorcev in izvajanje\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Oblikovanje utemeljenih pozivov za Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Omejitev odgovornosti**:\nTa dokument je bil preveden z uporabo AI prevajalske storitve [Co-op Translator](https://github.com/Azure/co-op-translator). Čeprav si prizadevamo za natančnost, vas prosimo, da upoštevate, da avtomatizirani prevodi lahko vsebujejo napake ali netočnosti. Izvirni dokument v njegovem izvirnem jeziku je treba obravnavati kot avtoritativni vir. Za kritične informacije je priporočljiv strokovni človeški prevod. Ne odgovarjamo za morebitna nesporazume ali napačne interpretacije, ki izhajajo iz uporabe tega prevoda.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/sr/.co-op-translator.json b/translations/sr/.co-op-translator.json index 402192cc6..717363087 100644 --- a/translations/sr/.co-op-translator.json +++ b/translations/sr/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "sr" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:13:41+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:52:27+00:00", "source_file": "README.md", "language_code": "sr" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "sr" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:58:20+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "sr" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:32:16+00:00", diff --git a/translations/sr/README.md b/translations/sr/README.md index 83419fc69..266800ed4 100644 --- a/translations/sr/README.md +++ b/translations/sr/README.md @@ -1,42 +1,42 @@ -# Phi Cookbook: Примери за праксу са Microsoft Phi моделима +# Phi Кувар: Практични примери са Мицрософтовим Phi моделима -[![Отворите и користите примере у GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Отворите у Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Отворите и користите примерке у GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Отворите у Дев контејнерима](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub сарадници](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub проблеми](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub захтеви за повлачење](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Добродошли](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![ПР су добродошли](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub посматрачи](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub пратиоци](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub форкови](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub звезде](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi је серија отворених AI модела које је развио Microsoft. +Phi је серија open source АИ модела које је развио Мицрософт. -Phi је тренутно најмоћнији и најисплативији мали језички модел (SLM), са врло добрим резултатима у мултијезику, резоновању, генерисању текста/четова, кодирању, сликама, звуку и другим сценаријима. +Phi је тренутно најмоћнији и најисплативији мали језички модел (SLM), са врло добрим резултатима у вишезначном језику, резоновању, генерисању текста/четова, кодирању, сликама, звуку и другим сценаријима. -Можете да распоредите Phi у облаку или на ивици уређаја, и лако градите генеративне AI апликације са ограниченом рачунарском снагом. +Можете распоредити Phi на облаку или на уређајима на ивици мреже, и лако градити генеративне АИ апликације са ограниченом рачунарском снагом. -Пратите ове кораке за почетак коришћења ових ресурса: -1. **Направите форк репозиторијума**: Кликните [![GitHub форкови](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +Пратите ове кораке да бисте почели користити овај ресурс: +1. **Направите форк оригиналног репозиторијума**: Кликните [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Клонирајте репозиторијум**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Придружите се Microsoft AI Discord заједници и упознајте стручњаке и друге програмере**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Придружите се Microsoft AI Discord заједници и упознајте експерте и колеге програмере**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/sr/cover.eb18d1b9605d754b.webp) ### 🌐 Подршка за више језика -#### Подржано преко GitHub Action (аутоматски и увек ажурирано) +#### Подржано путем GitHub Action (аутоматски и увек ажурирано) [Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](./README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **Више волите локално клонирање?** +> **Волите да клонирате локално?** > -> Овај репозиторијум укључује преко 50 превода на језике што значајно повећава величину преузимања. Да бисте клонирали без превода, користите sparse checkout: +> Овај репозиторијум садржи преко 50 превода, што значајно повећава величину преузимања. Да бисте клонирали без превода, користите sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,23 +52,23 @@ Phi је тренутно најмоћнији и најисплативији > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Ово вам даје све што вам је потребно за завршетак курса са много бржим преузимањем. +> Ово вам пружа све што вам је потребно за завршетак курса са много бржим преузимањем. ## Садржај - Увод - - [Добро дошли у Phi породицу](./md/01.Introduction/01/01.PhiFamily.md) - - [Подешавање вашег окружења](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Добродошли у Phi породицу](./md/01.Introduction/01/01.PhiFamily.md) + - [Подешавање окружења](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Разумевање кључних технологија](./md/01.Introduction/01/01.Understandingtech.md) - - [AI безбедност за Phi моделе](./md/01.Introduction/01/01.AISafety.md) + - [АИ сигурност за Phi моделе](./md/01.Introduction/01/01.AISafety.md) - [Подршка за Phi хардвер](./md/01.Introduction/01/01.Hardwaresupport.md) - [Phi модели и доступност на платформама](./md/01.Introduction/01/01.Edgeandcloud.md) - [Коришћење Guidance-ai и Phi](./md/01.Introduction/01/01.Guidance.md) - [GitHub Marketplace модели](https://github.com/marketplace/models) - [Azure AI каталог модела](https://ai.azure.com) -- Инференција Phi у различитим окружењима +- Интерференција Phi у различитим окружењима - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub модели](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry каталог модела](./md/01.Introduction/02/03.AzureAIFoundry.md) @@ -77,173 +77,176 @@ Phi је тренутно најмоћнији и најисплативији - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Инференција Phi породице - - [Инференција Phi на iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Инференција Phi на Android](./md/01.Introduction/03/Android_Inference.md) - - [Инференција Phi на Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Инференција Phi на AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Инференција Phi са Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Инференција Phi на локалном серверу](./md/01.Introduction/03/Local_Server_Inference.md) - - [Инференција Phi на даљинском серверу помоћу AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Инференција Phi са Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Инференција Phi--Vision локално](./md/01.Introduction/03/Vision_Inference.md) - - [Инференција Phi са Kaito AKS, Azure Containers (званична подршка)](./md/01.Introduction/03/Kaito_Inference.md) -- [Квантификација Phi породице](./md/01.Introduction/04/QuantifyingPhi.md) - - [Квантизација Phi-3.5 / 4 коришћењем llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Квантизација Phi-3.5 / 4 коришћењем Generative AI екстензија за onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Квантизација Phi-3.5 / 4 коришћењем Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Квантизација Phi-3.5 / 4 коришћењем Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) +- Интерференција Phi породице + - [Интерференција Phi на iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Интерференција Phi на Андроиду](./md/01.Introduction/03/Android_Inference.md) + - [Интерференција Phi на Jetson-у](./md/01.Introduction/03/Jetson_Inference.md) + - [Интерференција Phi на AI ПЦ-у](./md/01.Introduction/03/AIPC_Inference.md) + - [Интерференција Phi са Apple MLX Framework-ом](./md/01.Introduction/03/MLX_Inference.md) + - [Интерференција Phi на локалном серверу](./md/01.Introduction/03/Local_Server_Inference.md) + - [Интерференција Phi на удаљеном серверу уз коришћење AI Toolkit-а](./md/01.Introduction/03/Remote_Interence.md) + - [Интерференција Phi са Rust-ом](./md/01.Introduction/03/Rust_Inference.md) + - [Интерференција Phi--Визија локално](./md/01.Introduction/03/Vision_Inference.md) + - [Интерференција Phi са Kaito AKS, Azure контејнерима (званична подршка)](./md/01.Introduction/03/Kaito_Inference.md) +- [Квантовање Phi породице](./md/01.Introduction/04/QuantifyingPhi.md) + - [Квантовање Phi-3.5 / 4 уз помоћ llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Квантовање Phi-3.5 / 4 уз помоћ Generative AI проширења за onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Квантовање Phi-3.5 / 4 уз помоћ Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Квантовање Phi-3.5 / 4 уз помоћ Apple MLX Framework-а](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Евалуација Phi - - [Одговорно AI](./md/01.Introduction/05/ResponsibleAI.md) + - [Одговорни АИ](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry за евалуацију](./md/01.Introduction/05/AIFoundry.md) - - [Коришћење Promptflow за евалуацију](./md/01.Introduction/05/Promptflow.md) + - [Коришћење Promptflow-а за евалуацију](./md/01.Introduction/05/Promptflow.md) -- RAG са Azure AI Search - - [Како користити Phi-4-mini и Phi-4-multimodal (RAG) са Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- RAG са Azure AI претрагом + - [Како користити Phi-4-mini и Phi-4-multimodal (RAG) са Azure AI претрагом](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud локални хибридни RAG са SQLite FTS5 и phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Примери развоја Phi апликација - - Текст и апликације за чет - - Phi-4 примери - - [📓] [Ћаскање са Phi-4-mini ONNX моделом](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Ћаскање са Phi-4 локалним ONNX моделом у .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Конзолна апликација за чет у .NET са Phi-4 ONNX и Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 примери - - [Локални chatbot у прегледачу користећи Phi3, ONNX Runtime Web и WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Višestruki modeli - Interaktivni Phi-3-mini i OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Pravljenje omotača i korišćenje Phi-3 sa MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Optimizacija modela - Kako optimizovati Phi-3-min model za ONNX Runtime Web sa Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 aplikacija sa Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Višestruki modeli AI pokretana aplikacija za beleške primer](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Fino podešavanje i integracija prilagođenih Phi-3 modela sa Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Fino podešavanje i integracija prilagođenih Phi-3 modela sa Prompt flow u Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Evaluacija fino podešenog Phi-3 / Phi-3.5 modela u Microsoft Foundry sa fokusom na Microsoftove principe Odgovornog AI](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct uzorak predviđanja jezika (kineski/engleski)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Četbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Korišćenje Windows GPU za kreiranje Prompt flow rešenja sa Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Korišćenje Microsoft Phi-3.5 tflite za kreiranje Android aplikacije](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET primer koristeći lokalni ONNX Phi-3 model sa Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Konzolna čat .NET aplikacija sa Semantic Kernel i Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI Inference SDK Primeri zasnovani na kodu - - Phi-4 Primeri - - [📓] [Generisanje koda projekta korišćenjem Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 Primeri - - [Pravljenje sopstvenog Visual Studio Code GitHub Copilot čata sa Microsoft Phi-3 porodicom](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Kreiranje sopstvenog Visual Studio Code Chat Copilot agenta sa Phi-3.5 po GitHub modelima](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - Primeri naprednog rezonovanja - - Phi-4 Primeri - - [📓] [Phi-4-mini-razmišljanje ili Phi-4-razmišljanje primeri](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Fino podešavanje Phi-4-mini-razmišljanja sa Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Fino podešavanje Phi-4-mini-razmišljanja sa Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-razmišljanje sa GitHub modelima](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-razmišljanje sa Microsoft Foundry modelima](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demo prikazi - - [Phi-4-mini demo prikazi hostovani na Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demo prikazi hostovani na Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Primeri za vid - - Phi-4 Primeri - - [📓] [Korišćenje Phi-4-multimodal za čitanje slika i generisanje koda](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 Primeri - - [📓][Phi-3-vision-Slika tekst u tekst](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - Текст и чет апликације + - Phi-4 примерци + - [📓] [Чет са Phi-4-mini ONNX моделом](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Чет са Phi-4 локалним ONNX моделом .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Чет .NET конзолна апликација са Phi-4 ONNX уз Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Примери + - [Локални Чатбот у прегледачу користећи Phi3, ONNX Runtime Web и WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [OpenVino Чат](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Више модела - Интерктивни Phi-3-mini и OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Креирање омотача и коришћење Phi-3 са MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Оптимизација модела - Како оптимизовати Phi-3-min модел за ONNX Runtime Web помоћу Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 апликација са Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 Више Модела AI Покретана Апликација За Ноте Пример](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Фино подешавање и интеграција прилагођених Phi-3 модела уз Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Фино подешавање и интеграција прилагођених Phi-3 модела уз Prompt flow у Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Евалуација фино подешеног Phi-3 / Phi-3.5 модела у Microsoft Foundry са фокусом на Microsoft-ове принципе одговорног AI](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct пример предвиђања језика (кинески/енглески)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG Чатбот](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Коришћење Windows GPU за креирање Prompt flow решења са Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Коришћење Microsoft Phi-3.5 tflite за креирање Android апликације](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Q&A .NET пример коришћења локалног ONNX Phi-3 модела користећи Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Конзолна .NET апликација за ћаскање са Semantic Kernel и Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI Inference SDK Примери засновани на коду + - Phi-4 Примери + - [📓] [Генериши код пројекта користећи Phi-4-мултимодал](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Phi-3 / 3.5 Примери + - [Креирај свој Visual Studio Code GitHub Copilot Чат са Microsoft Phi-3 породицом](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Креирај свог Visual Studio Code Chat Copilot агента са Phi-3.5 од GitHub модела](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - Напредни примери закључивања + - Phi-4 Примери + - [📓] [Phi-4-mini-риећонг или Phi-4-риећонг примери](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Фино подешавање Phi-4-mini-риећонг са Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Фино подешавање Phi-4-mini-риећонг са Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-риећонг са GitHub моделима](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-риећонг са Microsoft Foundry моделима](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Демонстрације + - [Phi-4-mini демонстрације хостоване на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-мултимодал демонстрације хостоване на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Примери из вида + - Phi-4 Примери + - [📓] [Користи Phi-4-мултимодал за читање слика и генерисање кода](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 Примери + - [📓][Phi-3-vision-Слика текст у текст](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP ugradnja](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 Reciklaža](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Vizuelni asistent jezika - sa Phi3-Vision i OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Vid Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Vid OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vid višekadrovski ili višeslikovni primer](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vid Lokalni ONNX Model koristeći Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Meni zasnovani Phi-3 Vid Lokalni ONNX Model koristeći Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Primeri za rezonovanje i vid + - [📓][Phi-3-vision CLIP уграђивање](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [ДЕМO: Phi-3 Рециклажа](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - Визуелни језички асистент - са Phi3-Vision и OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision мулти-фрејм или мулти-слика пример](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision Локални ONNX модел користећи Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Мену базиран Phi-3 Vision Локални ONNX модел користећи Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Примери закључивања из вида - Phi-4-Reasoning-Vision-15B - - [📓] [Korišćenje Phi-4-Reasoning-Vision-15B za detekciju prelaženja van pešačkog prelaza](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Korišćenje Phi-4-Reasoning-Vision-15B za matematiku](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Korišćenje Phi-4-Reasoning-Vision-15B za detekciju UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - Primeri matematike - - Phi-4-Mini-Flash-Reasoning-Instruct Primeri [Demo matematike sa Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - - Audio primeri - - Phi-4 primeri - - [📓] [Ekstrakcija audio transkripata koristeći Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal audio primer](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal primer prevoda govora](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET konzolna aplikacija koristeći Phi-4-multimodal audio za analizu audio fajla i generisanje transkripta](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - MOE primeri - - Phi-3 / 3.5 primeri - - [📓] [Phi-3.5 Mešavina stručnjaka (MoEs) primer za društvene mreže](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Pravljenje Retrieval-Augmented Generation (RAG) pipeline sa NVIDIA NIM Phi-3 MOE, Azure AI Search i LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Коришћење Phi-4-Reasoning-Vision-15B за детекцију непрописног преласка улице](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Коришћење Phi-4-Reasoning-Vision-15B за математику](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Коришћење Phi-4-Reasoning-Vision-15B за детекцију корисничког интерфејса](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - Математички примери + - Phi-4-Мини-Фласх-Размишљање-Инструкције примери [Математички демо са Phi-4-Мини-Фласх-Размишљање-Инструкције](./md/02.Application/09.Math/MathDemo.ipynb) + + - Аудио примери + - Phi-4 Примери + - [📓] [Извлачење транскрипата из аудио снимака користећи Phi-4-мултимодал](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-мултимодал аудио пример](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-мултимодал пример превођења говора](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET конзолна апликација користећи Phi-4-мултимодал за анализу аудио фајла и генерисање транскрипта](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE Примери + - Phi-3 / 3.5 Примери + - [📓] [Phi-3.5 Мешавина експерата (MoEs) пример за друштвене мреже](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Креирање Retrieval-Augmented Generation (RAG) процеса са NVIDIA NIM Phi-3 MOE, Azure AI Search и LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Primeri poziva funkcija - - Phi-4 primeri 🆕 - - [📓] [Korišćenje poziva funkcija sa Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Korišćenje poziva funkcija za kreiranje višestrukih agenata sa Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Korišćenje poziva funkcija sa Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Korišćenje poziva funkcija sa ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Primeri mešanja multimodala - - Phi-4 primeri 🆕 - - [📓] [Korišćenje Phi-4-multimodal kao tehnološki novinar](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET konzolna aplikacija koristeći Phi-4-multimodal za analizu slika](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Fino podešavanje Phi primera - - [Scenariji fino podešavanja](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Fino podešavanje naspram RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Fino podešavanje neka Phi-3 postane industrijski ekspert](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Fino podešavanje Phi-3 sa AI Toolkit za VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Fino podešavanje Phi-3 sa Azure Machine Learning servisom](./md/03.FineTuning/Introduce_AzureML.md) - - [Fino podešavanje Phi-3 sa Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Fino podešavanje Phi-3 sa QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Fino podešavanje Phi-3 sa Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Fino podešavanje Phi-3 sa Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Fino podešavanje sa Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Fino podešavanje sa Microsoft Olive Praktična radionica](./md/03.FineTuning/olive-lab/readme.md) - - [Fino podešavanje Phi-3-vision sa Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Fino podešavanje Phi-3 sa Apple MLX okvirom](./md/03.FineTuning/FineTuning_MLX.md) - - [Fino podešavanje Phi-3-vision (službena podrška)](./md/03.FineTuning/FineTuning_Vision.md) - - [Фино подешавање Phi-3 са Kaito AKS , Azure контејнери (званична подршка)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Фино подешавање Phi-3 и 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - -- Практична радионица - - [Истраживање најсавременијих модела: LLM, SLM, локални развој и више](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Откључавање потенцијала NLP-а: фино подешавање са Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - -- Академски научни радови и публикације - - [Учебници су све што вам је потребно II: технички извештај phi-1.5](https://arxiv.org/abs/2309.05463) - - [Технички извештај Phi-3: Високо способан језички модел локално на вашем телефону](https://arxiv.org/abs/2404.14219) - - [Технички извештај Phi-4](https://arxiv.org/abs/2412.08905) - - [Технички извештај Phi-4-Mini: Компактни али моћни мултимодални језички модели преко Миксуре LoRA](https://arxiv.org/abs/2503.01743) + - Примери позивања функција + - Phi-4 Примери 🆕 + - [📓] [Коришћење позивања функција са Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Коришћење позивања функција да се креира више агената са Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Коришћење позивања функција са Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Коришћење позивања функција са ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Примери мешања мултимодалних података + - Phi-4 Примери 🆕 + - [📓] [Коришћење Phi-4-мултимодал као технолошког новинара](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET конзолна апликација користећи Phi-4-мултимодал за анализу слика](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Фино подешавање Phi примера + - [Сценарији фино подешавања](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Фино подешавање vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Фино подесити да Phi-3 постане индустријски експерт](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Фино подешавање Phi-3 са AI Toolkit за VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Фино подешавање Phi-3 са Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Фино подешавање Phi-3 са Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Фино подешавање Phi-3 са QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Фино подешавање Phi-3 са Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Фино подешавање Phi-3 са Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Фино подешавање са Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Рад у лабораторији за фино подешавање са Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) + - [Фино подешавање Phi-3-vision са Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Фајн-тунинг Phi-3 са Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Фајн-тунинг Phi-3-vision (званична подршка)](./md/03.FineTuning/FineTuning_Vision.md) + - [Фајн-тунинг Phi-3 са Kaito AKS , Azure Containers(званична подршка)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Фајн-тунинг Phi-3 и 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + +- Практични лабораторијски рад + - [Истраживање најсавременијих модела: LLMs, SLMs, локални развој и још много тога](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Отварање потенцијала NLP-а: фајн-тунинг са Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + +- Академски истраживачки радови и публикације + - [Textbooks Are All You Need II: phi-1.5 технички извештај](https://arxiv.org/abs/2309.05463) + - [Phi-3 технички извештај: Високо способан језички модел локално на вашем телефону](https://arxiv.org/abs/2404.14219) + - [Phi-4 технички извештај](https://arxiv.org/abs/2412.08905) + - [Phi-4-Mini технички извештај: Компактни али моћни мултимодални језички модели путем mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - [Оптимизација малих језичких модела за позивање функција у возилу](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Фино подешавање PHI-3 за одговарање на питања са више избора: Методологија, резултати и изазови](https://arxiv.org/abs/2501.01588) - - [Технички извештај Phi-4-резоновања](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Технички извештај Phi-4-mini-резоновања](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [(WhyPHI) Фајн-тунинг PHI-3 за одговарање на питања са више избора: Методологија, резултати и изазови](https://arxiv.org/abs/2501.01588) + - [Phi-4-reasoning технички извештај](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-reasoning технички извештај](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Коришћење Phi модела ### Phi на Microsoft Foundry -Можете научити како да користите Microsoft Phi и како да правите E2E решења на разним хардверским уређајима. Да бисте искусили Phi лично, почните са тестирањем модела и прилагођавањем Phi-ја вашим сценаријима коришћењем [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Више информација можете пронаћи у упутству Почетак рада са [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md). +Можете научити како да користите Microsoft Phi и како да изградите E2E решења на вашим различитим хардверским уређајима. Да бисте доживели Phi, почните тако што ћете се играти са моделима и прилагодити Phi својим сценаријима користећи [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Више информација можете сазнати у упутству Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Плаyгроунд** -Сваки модел има свој посебан плаyгроунд за тестирање модела [Azure AI Playground](https://aka.ms/try-phi3). +**Простор за испробавање** +Сваки модел има посебан простор за тестирање [Azure AI Playground](https://aka.ms/try-phi3). ### Phi на GitHub моделима -Можете научити како да користите Microsoft Phi и како да правите E2E решења на разним хардверским уређајима. Да бисте искусили Phi лично, почните са тестирањем модела и прилагођавањем Phi-ја вашим сценаријима коришћењем [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Више информација можете пронаћи у упутству Почетак рада са [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md). +Можете научити како да користите Microsoft Phi и како да изградите E2E решења на вашим различитим хардверским уређајима. Да бисте доживели Phi, почните тако што ћете се играти са моделом и прилагодити Phi својим сценаријима користећи [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Више информација можете сазнати у упутству Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Плаyгроунд** -Сваки модел има своју посебну [плаyгроунд за тестирање модела](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Простор за испробавање** +Сваки модел има посебан [простор за тестирање модела](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi на Hugging Face -Модел можете пронаћи и на [Hugging Face](https://huggingface.co/microsoft). +Модел можете наћи и на [Hugging Face](https://huggingface.co/microsoft) -**Плаyгроунд** - [Hugging Chat плаyгроунд](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Простор за испробавање** + [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) ## 🎒 Остали курсеви @@ -251,74 +254,76 @@ Phi је тренутно најмоћнији и најисплативији ### LangChain -[![LangChain4j за почетнике](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js за почетнике](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain за почетнике](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain for Beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- ### Azure / Edge / MCP / Агенти -[![AZD за почетнике](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI за почетнике](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP за почетнике](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI агенти за почетнике](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Генеративна серија вештачке интелигенције -[![Генеративна AI за почетнике](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Генеративна AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Генеративна AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Генеративна AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### Серсија генеративне вештачке интелигенције +[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- ### Основно учење -[![ML за почетнике](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Наука о подацима за почетнике](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI за почетнике](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Кибербезбедност за почетнике](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Веб развој за почетнике](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT за почетнике](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR развој за почетнике](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Copilot серија -[![Copilot за парно програмирање са AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot за C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot авантуре](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +### Серсија Copilot +[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## Одговорна вештачка интелигенција -Microsoft је посвећен помоћи нашим корисницима да користе наше AI производе одговорно, делећи своја сазнања и градећи партнерства заснована на поверењу кроз алатке као што су напомене о транспарентности и процене утицаја. Многи од ових ресурса могу се пронаћи на [https://aka.ms/RAI](https://aka.ms/RAI). -Приступ Microsoft-a одговорној вештачкој интелигенцији заснован је на нашим AI принципима праведности, поузданости и безбедности, приватности и безбедности, укључивости, транспарентности и одговорности. +Microsoft је посвећен помагању нашим корисницима да користе наше AI производе одговорно, делећи наша сазнања и градећи партнерства заснована на поверењу кроз алате као што су Transparency Notes и Impact Assessments. Многи од ових ресурса могу се пронаћи на [https://aka.ms/RAI](https://aka.ms/RAI). +Приступ компаније Microsoft одговорној вештачкој интелигенцији заснива се на нашим AI принципима правичности, поузданости и безбедности, приватности и сигурности, инклузивности, транспарентности и одговорности. -Великоразмерни модели природног језика, слика и говора — као они коришћени у овом примеру — потенцијално могу деловати на неправедан, непоуздан или увредљив начин, што може нанети штету. Молимо вас да консултујете [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) да бисте били информисани о ризицима и ограничењима. -Препоручени приступ за смањење ових ризика је да укључите систем безбедности у вашу архитектуру који може да детектује и спречи штетно понашање. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) пружа независни слој заштите, способан да детектује штетан кориснички и AI генерисани садржај у апликацијама и услугама. Azure AI Content Safety укључује API-је за текст и слике који вам омогућавају да детектујете материјал који је штетан. У оквиру Microsoft Foundry, сервис Content Safety вам омогућава да видите, истражите и испробате пример кода за детекцију штетног садржаја кроз различите модалитете. Следећа [документација за брзи почетак](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) вас води кроз прављење захтева ка сервису. +Велики модели за природни језик, слику и говор - као што су они коришћени у овом примеру - могу потенцијално да се понашају на начине који су неправедни, непоуздани или увредљиви, што може проузроковати штету. Молимо вас да консултујете [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) да бисте били информисани о ризицима и ограничењима. -Још један аспект који треба узети у обзир је укупне перформансе апликације. Код мултимодалних и мултимоделских апликација, перформансе подразумевамо као то да систем функционише онако како ви и ваши корисници очекујете, укључујући да не генерише штетне излазне резултате. Важно је да процените перформансе ваше укупне апликације користећи [оценитеље перформанси и квалитета и ризика и безбедности](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Такође имате могућност да креирате и процените са [прилагођеним оценитељима](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Можете анализирати вашу AI апликацију у вашем развојном окружењу користећи [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Узимајући или тест скуп података или циљ, ваше генерације у генеративној AI апликацији се кванитативно мере са уграђеним оценитељима или вашим изабраним прилагођеним оценитељима. Да бисте започели са azure ai evaluation sdk за процену вашег система, можете пратити [упутство за брзи почетак](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Када извршите процену, можете [визуелизовати резултате у Microsoft Foundry-ју](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Препоручени приступ за ублажавање ових ризика је укључивање безбедносног система у вашу архитектуру који може да детектује и спречи штетно понашање. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) пружа независни слој заштите, способан да детектује штетни кориснички и AI-генерисани садржај у апликацијама и услугама. Azure AI Content Safety укључује текстуалне и сликовне API-је који вам омогућавају да детектујете штетни материјал. У оквиру Microsoft Foundry, сервис Content Safety вам омогућава да прегледате, истражујете и испробате пример кода за детекцију штетног садржаја кроз различите модалитете. Следећа [документација за брзи почетак](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) вас води кроз прављење захтева сервису. -## Трговачке марке +Још један аспект који треба узети у обзир је укупне перформансе апликације. Код мултимодалних и мултимодел апликација, под перформансама мислимо да систем функционише онако како ви и ваши корисници очекујете, укључујући негенерисање штетних резултата. Важно је проценити перформансе ваше укупне апликације користећи [еvaluatorе за перформансе, квалитет, ризик и безбедност](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Такође имате могућност да креирате и евалвирате са [прилагођеним evaluatorима](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Овај пројекат може да садржи трговачке марке или логотипе пројеката, производа или услуга. Ауторизована употреба Microsoft трговачких марки или логотипа подлеже и мора пратити [Microsoft-ове смернице за трговачке марке и брендинг](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Употреба Microsoft трговачких марки или логотипа у модификованим верзијама овог пројекта не сме да изазове конфузију или да имплицира Microsoft спонзорство. Било каква употреба трговачких марки или логотипа трећих страна подлеже политикама тих трећих страна. +Можете евалвирати вашу AI апликацију у вашем развојном окружењу користећи [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Уз дати тест скуп података или циљ, генерације ваше генеративне AI апликације се квантитативно мере уграђеним или прилагођеним evaluatorима по вашем избору. Да бисте почели са Azure AI Evaluation SDK за евалуацију вашег система, можете пратити [водич за брзи почетак](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Када извршите покретање евалуације, можете [визуализовати резултате у Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## Добијање помоћи +## Заштитни знаци -Ако запнете или имате било каквих питања у вези са прављењем AI апликација, придружите се: +Овај пројекат може садржати заштитне знакове или логотипе пројеката, производа или услуга. Ауторизована употреба Мицрософт заштитних знакова или логотипа је предмет и мора пратити [Мицрософтове смернице за употребу заштитних знакова и брендова](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Употреба Мицрософт заштитних знакова или логотипа у изменљивим верзијама овог пројекта не сме изазивати забуну или имплицирати Мицрософтово спонзорство. Свака употреба заштитних знакова или логотипа трећих страна подлеже правилима тих трећих страна. + +## Како добити помоћ + +Ако заглавите или имате било каквих питања у вези изградње AI апликација, придружите се: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Ако имате повратне информације о производу или грешке током прављења посетите: +Ако имате повратне информације о производу или грешке приликом израде посетите: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Изјава о одрицању одговорности**: -Овај документ је преведен помоћу AI услуге за превођење [Co-op Translator](https://github.com/Azure/co-op-translator). Иако тежимо прецизности, молимо имайте у виду да аутоматизовани преводи могу садржати грешке или нетачности. Оригинални документ на његовом изворном језику треба сматрати ауторитативним извором. За критичне информације препоручује се професионални превод од стране људи. Нисмо одговорни за било каква неразумевања или погрешна тумачења која проистекну из употребе овог превода. +**Изјава о одрицању одговорности**: +Овај документ је преведен коришћењем услуге за аутоматски превод [Co-op Translator](https://github.com/Azure/co-op-translator). Иако тежимо тачности, имајте у виду да аутоматски преводи могу садржати грешке или нетачности. Оригинални документ на његовом изворном језику треба сматрати ауторитативним извором. За критичне информације препоручује се професионални људски превод. Нисмо одговорни за било каква неспоразума или погрешна тумачења која произилазе из коришћења овог превода. \ No newline at end of file diff --git a/translations/sr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/sr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..5dda51bce --- /dev/null +++ b/translations/sr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Хибридни RAG са Microsoft phi-4-mini и SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Аутор:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Фокус:** Мали језички модели (SLMs), SQLite FTS5 BM25, густе уграђености, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Мотивација: Дилема враћања кључних речи у локалним SLM-овима\n", + "Стандардне RAG архитектуре које се ослањају искључиво на густе векторске уграђености често не успевају да пронађу тачне нумеричке токене (нпр. `2,340,000 TL`, кодове уговора, бројеве рачуна). \n", + "Насупрот томе, ређи лексички претраживач (BM25) пропушта семантичке синониме и парафразирана питања.\n", + "\n", + "Овај приручник показује како имплементирати **високобрзински, у меморији хибридни претраживач** који комбинује:\n", + "1. **Густе векторе** (косинусна сличност)\n", + "2. **Ређи лексички претраживач** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Генерацију поткрепљених цитата (`[1]`, `[2]`)** са Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Двојни SQLite Шема (Густе Векторе + Виртуелна FTS5 BM25 Табела)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Мерење перформанси уноса и извршавања примерка\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Формулација утемељеног упита за Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Изјава о одрицању одговорности**:\nОвај документ је преведен коришћењем услуге за аутоматски превод [Co-op Translator](https://github.com/Azure/co-op-translator). Иако тежимо тачности, имајте у виду да аутоматски преводи могу садржати грешке или нетачности. Оригинални документ на његовом изворном језику треба сматрати ауторитативним извором. За критичне информације препоручује се професионални људски превод. Нисмо одговорни за било каква неспоразума или погрешна тумачења која произилазе из коришћења овог превода.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/sv/.co-op-translator.json b/translations/sv/.co-op-translator.json index ee545f471..a236f4a65 100644 --- a/translations/sv/.co-op-translator.json +++ b/translations/sv/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "sv" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:00:39+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:18:15+00:00", "source_file": "README.md", "language_code": "sv" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "sv" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:19:17+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "sv" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:07+00:00", diff --git a/translations/sv/README.md b/translations/sv/README.md index 95e95b79c..3b9b30e0a 100644 --- a/translations/sv/README.md +++ b/translations/sv/README.md @@ -1,42 +1,42 @@ # Phi Cookbook: Praktiska exempel med Microsofts Phi-modeller [![Öppna och använd exemplen i GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Öppna i Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Öppna&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Öppna i Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub bidragsgivare](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub ärenden](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub pull-förfrågningar](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub pull-begäranden](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Välkomna](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub observatörer](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Följ)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Forka)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub stjärnor](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Stjärna)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub bevakare](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub stjärnor](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi är en serie öppen källkods-AI-modeller utvecklade av Microsoft. +Phi är en serie open source AI-modeller utvecklade av Microsoft. -Phi är för närvarande den mest kraftfulla och kostnadseffektiva lilla språkmodellen (SLM), med mycket bra referensvärden inom flerspråkighet, resonemang, text-/chatskapande, kodning, bilder, ljud och andra scenarier. +Phi är för närvarande den mest kraftfulla och kostnadseffektiva lilla språkmodellen (SLM), med mycket bra resultat inom flerspråkighet, resonemang, text-/chattgenerering, kodning, bilder, ljud och andra scenarier. -Du kan distribuera Phi i molnet eller på kant-enheter, och du kan enkelt bygga generativa AI-applikationer med begränsad beräkningskraft. +Du kan distribuera Phi i molnet eller till edge-enheter, och du kan enkelt bygga generativa AI-applikationer med begränsad datorkraft. Följ dessa steg för att komma igång med att använda dessa resurser: -1. **Fork:a lagret**: Klicka [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Klon:a lagret**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Gå med i Microsoft AI Discord-communityn och träffa experter och andra utvecklare**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +1. **Forka Repositoriet**: Klicka [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Klona Repositoriet**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Gå med i Microsoft AI Discord Community och träffa experter och andra utvecklare**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/sv/cover.eb18d1b9605d754b.webp) -### 🌐 Stöd för flera språk +### 🌐 Flerspråkigt stöd #### Stöds via GitHub Action (Automatiserat & Alltid uppdaterat) -[Arabiska](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgariska](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Kinesiska (förenklad)](../zh-CN/README.md) | [Kinesiska (traditionell, Hongkong)](../zh-HK/README.md) | [Kinesiska (traditionell, Macau)](../zh-MO/README.md) | [Kinesiska (traditionell, Taiwan)](../zh-TW/README.md) | [Kroatiska](../hr/README.md) | [Tjeckiska](../cs/README.md) | [Danska](../da/README.md) | [Nederländska](../nl/README.md) | [Estniska](../et/README.md) | [Finska](../fi/README.md) | [Franska](../fr/README.md) | [Tyska](../de/README.md) | [Grekiska](../el/README.md) | [Hebreiska](../he/README.md) | [Hindi](../hi/README.md) | [Ungerska](../hu/README.md) | [Indonesiska](../id/README.md) | [Italienska](../it/README.md) | [Japanska](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreanska](../ko/README.md) | [Litauiska](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigeriansk pidgin](../pcm/README.md) | [Norska](../no/README.md) | [Persiska (Farsi)](../fa/README.md) | [Polska](../pl/README.md) | [Portugisiska (Brasilien)](../pt-BR/README.md) | [Portugisiska (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumänska](../ro/README.md) | [Ryska](../ru/README.md) | [Serbiska (kyrilliska)](../sr/README.md) | [Slovakiska](../sk/README.md) | [Slovenska](../sl/README.md) | [Spanska](../es/README.md) | [Swahili](../sw/README.md) | [Svenska](./README.md) | [Tagalog (filippinska)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thailändska](../th/README.md) | [Turkiska](../tr/README.md) | [Ukrainska](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamesiska](../vi/README.md) +[Arabiska](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgariska](../bg/README.md) | [Burmesiska (Myanmar)](../my/README.md) | [Kinesiska (Förenklad)](../zh-CN/README.md) | [Kinesiska (Traditionell, Hong Kong)](../zh-HK/README.md) | [Kinesiska (Traditionell, Macau)](../zh-MO/README.md) | [Kinesiska (Traditionell, Taiwan)](../zh-TW/README.md) | [Kroatiska](../hr/README.md) | [Tjeckiska](../cs/README.md) | [Danska](../da/README.md) | [Holländska](../nl/README.md) | [Estniska](../et/README.md) | [Finska](../fi/README.md) | [Franska](../fr/README.md) | [Tyska](../de/README.md) | [Grekiska](../el/README.md) | [Hebreiska](../he/README.md) | [Hindi](../hi/README.md) | [Ungerska](../hu/README.md) | [Indonesiska](../id/README.md) | [Italienska](../it/README.md) | [Japanska](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Koreanska](../ko/README.md) | [Lithauiska](../lt/README.md) | [Malajiska](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigeriansk Pidgin](../pcm/README.md) | [Norska](../no/README.md) | [Persiska (Farsi)](../fa/README.md) | [Polska](../pl/README.md) | [Portugisiska (Brasilien)](../pt-BR/README.md) | [Portugisiska (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Rumänska](../ro/README.md) | [Ryska](../ru/README.md) | [Serbiska (Kyrilliska)](../sr/README.md) | [Slovakiska](../sk/README.md) | [Slovenska](../sl/README.md) | [Spanska](../es/README.md) | [Swahili](../sw/README.md) | [Svenska](./README.md) | [Tagalog (Filippinska)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thailändska](../th/README.md) | [Turkiska](../tr/README.md) | [Ukrainska](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamesiska](../vi/README.md) -> **Föredrar du att klona lokalt?** +> **Föredrar att klona lokalt?** > -> Detta lager inkluderar 50+ språköversättningar vilket avsevärt ökar nedladdningsstorleken. För att klona utan översättningar, använd sparse checkout: +> Detta repository inkluderar över 50 språköversättningar vilket avsevärt ökar nedladdningsstorleken. För att klona utan översättningar, använd sparsamt utcheckning: > > **Bash / macOS / Linux:** > ```bash @@ -59,193 +59,196 @@ Följ dessa steg för att komma igång med att använda dessa resurser: - Introduktion - [Välkommen till Phi-familjen](./md/01.Introduction/01/01.PhiFamily.md) - - [Så sätter du upp din miljö](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Förståelse av nyckelteknologier](./md/01.Introduction/01/01.Understandingtech.md) + - [Ställa in din miljö](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Förstå nyckelteknologier](./md/01.Introduction/01/01.Understandingtech.md) - [AI-säkerhet för Phi-modeller](./md/01.Introduction/01/01.AISafety.md) - [Phi hårdvarustöd](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi-modeller & tillgänglighet över plattformar](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Phi-modeller & Tillgänglighet över plattformar](./md/01.Introduction/01/01.Edgeandcloud.md) - [Använda Guidance-ai och Phi](./md/01.Introduction/01/01.Guidance.md) - [GitHub Marketplace-modeller](https://github.com/marketplace/models) - - [Azure AI Model Catalog](https://ai.azure.com) + - [Azure AI Modellkatalog](https://ai.azure.com) -- Inferenz av Phi i olika miljöer +- Inferens Phi i olika miljöer - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub-modeller](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry Model Catalog](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Microsoft Foundry Modellkatalog](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - -- Inferenz Phi-familjen - - [Inference Phi på iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Inference Phi på Android](./md/01.Introduction/03/Android_Inference.md) - - [Inference Phi på Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Inference Phi på AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Inference Phi med Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Inference Phi på lokal server](./md/01.Introduction/03/Local_Server_Inference.md) - - [Inference Phi på fjärrserver med AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Inference Phi med Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Inference Phi–Vision lokalt](./md/01.Introduction/03/Vision_Inference.md) - - [Inference Phi med Kaito AKS, Azure Containers (officiellt stöd)](./md/01.Introduction/03/Kaito_Inference.md) -- [Kvantifiera Phi-familjen](./md/01.Introduction/04/QuantifyingPhi.md) + - [Foundry Lokalt](./md/01.Introduction/02/07.FoundryLocal.md) + +- Inferens Phi-familjen + - [Inferens Phi i iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Inferens Phi i Android](./md/01.Introduction/03/Android_Inference.md) + - [Inferens Phi i Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Inferens Phi i AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Inferens Phi med Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Inferens Phi i lokal server](./md/01.Introduction/03/Local_Server_Inference.md) + - [Inferens Phi i fjärrserver med AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Inferens Phi med Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Inferens Phi–Vision lokalt](./md/01.Introduction/03/Vision_Inference.md) + - [Inferens Phi med Kaito AKS, Azure Containers (officiellt stöd)](./md/01.Introduction/03/Kaito_Inference.md) +- [Kvantisera Phi-familjen](./md/01.Introduction/04/QuantifyingPhi.md) - [Kvantisera Phi-3.5 / 4 med llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - [Kvantisera Phi-3.5 / 4 med Generative AI extensions för onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [Kvantisera Phi-3.5 / 4 med Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - [Kvantisera Phi-3.5 / 4 med Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- Utvärdera Phi +- Utvärdering Phi - [Ansvarsfull AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry för utvärdering](./md/01.Introduction/05/AIFoundry.md) - [Använda Promptflow för utvärdering](./md/01.Introduction/05/Promptflow.md) - RAG med Azure AI Search - [Hur man använder Phi-4-mini och Phi-4-multimodal (RAG) med Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud Local Hybrid RAG med SQLite FTS5 och phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi applikationsutvecklingsexempel +- Phi applikationsutvecklingsprov - Text- & chattapplikationer - Phi-4-exempel - - [📓] [Chatta med Phi-4-mini ONNX-modell](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Chatta med Phi-4 lokal ONNX-modell .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Chat .NET konsolapp med Phi-4 ONNX via Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5-exempel - - [Lokal chatbot i webbläsare med Phi3, ONNX Runtime Web och WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [📓] [Chatta med Phi-4-mini ONNX Modell](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Chatta med Phi-4 lokal ONNX Modell .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Chat .NET konsolapp med Phi-4 ONNX med Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Exempel + - [Lokal Chatbot i webbläsaren med Phi3, ONNX Runtime Web och WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - Interaktiv Phi-3-mini och OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Skapa ett wrapper och använd Phi-3 med MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Modeloptimering - Hur man optimerar Phi-3-minimodellen för ONNX Runtime Web med Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Multi Modell - Interaktiv Phi-3-mini och OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Bygga en wrapper och använda Phi-3 med MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Modelloptimering - Hur man optimerar Phi-3-minimodellen för ONNX Runtime Web med Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [WinUI3-app med Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Multi Model AI-drivet anteckningsexempel](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Fintsuning och integrera anpassade Phi-3-modeller med Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Fintsuning och integrera anpassade Phi-3-modeller med Prompt flow i Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Utvärdera den fintsunade Phi-3 / Phi-3.5-modellen i Microsoft Foundry med fokus på Microsofts ansvarsfulla AI-principer](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct språklig förutsägelseexempel (kinesiska/engelska)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + -[WinUI3 Multi Model AI-driven Anteckningsapp Exempel](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Finjustera och integrera egna Phi-3 modeller med Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Finjustera och integrera egna Phi-3 modeller med Prompt flow i Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Utvärdera den finjusterade Phi-3 / Phi-3.5 modellen i Microsoft Foundry med fokus på Microsofts ansvariga AI-principer](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct språkprovsmodell (kinesiska/engelska)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Använd Windows GPU för att skapa Prompt flow-lösning med Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Använd Microsoft Phi-3.5 tflite för att skapa Android-app](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET-exempel med lokal ONNX Phi-3-modell med Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Konsolchat .NET-app med Semantic Kernel och Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Använda Windows GPU för att skapa Prompt flow-lösning med Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Använda Microsoft Phi-3.5 tflite för att skapa Android-app](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Frågor & Svar .NET-exempel med lokal ONNX Phi-3-modell med Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Konsolchatt .NET-app med Semantic Kernel och Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK kodbaserade exempel - - Phi-4-exempel + - Azure AI Inference SDK Kodbaserade Exempel + - Phi-4 Exempel - [📓] [Generera projektkod med Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5-exempel + - Phi-3 / 3.5 Exempel - [Bygg din egen Visual Studio Code GitHub Copilot Chat med Microsoft Phi-3-familjen](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Skapa din egen Visual Studio Code Chat Copilot Agent med Phi-3.5 av GitHub-modeller](./md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Skapa din egen Visual Studio Code Chat Copilot Agent med Phi-3.5 via GitHub-modeller](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - Avancerade resonemangsexempel - - Phi-4-exempel - - [📓] [Phi-4-mini-resonemang eller Phi-4-resonemangsexempel](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Fintsuning av Phi-4-mini-resonemang med Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Fintsuning av Phi-4-mini-resonemang med Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - Avancerade Resonemangsexempel + - Phi-4 Exempel + - [📓] [Phi-4-mini-resonemang eller Phi-4-resonemangs exempel](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Finjustera Phi-4-mini-resonemang med Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Finjustera Phi-4-mini-resonemang med Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-resonemang med GitHub-modeller](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-resonemang med Microsoft Foundry-modeller](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demos - [Phi-4-mini demos värd på Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodala demos värd på Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodala demos värd på Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - Visionsexempel - - Phi-4-exempel - - [📓] [Använd Phi-4-multimodal för att läsa bilder och generera kod](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5-exempel + - Phi-4 Exempel + - [📓] [Använd Phi-4-multimodal för att läsa bilder och generera kod](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 Exempel - [📓][Phi-3-vision-Bild text till text](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision CLIP Inbäddning](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [DEMO: Phi-3 Återvinning](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Visuell språkassistent - med Phi3-Vision och OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3-vision - Visuell språkasistent - med Phi3-Vision och OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision multi-frame eller multi-image exempel](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision Lokal ONNX-modell med Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menybaserad Phi-3 Vision Lokal ONNX-modell med Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Phi-3.5 Vision multi-frame eller multi-bild exempel](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision Lokal ONNX Modell med Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Menybaserad Phi-3 Vision Lokal ONNX Modell med Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - Resonemang-Visionsexempel - - Phi-4-Reasoning-Vision-15B - - [📓] [Använda Phi-4-Reasoning-Vision-15B för att upptäcka olaglig gång](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Använda Phi-4-Reasoning-Vision-15B för matematik](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Använda Phi-4-Reasoning-Vision-15B för att upptäcka UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - Resonemangs-Vision Exempel + - Phi-4-Resonemangs-Vision-15B + - [📓] [Använda Phi-4-Resonemangs-Vision-15B för att upptäcka olaglig vägövergång (jaywalking)](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Använda Phi-4-Resonemangs-Vision-15B för matematik](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Använda Phi-4-Resonemangs-Vision-15B för att upptäcka UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - Mattexempel - - Phi-4-Mini-Flash-Reasoning-Instruct-exempel [Mattematisk demo med Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Matematikexempel + - Phi-4-Mini-Flash-Reasoning-Instruct Exempel [Matematikdemo med Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - Ljudexempel - - Phi-4-exempel - - [📓] [Extrahera ljudtranskript med Phi-4-multimodala](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodalt ljudexempel](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal talöversättningsexempel](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET-konsolapplikation som använder Phi-4-multimodal ljud för att analysera en ljudfil och generera transkript](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - MOE-exempel - - Phi-3 / 3.5-exempel - - [📓] [Phi-3.5 Mixture of Experts Models (MoEs) Sociala Medier-exempel](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Bygga en Retrieval-Augmented Generation (RAG) pipeline med NVIDIA NIM Phi-3 MOE, Azure AI Search och LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - Phi-4 Exempel + - [📓] [Extrahera ljudtranskript med Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal Ljudexempel](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal Talöversättningsexempel](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET konsolapplikation med Phi-4-multimodal Ljud för att analysera ljudfil och generera transkript](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE Exempel + - Phi-3 / 3.5 Exempel + - [📓] [Phi-3.5 Mixture of Experts Models (MoEs) Sociala Medier Exempel](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Bygga en Retrieval-Augmented Generation (RAG) Pipeline med NVIDIA NIM Phi-3 MOE, Azure AI Search och LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Funktionsanrops-exempel - - Phi-4-exempel 🆕 - - [📓] [Använda funktionsanrop med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Använda funktionsanrop för att skapa multi-agenter med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Använda funktionsanrop med Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Använda funktionsanrop med ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Multimodal Mixing-exempel - - Phi-4-exempel 🆕 - - [📓] [Använda Phi-4-multimodal som teknikjournalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET-konsolapplikation som använder Phi-4-multimodal för att analysera bilder](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Fintsuning Phi-exempel - - [Fintsuning-scenarier](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Fintsuning vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Fintsuning för att låta Phi-3 bli en branschexpert](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Fintsuning Phi-3 med AI Toolkit för VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Fintsuning Phi-3 med Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Fintsuning Phi-3 med Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Fintsuning Phi-3 med QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Fintsuning Phi-3 med Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Fintsuning Phi-3 med Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Fintsuning med Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Fintsuning med Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - - [Fintsuning Phi-3-vision med Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Fintsuning Phi-3 med Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Fintsuning Phi-3-vision (officiellt stöd)](./md/03.FineTuning/FineTuning_Vision.md) + - Funktionanrops Exempel + - Phi-4 Exempel 🆕 + - [📓] [Använda Funktionanrop med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Använda Funktionanrop för att skapa multi-agenter med Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Använda Funktionanrop med Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Använda Funktionanrop med ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Multimodal Blandningsexempel + - Phi-4 Exempel 🆕 + - [📓] [Använda Phi-4-multimodal som teknologijournalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET konsolapplikation med Phi-4-multimodal för att analysera bilder](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Finjustering av Phi Exempel + - [Finjusteringsscenarier](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Finjustering vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Finjustering Låt Phi-3 bli en branschexpert](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Finjustering Phi-3 med AI Toolkit för VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Finjustering Phi-3 med Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Finjustering Phi-3 med Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Finjustering Phi-3 med QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Finjustering Phi-3 med Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Finjustering Phi-3 med Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Finjustering med Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Finjustering med Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) + - [Finjustering Phi-3-vision med Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Finjustera Phi-3 med Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Finjustera Phi-3-vision (officiellt stöd)](./md/03.FineTuning/FineTuning_Vision.md) - [Finjustera Phi-3 med Kaito AKS, Azure Containers (officiellt stöd)](./md/03.FineTuning/FineTuning_Kaito.md) - [Finjustera Phi-3 och 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Praktisk labb - - [Utforska toppmoderna modeller: LLMs, SLMs, lokal utveckling med mera](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Lås upp NLP:s potential: Finjustering med Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Utforska banbrytande modeller: LLMs, SLMs, lokal utveckling och mer](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Lås upp NLP-potential: Finjustering med Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Akademiska forskningsartiklar och publikationer +- Akademiska forskningspapper och publikationer - [Textbooks Are All You Need II: phi-1.5 teknisk rapport](https://arxiv.org/abs/2309.05463) - - [Phi-3 teknisk rapport: En högpresterande språkmodell lokalt på din telefon](https://arxiv.org/abs/2404.14219) + - [Phi-3 teknisk rapport: En högkompetent språkmodell lokalt på din telefon](https://arxiv.org/abs/2404.14219) - [Phi-4 teknisk rapport](https://arxiv.org/abs/2412.08905) - [Phi-4-Mini teknisk rapport: Kompakta men kraftfulla multimodala språkmodeller via Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Optimering av små språkmodeller för funktionsanrop i fordon](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Finjustering av PHI-3 för flervalsfrågor: Metodik, resultat och utmaningar](https://arxiv.org/abs/2501.01588) - - [Phi-4-reasoning teknisk rapport](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini-reasoning teknisk rapport](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Optimera små språkmodeller för funktionsanrop i fordon](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Finjustera PHI-3 för flervalsfrågor: Metodik, resultat och utmaningar](https://arxiv.org/abs/2501.01588) + - [Phi-4-resonerande teknisk rapport](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-resonerande teknisk rapport](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Användning av Phi-modeller +## Använda Phi-modeller ### Phi på Microsoft Foundry -Du kan lära dig hur du använder Microsoft Phi och hur du bygger E2E-lösningar på dina olika hårdvaruenheter. För att testa Phi själv, börja med att leka med modellerna och anpassa Phi för dina scenarier med hjälp av [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Du kan lära dig mer i Kom igång med [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Du kan lära dig hur du använder Microsoft Phi och hur du bygger end-to-end-lösningar för dina olika hårdvaruenheter. För att uppleva Phi själv, börja med att leka med modellerna och anpassa Phi för dina scenarier med hjälp av [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Du kan lära dig mer i Kom igång med [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Testmiljö** -Varje modell har en dedikerad testmiljö för att prova modellen [Azure AI Playground](https://aka.ms/try-phi3). +**Lekplats** +Varje modell har en dedikerad lekplats för att testa modellen [Azure AI Playground](https://aka.ms/try-phi3). ### Phi på GitHub-modeller -Du kan lära dig hur du använder Microsoft Phi och hur du bygger E2E-lösningar på dina olika hårdvaruenheter. För att testa Phi själv, börja med att leka med modellen och anpassa Phi för dina scenarier med hjälp av [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Du kan lära dig mer i Kom igång med [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Du kan lära dig hur du använder Microsoft Phi och hur du bygger end-to-end-lösningar för dina olika hårdvaruenheter. För att uppleva Phi själv, börja med att leka med modellen och anpassa Phi för dina scenarier med hjälp av [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Du kan lära dig mer i Kom igång med [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Testmiljö** -Varje modell har en dedikerad [testmiljö för att prova modellen](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Lekplats** +Varje modell har en dedikerad [lekplats för att testa modellen](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi på Hugging Face Du kan också hitta modellen på [Hugging Face](https://huggingface.co/microsoft) -**Testmiljö** -[Hugging Chat testmiljö](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Lekplats** + [Hugging Chat lekplats](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Andra kurser + ## 🎒 Andra kurser Vårt team producerar andra kurser! Kolla in: @@ -256,15 +259,15 @@ Vårt team producerar andra kurser! Kolla in: [![LangChain för nybörjare](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agentprogram +### Azure / Edge / MCP / Agenter [![AZD för nybörjare](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI för nybörjare](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP för nybörjare](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI-agenter för nybörjare](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agenter för nybörjare](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Generativ AI-serie +### Serie om generativ AI [![Generativ AI för nybörjare](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Generativ AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) [![Generativ AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) @@ -272,7 +275,7 @@ Vårt team producerar andra kurser! Kolla in: --- -### Kärnkunskap +### Grundläggande lärande [![ML för nybörjare](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Datavetenskap för nybörjare](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI för nybörjare](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) @@ -283,36 +286,38 @@ Vårt team producerar andra kurser! Kolla in: --- -### Copilot-serie +### Copilot-serien [![Copilot för AI-parprogrammering](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot för C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![Copilot-äventyr](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Ansvarsfull AI +## Ansvarsfull AI + +Microsoft är engagerat i att hjälpa våra kunder använda våra AI-produkter ansvarsfullt, dela med sig av våra lärdomar och bygga förtroendebaserade partnerskap genom verktyg som Transparency Notes och Impact Assessments. Många av dessa resurser finns på [https://aka.ms/RAI](https://aka.ms/RAI). +Microsofts tillvägagångssätt för ansvarsfull AI är grundat i våra AI-principer om rättvisa, pålitlighet och säkerhet, integritet och säkerhet, inkludering, transparens och ansvar. + +Storskaliga modeller för naturligt språk, bild och tal – som de som används i detta exempel – kan potentiellt bete sig på sätt som är orättvisa, opålitliga eller stötande, vilket kan orsaka skada. Vänligen konsultera [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) för att bli informerad om risker och begränsningar. -Microsoft är engagerade i att hjälpa våra kunder använda våra AI-produkter ansvarsfullt, dela med oss av våra lärdomar och bygga förtroendebaserade partnerskap genom verktyg som Transparency Notes och Impact Assessments. Många av dessa resurser finns på [https://aka.ms/RAI](https://aka.ms/RAI). -Microsofts syn på ansvarsfull AI grundar sig på våra AI-principer för rättvisa, tillförlitlighet och säkerhet, integritet och säkerhet, inkludering, transparens och ansvarsskyldighet. -Storskaliga språk-, bild- och talmodeller – som de som används i detta exempel – kan potentiellt bete sig på sätt som är orättvisa, opålitliga eller stötande, vilket kan orsaka skada. Vänligen konsultera [Azure OpenAI-tjänstens Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) för information om risker och begränsningar. -Den rekommenderade metoden för att mildra dessa risker är att inkludera ett säkerhetssystem i din arkitektur som kan upptäcka och förhindra skadligt beteende. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) erbjuder ett oberoende skyddslager som kan upptäcka skadligt användargenererat och AI-genererat innehåll i applikationer och tjänster. Azure AI Content Safety inkluderar text- och bild-API:er som gör det möjligt för dig att upptäcka material som är skadligt. Inom Microsoft Foundry tillåter Content Safety-tjänsten dig att visa, utforska och prova exempel på kod för att upptäcka skadligt innehåll över olika modaliteter. Följande [snabbstartsdokumentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) guidar dig genom att göra förfrågningar till tjänsten. +Den rekommenderade metoden för att mildra dessa risker är att inkludera ett säkerhetssystem i din arkitektur som kan upptäcka och förhindra skadligt beteende. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) erbjuder ett fristående skyddsskikt som kan upptäcka skadligt användar- och AI-genererat innehåll i applikationer och tjänster. Azure AI Content Safety inkluderar text- och bild-API:er som gör det möjligt att upptäcka material som är skadligt. Inom Microsoft Foundry låter Content Safety-tjänsten dig visa, utforska och prova exempel på kod för att upptäcka skadligt innehåll över olika modaliteter. Följande [quickstart-dokumentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) guidar dig genom att göra förfrågningar till tjänsten. -En annan aspekt att ta hänsyn till är den övergripande applikationsprestandan. Med multimodala och multimodellsapplikationer anser vi att prestanda betyder att systemet fungerar som du och dina användare förväntar sig, inklusive att inte generera skadliga resultat. Det är viktigt att bedöma prestandan för din övergripande applikation med hjälp av [Evaluatorer för Prestanda och Kvalitet samt Risk och Säkerhet](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Du har också möjlighet att skapa och utvärdera med [anpassade evaluatorer](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +En annan aspekt att ta hänsyn till är den övergripande applikationsprestandan. Med multimodala och multimodellsapplikationer avser vi med prestanda att systemet fungerar som du och dina användare förväntar sig, inklusive att det inte genererar skadliga resultat. Det är viktigt att bedöma prestandan för din totala applikation med hjälp av [Evaluatörer för prestanda och kvalitet samt risk och säkerhet](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Du har också möjlighet att skapa och evaluera med [egna evaluatörer](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Du kan utvärdera din AI-applikation i din utvecklingsmiljö med hjälp av [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Med antingen en testdatamängd eller ett mål mäts dina generativa AI-applikationsgenereringar kvantitativt med inbyggda evaluatorer eller anpassade evaluatorer efter eget val. För att komma igång med Azure AI Evaluation SDK för att utvärdera ditt system kan du följa [snabbstartsguiden](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). När du har genomfört en utvärderingskörning kan du [visualisera resultaten i Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Du kan utvärdera din AI-applikation i din utvecklingsmiljö med hjälp av [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Med antingen en testdatamängd eller ett mål mäts dina generativa AI-applikationsgenereringar kvantitativt med inbyggda evaluatörer eller egna evaluatörer du väljer. För att komma igång med Azure AI Evaluation SDK för att utvärdera ditt system kan du följa [quickstart-guiden](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). När du har genomfört ett utvärderingskör kan du [visualisera resultaten i Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Varumärken -Detta projekt kan innehålla varumärken eller logotyper för projekt, produkter eller tjänster. Auktoriserad användning av Microsofts varumärken eller logotyper är föremål för och måste följa [Microsofts riktlinjer för varumärken och varumärkesprofil](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Användning av Microsofts varumärken eller logotyper i modifierade versioner av detta projekt får inte skapa förvirring eller antyda Microsofts sponsring. All användning av tredjeparts varumärken eller logotyper är föremål för dessa tredjeparts policyer. +Det här projektet kan innehålla varumärken eller logotyper för projekt, produkter eller tjänster. Auktoriserad användning av Microsofts varumärken eller logotyper är underställd och måste följa [Microsofts riktlinjer för varumärken och varumärkesanvändning](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Användning av Microsofts varumärken eller logotyper i modifierade versioner av detta projekt får inte leda till förvirring eller antyda att Microsoft sponsrar. All användning av tredjeparts varumärken eller logotyper lyder under dessa tredje parts policyer. ## Få hjälp -Om du kör fast eller har frågor om att bygga AI-appar, gå med i: +Om du fastnar eller har frågor om att bygga AI-appar, gå med i: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Om du har produktfeedback eller stöter på fel medan du bygger, besök: +Om du har produktfeedback eller stöter på fel under byggandet, besök: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) @@ -320,5 +325,5 @@ Om du har produktfeedback eller stöter på fel medan du bygger, besök: **Ansvarsfriskrivning**: -Detta dokument har översatts med hjälp av AI-översättningstjänsten [Co-op Translator](https://github.com/Azure/co-op-translator). Även om vi strävar efter noggrannhet, var vänlig observera att automatiska översättningar kan innehålla fel eller brister. Det ursprungliga dokumentet på dess modersmål bör betraktas som den auktoritativa källan. För kritisk information rekommenderas professionell mänsklig översättning. Vi ansvarar inte för eventuella missförstånd eller feltolkningar som uppstår från användningen av denna översättning. +Detta dokument har översatts med hjälp av AI-översättningstjänsten [Co-op Translator](https://github.com/Azure/co-op-translator). Även om vi strävar efter noggrannhet, var vänlig notera att automatiska översättningar kan innehålla fel eller brister. Det ursprungliga dokumentet på dess modersmål bör betraktas som den auktoritativa källan. För kritisk information rekommenderas professionell mänsklig översättning. Vi ansvarar inte för några missförstånd eller feltolkningar som uppstår till följd av användningen av denna översättning. \ No newline at end of file diff --git a/translations/sv/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/sv/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..63486a2e9 --- /dev/null +++ b/translations/sv/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrid RAG med Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Författare:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Fokus:** Små Språkmodeller (SLM), SQLite FTS5 BM25, Täta Embeddings, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivation: Problemet med Nyckelordsåterkallning i Lokala SLM\n", + "Standard RAG-arkitekturer som enbart bygger på täta vektor-embeddings misslyckas ofta med att återkalla exakta numeriska token (t.ex. `2,340,000 TL`, kontraktskoder, kontonummer). \n", + "Däremot missar sparsam lexikal sökning (BM25) semantiska synonymer och omformulerade frågor.\n", + "\n", + "Denna kokbok visar hur man implementerar en **högpresterande, minnesbaserad hybridåterhämtningsmotor** som kombinerar:\n", + "1. **Täta vektorer** (Cosine Similarity)\n", + "2. **Spars lexikal sökning** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Förankrad citeringsgenerering (`[1]`, `[2]`)** med Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Dubbel SQLite-schema (täta vektorer + virtuell FTS5 BM25-tabell)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Exempelinsamling och körningsbenchmark\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Grundad uppmaningsformulering för Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Ansvarsfriskrivning**:\nDetta dokument har översatts med hjälp av AI-översättningstjänsten [Co-op Translator](https://github.com/Azure/co-op-translator). Även om vi strävar efter noggrannhet, var vänlig notera att automatiska översättningar kan innehålla fel eller brister. Det ursprungliga dokumentet på dess modersmål bör betraktas som den auktoritativa källan. För kritisk information rekommenderas professionell mänsklig översättning. Vi ansvarar inte för några missförstånd eller feltolkningar som uppstår till följd av användningen av denna översättning.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/sw/.co-op-translator.json b/translations/sw/.co-op-translator.json index 6b0c92969..e78ed7a38 100644 --- a/translations/sw/.co-op-translator.json +++ b/translations/sw/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "sw" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:22:35+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:36:29+00:00", "source_file": "README.md", "language_code": "sw" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "sw" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:41:42+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "sw" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:49+00:00", diff --git a/translations/sw/README.md b/translations/sw/README.md index 018b8d520..fef4f27d1 100644 --- a/translations/sw/README.md +++ b/translations/sw/README.md @@ -3,40 +3,40 @@ [![Fungua na tumia sampuli katika GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Fungua katika Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![Wachangiaji wa GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![Waibukizi wa GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![Masuala ya GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Maombi ya kuvuta GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Karibu](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Maombi ya GitHub ya kuvuta](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![Karibu PRs](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![Watazamaji wa GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![Fikicha za GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![Tawi za GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![Nyota za GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi ni mfululizo wa mifano ya AI ya chanzo huria iliyotengenezwa na Microsoft. +Phi ni mfululizo wa mifano ya AI ya chanzo wazi iliyotengenezwa na Microsoft. -Phi kwa sasa ni mfano mdogo wa lugha (SLM) wenye nguvu zaidi na wa gharama nafuu, wenye viwango bora katika lugha nyingi, reasoning, uzalishaji wa maandishi/mazungumzo, upangaji wa msimbo, picha, sauti na hali nyinginezo. +Phi kwa sasa ndio mfano mdogo wa lugha (SLM) wenye nguvu zaidi na gharama nafuu, na viwango bora sana katika lugha nyingi, uelewa, uzalishaji wa maandishi/mazungumzo, uandishi wa programu, picha, sauti na matukio mengine. -Unaweza kupeleka Phi mawinguni au kwenye vifaa vya edge, na unaweza kwa urahisi kujenga programu za AI za uzalishaji na nguvu chache za kompyuta. +Unaweza kupeleka Phi kwenye wingu au vifaa vya edge, na unaweza kwa urahisi kujenga programu za AI zinazozalisha kwa nguvu kidogo za kompyuta. Fuata hatua hizi kuanza kutumia rasilimali hizi: -1. **Fungua Tawi**: Bonyeza [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +1. **Fungua Hazina**: Bonyeza [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Nakili Hazina**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Jiunge na Jamii ya Microsoft AI Discord na kutana na wataalamu na waendelezaji wenzako**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Jiunge na Jumuiya ya Microsoft AI Discord na kukutana na wataalamu na watengenezaji wenza**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![jalada](../../translated_images/sw/cover.eb18d1b9605d754b.webp) +![cover](../../translated_images/sw/cover.eb18d1b9605d754b.webp) -### 🌐 Msaada wa Lugha nyingi +### 🌐 Msaada wa Lugha Nyingi -#### Umeungwa mkono kupitia Hatua ya GitHub (mojawapo & daima ya kisasa) +#### Inasaidiwa kupitia Kitendo cha GitHub (Kiotomatiki na Kila Wakati Kipya) -[Kiarabu](../ar/README.md) | [Kibengali](../bn/README.md) | [Kibolgaria](../bg/README.md) | [Kibama (Myanmar)](../my/README.md) | [Kichina (Iliyorahisishwa)](../zh-CN/README.md) | [Kichina (Kisasa, Hong Kong)](../zh-HK/README.md) | [Kichina (Kisasa, Macau)](../zh-MO/README.md) | [Kichina (Kisasa, Taiwan)](../zh-TW/README.md) | [Kroeshia](../hr/README.md) | [Kicheki](../cs/README.md) | [Kidenmaki](../da/README.md) | [Kiholanzi](../nl/README.md) | [Kiestonia](../et/README.md) | [Kifini](../fi/README.md) | [Kifaransa](../fr/README.md) | [Kijerumani](../de/README.md) | [Kigiriki](../el/README.md) | [Kiebrania](../he/README.md) | [Kihindi](../hi/README.md) | [Kihungaria](../hu/README.md) | [Kiindonesia](../id/README.md) | [Kiitaliano](../it/README.md) | [Kijapani](../ja/README.md) | [Kikannada](../kn/README.md) | [Kikmere](../km/README.md) | [Kikorea](../ko/README.md) | [Kilithuania](../lt/README.md) | [Kimelayu](../ms/README.md) | [Kimalayalam](../ml/README.md) | [Kimarathi](../mr/README.md) | [Kinepali](../ne/README.md) | [Pidgin ya Nijeria](../pcm/README.md) | [Kinorwe](../no/README.md) | [Kiajemi (Farsi)](../fa/README.md) | [Kipolishi](../pl/README.md) | [Kireno (Brazil)](../pt-BR/README.md) | [Kireno (Ureno)](../pt-PT/README.md) | [Kipunjabi (Gurmukhi)](../pa/README.md) | [Kiromania](../ro/README.md) | [Kirusi](../ru/README.md) | [Kiserbia (Sirilik)](../sr/README.md) | [Kislovakia](../sk/README.md) | [Kislovenia](../sl/README.md) | [Kihispania](../es/README.md) | [Kiswahili](./README.md) | [Kiswidi](../sv/README.md) | [Kitagalogi (Kifilipino)](../tl/README.md) | [Kitamili](../ta/README.md) | [Kitelugu](../te/README.md) | [Kithai](../th/README.md) | [Kituruki](../tr/README.md) | [Kiukraini](../uk/README.md) | [Kiurdu](../ur/README.md) | [Kivenetiamu](../vi/README.md) +[Kiarabu](../ar/README.md) | [Kibengali](../bn/README.md) | [Kibolgaria](../bg/README.md) | [Kiburma (Myanma)](../my/README.md) | [Kichina (Rahisi)](../zh-CN/README.md) | [Kichina (Kiasili, Hong Kong)](../zh-HK/README.md) | [Kichina (Kiasili, Macau)](../zh-MO/README.md) | [Kichina (Kiasili, Taiwan)](../zh-TW/README.md) | [Kikroeshia](../hr/README.md) | [Kicheki](../cs/README.md) | [Kidani](../da/README.md) | [Kiholanzi](../nl/README.md) | [Kiestonia](../et/README.md) | [Kifini](../fi/README.md) | [Kifaransa](../fr/README.md) | [Kijerumani](../de/README.md) | [Kigiriki](../el/README.md) | [Kiebrania](../he/README.md) | [Kihindi](../hi/README.md) | [Kihungari](../hu/README.md) | [Kiindonesia](../id/README.md) | [Kiitaliano](../it/README.md) | [Kijapani](../ja/README.md) | [Kikannada](../kn/README.md) | [Kikhmer](../km/README.md) | [Kikorea](../ko/README.md) | [Kilithuania](../lt/README.md) | [Kimalae](../ms/README.md) | [Kimalayalam](../ml/README.md) | [Kimarathi](../mr/README.md) | [Kinepali](../ne/README.md) | [Pijin ya Nigeria](../pcm/README.md) | [Kinorwe](../no/README.md) | [Kiajemi (Farsi)](../fa/README.md) | [Kipolandi](../pl/README.md) | [Kireno (Brazil)](../pt-BR/README.md) | [Kireno (Ureno)](../pt-PT/README.md) | [Kipunjabi (Gurmukhi)](../pa/README.md) | [Kiromania](../ro/README.md) | [Kirusi](../ru/README.md) | [Kiserbia (Kiriliki)](../sr/README.md) | [Kislovakia](../sk/README.md) | [Kislovenia](../sl/README.md) | [Kihispania](../es/README.md) | [Kiswahili](./README.md) | [Kiswidi](../sv/README.md) | [Kitagalog (Kifilipino)](../tl/README.md) | [Kitamili](../ta/README.md) | [Kitelugu](../te/README.md) | [Kithai](../th/README.md) | [Kituruki](../tr/README.md) | [Kiukraine](../uk/README.md) | [Kiurdu](../ur/README.md) | [Kivietinamu](../vi/README.md) -> **Ungependa Kunakili Kwa Ndani?** +> **Unapendelea Nakili Kwenye Kompyuta Mpaka?** > -> Hazina hii ina tafsiri zaidi ya 50 za lugha ambazo huongeza ukubwa wa kupakua. Ili nakili bila tafsiri, tumia sparse checkout: +> Hazina hii ina tafsiri za lugha zaidi ya 50 ambazo huongeza sana ukubwa wa kupakua. Ili kunakili bila tafsiri, tumia sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,202 +52,205 @@ Fuata hatua hizi kuanza kutumia rasilimali hizi: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Hii inakupa kila kitu unachohitaji ili kukamilisha kozi kwa upakuaji wa kasi zaidi. +> Hii inakupa kila kitu unachohitaji kukamilisha kozi kwa kupakua kwa kasi zaidi. ## Jedwali la Yaliyomo - Utangulizi - - [Karibu kwa Familia ya Phi](./md/01.Introduction/01/01.PhiFamily.md) + - [Karibu katika Familia ya Phi](./md/01.Introduction/01/01.PhiFamily.md) - [Kuweka mazingira yako](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Kuelewa Teknolojia Muhimu](./md/01.Introduction/01/01.Understandingtech.md) - [Usalama wa AI kwa Mifano ya Phi](./md/01.Introduction/01/01.AISafety.md) - [Msaada wa Vifaa vya Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Mifano ya Phi na Upatikana kwenye majukwaa](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Mifano ya Phi & Upatikanaji kwenye majukwaa](./md/01.Introduction/01/01.Edgeandcloud.md) - [Kutumia Guidance-ai na Phi](./md/01.Introduction/01/01.Guidance.md) - [Mifano ya Soko la GitHub](https://github.com/marketplace/models) - - [Katalogi ya Mfano wa AI ya Azure](https://ai.azure.com) + - [Katalogi ya Mifano ya Azure AI](https://ai.azure.com) -- Kutathmini Phi katika mazingira tofauti +- Kutoa maamuzi ya Phi katika mazingira tofauti - [Hugging face](./md/01.Introduction/02/01.HF.md) - [Mifano ya GitHub](./md/01.Introduction/02/02.GitHubModel.md) - - [Katalogi ya Mfano ya Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Katalogi ya Mfano wa Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Kutathmini Familia ya Phi - - [Kutathmini Phi katika iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Kutathmini Phi katika Android](./md/01.Introduction/03/Android_Inference.md) - - [Kutathmini Phi katika Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Kutathmini Phi katika AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Kutathmini Phi na Mfumo wa Apple MLX](./md/01.Introduction/03/MLX_Inference.md) - - [Kutathmini Phi katika Server ya Ndani](./md/01.Introduction/03/Local_Server_Inference.md) - - [Kutathmini Phi katika Server ya Mbali kwa kutumia AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Kutathmini Phi na Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Kutathmini Phi--Mwangaza Ndani](./md/01.Introduction/03/Vision_Inference.md) - - [Kutathmini Phi na Kaito AKS, Vyombo vya Azure (msaada rasmi)](./md/01.Introduction/03/Kaito_Inference.md) +- Kutoa maamuzi Familia ya Phi + - [Kutoa maamuzi Phi katika iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Kutoa maamuzi Phi katika Android](./md/01.Introduction/03/Android_Inference.md) + - [Kutoa maamuzi Phi katika Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Kutoa maamuzi Phi katika AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Kutoa maamuzi Phi kwa kutumia Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Kutoa maamuzi Phi katika Server ya Ndani](./md/01.Introduction/03/Local_Server_Inference.md) + - [Kutoa maamuzi Phi katika Server ya Mbali kwa kutumia AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Kutoa maamuzi Phi kwa Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Kutoa maamuzi Phi--Maono ndani](./md/01.Introduction/03/Vision_Inference.md) + - [Kutoa maamuzi Phi na Kaito AKS, Azure Containers (msaada rasmi)](./md/01.Introduction/03/Kaito_Inference.md) - [Kupima Familia ya Phi](./md/01.Introduction/04/QuantifyingPhi.md) - - [Kupima Phi-3.5 / 4 kwa kutumia llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Kupima Phi-3.5 / 4 kwa kutumia nyongeza za AI za uzalishaji kwa onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Kupima Phi-3.5 / 4 kwa kutumia Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Kupima Phi-3.5 / 4 kwa kutumia Mfumo wa Apple MLX](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Kurekebisha Phi-3.5 / 4 kwa kutumia llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Kurekebisha Phi-3.5 / 4 kwa kutumia nyongeza za AI za kuanzisha kwa onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Kurekebisha Phi-3.5 / 4 kwa kutumia Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Kurekebisha Phi-3.5 / 4 kwa kutumia Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- Tathmini ya Phi - - [AI Inayowajibika](./md/01.Introduction/05/ResponsibleAI.md) +- Tathmini Phi + - [AI ya Kuwajibika](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry kwa Tathmini](./md/01.Introduction/05/AIFoundry.md) - [Kutumia Promptflow kwa Tathmini](./md/01.Introduction/05/Promptflow.md) - RAG na Azure AI Search - - [Jinsi ya kutumia Phi-4-mini na Phi-4-multimodal(RAG) na Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Jinsi ya kutumia Phi-4-mini na Phi-4-multimodal(RAG) pamoja na Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG ya Mchanganyiko wa Zero-Wingu Ndani na SQLite FTS5 na phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Sampuli za maendeleo ya programu za Phi - - Programu za Maandishi na Mazungumzo - - Sampuli za Phi-4 - - [📓] [Mazungumzo na Mfano wa Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Mazungumzo na Mfano wa Phi-4 wa ndani ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Programu ya mazungumzo ya Console ya .NET na Phi-4 ONNX kwa kutumia Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Sampuli za Phi-3 / 3.5 - - [Chatbot wa Ndani katika kivinjari kwa kutumia Phi3, ONNX Runtime Web na WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - Phi-3-mini ya Kuingiliana na OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Kujenga kifuniko na kutumia Phi-3 na MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - Programu za Maandishi & Mazungumzo + - Sampuli za Phi-4 + - [📓] [Mzungumze na Mfano wa Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Mzungumze na Mfano wa Phi-4 wa ndani wa ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Programu ya konsoli ya mzungumzo .NET na Phi-4 ONNX kwa kutumia Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Sampuli za Phi-3 / 3.5 + - [Bot ya mazungumzo ya ndani kwenye kivinjari ikitumia Phi3, ONNX Runtime Web na WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Mazungumzo ya OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Mfano wa Multi Model - Phi-3-mini ya Kuisiliana na OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Kuunda kifuniko na kutumia Phi-3 na MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [Uboreshaji wa Mfano - Jinsi ya kuboresha mfano wa Phi-3-min kwa ONNX Runtime Web na Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 App na Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 Mfano wa AI wa Multimodeli Powered Notes App](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Kufinyaza na Kuunganisha mifano ya desturi ya Phi-3 na Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Kufinyaza na Kuunganisha mifano ya desturi ya Phi-3 na Prompt flow katika Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Kutathmini Mfano wa Phi-3 / Phi-3.5 uliobinafsishwa katika Microsoft Foundry ukizingatia Kanuni za Haki za Microsoft za AI](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [Programu ya WinUI3 na Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[Mfano wa Programu ya WinUI3 Multi Model AI Powered Notes](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Tengeneza upya na Unganisha mifano maalum ya Phi-3 na mtiririko wa Prompt](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Tengeneza upya na Unganisha mifano maalum ya Phi-3 na mtiririko wa Prompt katika Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Pima Mfano wa Phi-3 / Phi-3.5 ulioboreshwa katika Microsoft Foundry ukizingatia Kanuni za AI za Kuwajibika za Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Mfano wa utabiri wa lugha wa Phi-3.5-mini-instruct (Kichina/Kiingereza)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Kutumia Windows GPU kuunda suluhisho la Prompt flow na Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Bot ya Mazungumzo ya Phi-3.5-Instruct WebGPU RAG](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Kutumia Windows GPU kuunda suluhisho la mtiririko wa Prompt na Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Kutumia Microsoft Phi-3.5 tflite kuunda programu ya Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A Mfano wa .NET kutumia mfano wa ndani wa ONNX Phi-3 kwa kutumia Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Programu ya mazungumzo ya Console .NET na Semantic Kernel na Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Mfano wa maswali na majibu .NET ukitumia mfano wa Phi-3 wa ndani wa ONNX na Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Programu ya mazungumzo ya console .NET na Semantic Kernel na Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Sampuli za Azure AI Inference SDK zinazotegemea Msimbo + - Sampuli za Azure AI Inference SDK Zinazotumia Msimbo - Sampuli za Phi-4 - - [📓] [Tengeneza msimbo wa mradi kwa kutumia Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [📓] [Tengeneza msimbo wa mradi ukitumia Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Sampuli za Phi-3 / 3.5 - - [Jenga mazungumzo yako ya Visual Studio Code GitHub Copilot na Familia ya Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Unda Wakala wa Mazungumzo wa Visual Studio Code Copilot na Phi-3.5 kwa kutumia Mifano ya GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Jenga GitHub Copilot Chat yako mwenyewe ya Visual Studio Code na Familia ya Phi-3 ya Microsoft](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Unda Wakala wa Chat Copilot wa Visual Studio Code na Phi-3.5 kwa kutumia Mifano ya GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - Sampuli za Ufafanuzi wa Juu + - Sampuli za Ufahamu wa Juu - Sampuli za Phi-4 - - [📓] [Sampuli za Phi-4-mini-ufafanuzi au Phi-4-ufafanuzi](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Kufinyaza Phi-4-mini-ufafanuzi kwa Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Kufinyaza Phi-4-mini-ufafanuzi kwa Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-ufafanuzi na Mifano ya GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-ufafanuzi na Mifano ya Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Maonesho - - [Demos za Phi-4-mini zilizoandaliwa kwenye Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Demos za Phi-4-multimodal zilizoandaliwa kwenye Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [📓] [Sampuli za ufahamu wa Phi-4-mini au Phi-4](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Kuongeza usahihi wa Phi-4-mini-ufahamu kwa Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Kuongeza usahihi wa Phi-4-mini-ufahamu kwa Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-ufahamu na Mifano ya GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-ufahamu na Mifano ya Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Maonyesho + - [Maonyesho ya Phi-4-mini yanayosimamiwa kwenye Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Maonyesho ya Phi-4-multimodal yanayosimamiwa kwenye Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - Sampuli za Maono - Sampuli za Phi-4 - - [📓] [Tumia Phi-4-multimodal kusoma picha na kutoa msimbo](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [📓] [Tumia Phi-4-multimodal kusoma picha na kuunda msimbo](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Sampuli za Phi-3 / 3.5 - - [📓][Phi-3-maono-Picha maandishi kwa maandishi](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-maono-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-maono CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Urejeleaji wa Phi-3](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-maono - Msaidizi wa lugha wa kuona - na Phi3-Maono na OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Maono Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Maono OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Maono mfano wa multi-frame au picha nyingi](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Maono Mfano wa Ndani wa ONNX kutumia Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menyu inayoanzisha Mfano wa Ndani wa ONNX wa Phi-3 Maono kutumia Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Sampuli za Ufafanuzi-Maono - - Phi-4-Ufafanuzi-Maono-15B - - [📓] [Kutumia Phi-4-Ufafanuzi-Maono-15B kugundua kupita barabara mahali pasipo halali](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Kutumia Phi-4-Ufafanuzi-Maono-15B kwa hesabu](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Kutumia Phi-4-Ufafanuzi-Maono-15B kugundua UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓][Phi-3-vision-Picha maandishi hadi maandishi](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vision CLIP Uingizaji](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [ONYESHO: Urejelezaji wa Phi-3](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - Msaidizi wa lugha wa kuona - na Phi3-Vision na OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Mfano wa kielelezo cha picha nyingi au fremu nyingi wa Phi-3.5 Vision](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Mfano wa Ndani wa Phi-3 Vision Local ONNX ukitumia Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Mfano wa Phi-3 Vision Local ONNX unaotegemea menyu ukitumia Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Sampuli za Ufahamu wa Maono + - Phi-4-Ufahamu-Maono-15B + - [📓] [Kutumia Phi-4-Ufahamu-Maono-15B kugundua kutembea kupita barabara kibinafsi](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Kutumia Phi-4-Ufahamu-Maono-15B kwa hesabu](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Kutumia Phi-4-Ufahamu-Maono-15B kugundua UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Sampuli za Hisabati - - Sampuli za Phi-4-Mini-Flash-Ufafanuzi-Maelekezo [Demo ya Hisabati na Phi-4-Mini-Flash-Ufafanuzi-Maelekezo](./md/02.Application/09.Math/MathDemo.ipynb) + - Sampuli za Phi-4-Mini-Flash-Reasoning-Instruct [Onyesho la Hisabati na Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - Sampuli za Sauti - Sampuli za Phi-4 - [📓] [Kutoa maandishi ya sauti kwa kutumia Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Mfano wa Sauti wa Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Mfano wa Tafsiri ya Hotuba ya Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [Programu ya console ya .NET ikitumia Phi-4-multimodal ya Sauti kuchambua faili la sauti na kutoa maandishi](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [📓] [Mfano wa Tafsiri ya Hotuba wa Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [Programu ya console .NET ikitumia Phi-4-multimodal Sauti kuchambua faili la sauti na kutengeneza maandishi](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - Sampuli za MOE - Sampuli za Phi-3 / 3.5 - - [📓] [Mchanganyiko wa Mifano wa Wataalam wa Phi-3.5 (MoEs) Mfano wa Mitandao ya Kijamii](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Kujenga Mlolongo wa Uzalishaji Uliongezwa kwa Utafutaji (RAG) na NVIDIA NIM Phi-3 MOE, Azure AI Search, na LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Mfano wa Mchanganyiko wa Wataalam wa Phi-3.5 (MoEs) kwa Mitandao ya Kijamii](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Kuunda Mlolongo wa Uzalishaji Ulioboresha (RAG) na NVIDIA NIM Phi-3 MOE, Azure AI Search, na LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Sampuli za Kupiga Simu za Vitendo + - Sampuli za Kupigia Simu Kazi - Sampuli za Phi-4 🆕 - - [📓] [Kutumia Kupiga Simu za Vitendo Na Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Kutumia Kupiga Simu za Vitendo kuunda mawakala wengi Na Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Kutumia Kupiga Simu za Vitendo na Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Kutumia Kupiga Simu za Vitendo na ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - [📓] [Kutumia Kupigia Simu Kazi na Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Kutumia Kupigia Simu Kazi kuunda mawakala wengi na Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Kutumia Kupigia Simu Kazi na Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Kutumia Kupigia Simu Kazi na ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - Sampuli za Mchanganyiko wa Multimodal - Sampuli za Phi-4 🆕 - - [📓] [Kutumia Phi-4-multimodal kama Mwandishi wa Teknolojia](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [Programu ya console ya .NET ikitumia Phi-4-multimodal kuchambua picha](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Kufinyaza Samples za Phi - - [Hali za Kufinyaza](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Kufinyaza dhidi ya RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Kufinyaza Ili Phi-3 aweke ujuzi wa sekta](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Kufinyaza Phi-3 na AI Toolkit kwa VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Kufinyaza Phi-3 kwa Huduma ya Azure Machine Learning](./md/03.FineTuning/Introduce_AzureML.md) - - [Kufinyaza Phi-3 na Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Kufinyaza Phi-3 na QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Kufinyaza Phi-3 na Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Kufinyaza Phi-3 na Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Kufinyaza na Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Kufinyaza na Maabara ya Mikono ya Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [Kufinyaza Phi-3-maono na Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Kufinyaza Phi-3 na Mfumo wa Apple MLX](./md/03.FineTuning/FineTuning_MLX.md) - - [Kufinyaza Phi-3-maono (msaada rasmi)](./md/03.FineTuning/FineTuning_Vision.md) - - [Urekebishaji wa Phi-3 na Kaito AKS, Kontena za Azure (Msaada Rasmi)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Urekebishaji wa Phi-3 na 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - -- Maabara ya Vitendo - - [Kutafakari mifano ya kisasa: LLMs, SLMs, maendeleo ya ndani na zaidi](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Kufungua Uwezo wa NLP: Urekebishaji kwa Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) - -- Makala za Utafiti wa Kielimu na Machapisho - - [Vitabu Pekee Ndiyo Inayohitajika II: ripoti ya kiteknolojia ya phi-1.5](https://arxiv.org/abs/2309.05463) - - [Ripoti ya Kiteknolojia ya Phi-3: Mfano wa Lugha wenye Uwezo Mkubwa Kwenye Simu Yako](https://arxiv.org/abs/2404.14219) - - [Ripoti ya Kiteknolojia ya Phi-4](https://arxiv.org/abs/2412.08905) - - [Ripoti ya Kiteknolojia ya Phi-4-Mini: Mifano Midogo Lakini Yenye Nguvu ya Lugha Mtandao kupitia Mchanganyiko wa LoRAs](https://arxiv.org/abs/2503.01743) - - [Kuboresha Mifano Midogo ya Lugha kwa Kupiga Simu za Kazi Ndani ya Gari](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Urekebishaji wa PHI-3 kwa Majibu ya Maswali ya Chaguo Nyingi: Mbinu, Matokeo, na Changamoto](https://arxiv.org/abs/2501.01588) - - [Ripoti ya Kiteknolojia ya Phi-4-uwezo wa kufikiri](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Ripoti ya Kiteknolojia ya Phi-4-mini-uwezo wa kufikiri](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [📓] [Kutumia Phi-4-multimodal kama Mwanahabari wa Teknolojia](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [Programu ya console .NET ikitumia Phi-4-multimodal kuchambua picha](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Kufinya Mifano ya Phi + - [Mazingira ya Kufinya](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Kufinya vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Kufinya Acha Phi-3 awe mtaalamu wa viwanda](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Kufinya Phi-3 na AI Toolkit kwa VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Kufinya Phi-3 na Huduma ya Azure Machine Learning](./md/03.FineTuning/Introduce_AzureML.md) + - [Kufinya Phi-3 na Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Kufinya Phi-3 na QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Kufinya Phi-3 na Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Kufinya Phi-3 na Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Kufinya na Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Kufinya na mafunzo ya Mikono ya Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) + - [Kufinya Phi-3-vision na Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Kurekebisha Phi-3 kwa kutumia Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Kurekebisha Phi-3-vision (msaada rasmi)](./md/03.FineTuning/FineTuning_Vision.md) + - [Kurekebisha Phi-3 kwa Kaito AKS , Kontena za Azure (Msaada Rasmi)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Kurekebisha Phi-3 na 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + +- Kazi ya Mikono Maabara + - [Kuchunguza mifano ya hali ya juu: LLMs, SLMs, maendeleo ya hapa hapa na zaidi](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Kufungua Uwezo wa NLP: Kurekebisha kwa Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + +- Karatasi za Utafiti za Kielimu na Machapisho + - [Vitabu ni Vyote Unavyohitaji II: ripoti ya kiufundi ya phi-1.5](https://arxiv.org/abs/2309.05463) + - [Ripoti ya Kiufundi ya Phi-3: Mfano wa Lugha Ulio na Uwezo Mkubwa Kwenye Simu Yako](https://arxiv.org/abs/2404.14219) + - [Ripoti ya Kiufundi ya Phi-4](https://arxiv.org/abs/2412.08905) + - [Ripoti ya Kiufundi ya Phi-4-Mini: Mifano Midogo Lakini Yenye Nguvu ya Lugha Multi-modality Kupitia Mchanganyiko wa LoRAs](https://arxiv.org/abs/2503.01743) + - [Kuboresha Mifano Midogo ya Lugha kwa Kuitisha Kazi Ndani ya Gari](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Kurekebisha PHI-3 kwa Majibu ya Maswali ya Chaguo Nyingi: Mbinu, Matokeo, na Changamoto](https://arxiv.org/abs/2501.01588) + - [Ripoti ya Kiufundi ya Phi-4-Reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Ripoti ya Kiufundi ya Phi-4-Mini-Reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Kutumia Mifano ya Phi ### Phi kwenye Microsoft Foundry -Unaweza kujifunza jinsi ya kutumia Microsoft Phi na jinsi ya kujenga suluhisho za E2E katika vifaa vyako tofauti vya vifaa. Ili kujionea Phi mwenyewe, anza kwa kucheza na mifano na kubinafsisha Phi kwa hali zako ukitumia [Katalogi ya Mfano wa AI ya Microsoft Foundry Azure](https://aka.ms/phi3-azure-ai) unaweza kujifunza zaidi katika Kuanzisha na [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Unaweza kujifunza jinsi ya kutumia Microsoft Phi na jinsi ya kujenga suluhisho za E2E katika vifaa vyako tofauti vya vifaa. Ili kujaribu Phi kwa mwenyewe, anza kwa kucheza na mifano na kubinafsisha Phi kwa hali zako kwa kutumia [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) unaweza kujifunza zaidi katika Kuanzia na [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Uwanja wa Mchezo** -Kila mfano una uwanja wake wa majaribio wa kujaribu mfano huo [Azure AI Playground](https://aka.ms/try-phi3). +**Sehemu ya Kucheza** +Kila mfano una sehemu maalum ya kujaribu mfano [Azure AI Playground](https://aka.ms/try-phi3). ### Phi kwenye Mifano ya GitHub -Unaweza kujifunza jinsi ya kutumia Microsoft Phi na jinsi ya kujenga suluhisho za E2E katika vifaa vyako tofauti vya vifaa. Ili kujionea Phi mwenyewe, anza kwa kucheza na mfano na kubinafsisha Phi kwa hali zako ukitumia [Katalogi ya Mfano ya GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) unaweza kujifunza zaidi katika Kuanzisha na [Katalogi ya Mfano ya GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) +Unaweza kujifunza jinsi ya kutumia Microsoft Phi na jinsi ya kujenga suluhisho za E2E katika vifaa vyako tofauti vya vifaa. Ili kujaribu Phi kwa mwenyewe, anza kwa kucheza na mfano na kubinafsisha Phi kwa hali zako kwa kutumia [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) unaweza kujifunza zaidi katika Kuanzia na [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Uwanja wa Mchezo** -Kila mfano una uwanja wa majaribio wa [kujaribu mfano](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Sehemu ya Kucheza** +Kila mfano una [sehemu maalum ya kujaribu mfano](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi kwenye Hugging Face -Unaweza pia kupata mfano kwenye [Hugging Face](https://huggingface.co/microsoft) +Pia unaweza kupata mfano kwenye [Hugging Face](https://huggingface.co/microsoft) -**Uwanja wa Mchezo** - [Uwanja wa Mchezo wa Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Sehemu ya Kucheza** + [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) ## 🎒 Kozi Nyingine -Timu yetu huandaa kozi nyingine! Angalia: +Timu yetu hutengeneza kozi nyingine! Angalia: ### LangChain @@ -256,19 +259,19 @@ Timu yetu huandaa kozi nyingine! Angalia: [![LangChain kwa Waanzilishi](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Wakala +### Azure / Edge / MCP / Mawakala [![AZD kwa Waanzilishi](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI kwa Waanzilishi](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP kwa Waanzilishi](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Wakala wa AI kwa Waanzilishi](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Mawakala wa AI kwa Waanzilishi](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Mfululizo wa AI wa Kuumba -[![AI wa Kuumba kwa Waanzilishi](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI wa Kuumba (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![AI wa Kuumba (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![AI wa Kuumba (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### Mfululizo wa AI wa Kizazi +[![AI wa Kizazi kwa Waanzilishi](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI wa Kizazi (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![AI wa Kizazi (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![AI wa Kizazi (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- @@ -284,41 +287,43 @@ Timu yetu huandaa kozi nyingine! Angalia: --- ### Mfululizo wa Copilot -[![Copilot kwa Uandishi wa Programu Zaidi ya AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot kwa Uandishi wa Programu za AI kwa Pairs](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot kwa C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Ushindi wa Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## AI Iliyo Wajibu +## AI Inayohusika + +Microsoft imejizatiti kusaidia wateja wetu kutumia bidhaa zetu za AI kwa uwajibikaji, kushiriki uzoefu wetu, na kujenga ushirikiano unaotegemea uaminifu kupitia zana kama Vidokezo vya Uwajibikaji na Tathmini za Athari. Rasilimali nyingi za aina hii zinaweza kupatikana kwenye [https://aka.ms/RAI](https://aka.ms/RAI). +Mbinu ya Microsoft kwa AI inayohusika inatokana na kanuni zetu za AI za haki, kuaminika na usalama, faragha na usalama, ujumuishaji, uwazi, na uwajibikaji. + +Mifano mikubwa ya lugha asilia, picha, na sauti - kama ile inayotumika katika mfano huu - inaweza kuonekana kwa njia zisizo za haki, zisizo za kuaminika, au chukizo, na kusababisha madhara. Tafadhali kawaangalie [kidokezo cha uwazi cha huduma ya Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ili uwe na taarifa kuhusu hatari na vizingiti. -Microsoft imejizatiti kusaidia wateja wetu kutumia bidhaa zetu za AI kwa uwajibikaji, kushiriki masomo yetu, na kujenga ushirikiano wa msingi wa imani kupitia zana kama Vidokezo vya Uwajibikaji na Tathmini za Athari. Rasilimali nyingi za aina hii zinaweza kupatikana katika [https://aka.ms/RAI](https://aka.ms/RAI). -Mbinu ya Microsoft kuhusu AI yenye uwajibikaji imejikita katika kanuni zetu za AI za usawa, uaminifu na usalama, faragha na usalama, ujumuishaji, uwazi, na uwajibikaji. -Mifano mikubwa ya lugha asilia, picha, na sauti - kama ile inayotumika katika mfano huu - inaweza kutenda kwa njia zisizo za haki, zisizotegemewa, au zenye kukera, na kusababisha madhara. Tafadhali angalia [Kumbuka la Uwajibikaji wa Huduma ya Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ili ujue kuhusu hatari na mipaka. -Njia inayopendekezwa kupunguza hatari hizi ni kujumuisha mfumo wa usalama katika usanifu wako ambao unaweza kugundua na kuzuia tabia mbaya. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) hutoa tabaka huru la ulinzi, linaloweza kugundua maudhui hatarishi yaliyotengenezwa na watumiaji na AI katika programu na huduma. Azure AI Content Safety inajumuisha API za maandishi na picha ambazo zinakuwezesha kugundua nyenzo zenye madhara. Ndani ya Microsoft Foundry, huduma ya Content Safety inakuwezesha kuona, kuchunguza na kujaribu mfano wa msimbo wa kugundua maudhui hatarishi katika aina mbalimbali. Hati za [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) zitaelekeza hatua kwa hatua jinsi ya kutuma maombi kwa huduma hii. +Njia inayopendekezwa ya kupunguza hatari hizi ni kujumuisha mfumo wa usalama katika usanifu wako ambao unaweza kugundua na kuzuia tabia hatarishi. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) hutoa safu huru ya ulinzi, inayoweza kugundua yaliyomo hatarishi yanayotokana na watumiaji na AI katika programu na huduma. Azure AI Content Safety inajumuisha API za maandishi na picha zinazokuwezesha kugundua nyenzo zinazoweza kuwa hatari. Ndani ya Microsoft Foundry, huduma ya Content Safety inakuwezesha kuona, kuchunguza na kujaribu msimbo wa mfano wa kugundua yaliyomo hatarishi kwa njia mbalimbali. Nyaraka ifuatayo ya [mwongozo wa haraka](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) inakuonesha jinsi ya kufanya maombi kwenye huduma. -Jambo lingine la kuzingatia ni utendakazi wa jumla wa programu. Kwa programu zenye modal nyingi na modeli nyingi, tunachukulia utendakazi kama mfumo unavyoenda kama unavyotarajia wewe na watumiaji wako, ikijumuisha kutotengeneza matokeo yenye madhara. Ni muhimu kutathmini utendakazi wa programu yako kwa ujumla kwa kutumia [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Pia una uwezo wa kuunda na kutathmini kwa kutumia [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Jambo lingine la kuzingatia ni utendaji wa jumla wa programu. Kwa programu ambazo zinatumia njia nyingi na mifano mingi, tunazingatia utendaji kuwa mfumo unafanya kazi kama unavyotarajia wewe na watumiaji wako, ikiwa ni pamoja na kutozalisha matokeo hatarishi. Ni muhimu kutathmini utendaji wa programu yako kwa ujumla kwa kutumia [viongezi vya Utendaji, Ubora, Hatari, na Usalama](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Pia una uwezo wa kuunda na kutathmini kwa kutumia [viongezi maalum](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Unaweza kutathmini programu yako ya AI katika mazingira yako ya maendeleo kwa kutumia [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Ukiwa na seti ya data ya majaribio au lengo, uzalishaji wa programu yako ya generative AI hupimwa kwa kiasi kwa kutumia wataathiri waliopo au wataathiri wa kawaida wa chaguo lako. Ili kuanza kutumia azure ai evaluation sdk kutathmini mfumo wako, unaweza kufuata [mwongozo wa quickstart](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Mara tu utakapoendesha tathmini, unaweza [kuonyesha matokeo katika Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Unaweza kutathmini programu yako ya AI katika mazingira yako ya maendeleo kwa kutumia [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Ukipata seti ya majaribio au lengo, uzalishaji wa programu yako ya AI ya kizazi hupimwa kikumi kikubwa kwa viongezi vilivyotengenezwa au viongezi maalum unavyovipenda. Ili kuanza na azure ai evaluation sdk kutathmini mfumo wako, unaweza kufuata [mwongozo wa haraka](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Mara utakapoendesha tathmini, unaweza [kuona matokeo katika Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## Hati miliki +## Alama za Biashara -Mradi huu unaweza kuwa na alama za biashara au nembo za miradi, bidhaa, au huduma. Matumizi ya idhini ya alama za biashara au nembo za Microsoft yanategemea na lazima yafuatwe kwa mujibu wa [Miongozo ya Alama za Biashara & Brand za Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Matumizi ya alama za biashara au nembo za Microsoft katika toleo lililobadilishwa la mradi huu hayapaswi kusababisha mkanganyiko au kuashiria udhamini wa Microsoft. Matumizi yoyote ya alama za biashara au nembo za watu wengine yanategemea sera za pande hizo za tatu. +Mradi huu unaweza kuwa na alama za biashara au nembo za miradi, bidhaa, au huduma. Matumizi yaliyothibitishwa ya alama za biashara au nembo za Microsoft yanahusiana na na lazima yafuatilie [Miongozo ya Alama za Biashara na Brand ya Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Matumizi ya alama za biashara au nembo za Microsoft katika matoleo yaliyobadilishwa ya mradi huu hayapaswi kusababisha mkanganyiko au kuashiria udhamini wa Microsoft. Matumizi yoyote ya alama za biashara au nembo za watu wa tatu yanategemea sera za watu hao wa tatu. ## Kupata Msaada -Ikiwa utakumbwa na matatizo au una maswali yoyote kuhusu kujenga programu za AI, jiunge: +Ukikumbwa na changamoto au ikiwa una maswali kuhusu kujenga programu za AI, jiunge: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Ikiwa una maoni kuhusu bidhaa au makosa wakati wa ujenzi, tembelea: +Ikiwa una maoni au hitilafu kuhusu bidhaa wakati wa kujenga, tembelea: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Kisio cha lawama**: -Hati hii imetafsiriwa kwa kutumia huduma ya utafsiri wa AI [Co-op Translator](https://github.com/Azure/co-op-translator). Ingawa tunajitahidi kwa usahihi, tafadhali fahamu kwamba tafsiri za kiotomatiki zinaweza kuwa na makosa au usahihi mdogo. Hati ya asili katika lugha yake ya asili inapaswa kuchukuliwa kama chanzo chenye mamlaka. Kwa taarifa muhimu, inashauriwa kutumia utafsiri wa kibinadamu wa kitaalamu. Hatuwezi kuwajibika kwa jambo lolote la kutoelewana au tafsiri potofu zitokanazo na matumizi ya tafsiri hii. +**Kionyozo**: +Hati hii imetafsiriwa kwa kutumia huduma ya tafsiri ya AI [Co-op Translator](https://github.com/Azure/co-op-translator). Ingawa tunajitahidi kupata usahihi, tafadhali fahamu kwamba tafsiri za kiotomatiki zinaweza kuwa na makosa au upungufu wa usahihi. Hati ya asili katika lugha yake halisi inapaswa kuchukuliwa kama chanzo cha mamlaka. Kwa taarifa muhimu, tafsiri ya kitaalamu inayofanywa na binadamu inapendekezwa. Hatutojibu kwa kuelewa vibaya au tafsiri potofu zinazotokea kutokana na matumizi ya tafsiri hii. \ No newline at end of file diff --git a/translations/sw/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/sw/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..1f9647083 --- /dev/null +++ b/translations/sw/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrid RAG na Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Mwandishi:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Mazingira:** Modeli Ndogo za Lugha (SLMs), SQLite FTS5 BM25, Mchanganyiko wa Maundaji Yenye Mwinuko (Dense Embeddings), Mchanganyiko wa Daraja la Kurejea (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motisha: Dilema ya Kurejea Maneno Muhimu katika SLM za Ndani\n", + "Miundo ya kawaida ya RAG inayotegemea tu mchanganyiko wa vector zenye mwinuko mara nyingi huishia kushindwa kupata maneno sahihi ya nambari (mfano, `2,340,000 TL`, nambari za mikataba, nambari za akaunti). \n", + "Kinyume chake, utafutaji wa rafu wa maneno (BM25) hupoteza kielelezo cha maana na maswali yaliyotafsiriwa kwa maneno mengine.\n", + "\n", + "Kitabu hiki kinadhihirisha jinsi ya kutekeleza **injini ya Urejeaji ya Hybrid ya Kasi Juu, ya Kwenye Kumbukumbu** inayochanganya:\n", + "1. **Vectors zenye Mwinuko** (Ulinganisho wa Cosine)\n", + "2. **Utafutaji wa Rafu wa Maneno (Lexical) uliopachikwa (SQLite FTS5 BM25)**\n", + "3. **Mchanganyiko wa Daraja la Kurejea (RRF, $k=60$)**\n", + "4. **Uundaji wa Marejeo Yaliyo Thibitishwa (`[1]`, `[2]`)** na Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Muundo wa SQLite Mbili (Vectors Zenye Unene + Jedwali Halisi la FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Upimaji wa Mfano wa Uingilizi na Utekelezaji\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Uundaji wa Mwongozo Unaothibitishwa kwa Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Kionyozo**:\nHati hii imetafsiriwa kwa kutumia huduma ya tafsiri ya AI [Co-op Translator](https://github.com/Azure/co-op-translator). Ingawa tunajitahidi kupata usahihi, tafadhali fahamu kwamba tafsiri za kiotomatiki zinaweza kuwa na makosa au upungufu wa usahihi. Hati ya asili katika lugha yake halisi inapaswa kuchukuliwa kama chanzo cha mamlaka. Kwa taarifa muhimu, tafsiri ya kitaalamu inayofanywa na binadamu inapendekezwa. Hatutojibu kwa kuelewa vibaya au tafsiri potofu zinazotokea kutokana na matumizi ya tafsiri hii.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/ta/.co-op-translator.json b/translations/ta/.co-op-translator.json index 9c8bff41a..9dc027277 100644 --- a/translations/ta/.co-op-translator.json +++ b/translations/ta/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "ta" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T07:51:26+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:25:20+00:00", "source_file": "README.md", "language_code": "ta" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "ta" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:31:13+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "ta" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-10-11T11:33:28+00:00", diff --git a/translations/ta/README.md b/translations/ta/README.md index b9d28c3e2..bc215cc28 100644 --- a/translations/ta/README.md +++ b/translations/ta/README.md @@ -1,42 +1,42 @@ -# Phi வாணிகம்: Microsoft இன் Phi மாதிரிகளுடன் நேரடி உதாரணங்கள் +# Phi Cookbook: Microsoft இன் Phi மாதிரிகளுடன் கையால் செய்யும் உதாரணங்கள் -[![GitHub Codespaces இல் மாதிரிகளை திறந்து பயன்படுத்தவும்](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![GitHub Codespaces இல் மாதிரிகளைத் திறக்கவும் பயன்படுத்தவும்](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Dev Containers இல் திறக்கவும்](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub பங்களிப்பாளர்கள்](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub சிக்கல்கள்](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub புல் கோரிக்கைகள்](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs வரவேற்கப்படும்](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub பிரச்சனைகள்](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub புல்-ரிக்வஸ்டுகள்](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs வரவேற்கப்படுகின்றன](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![GitHub பார்வையாளர்கள்](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ஃபோர்க்](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub நட்சத்திரங்கள்](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub பங்கை பிரிக்கைகள்](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub நட்சூறுகள்](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi என்பது Microsoft உருவாக்கிய திறந்த மூல AI மாதிரிகளின் தொடர் ஆகும். +Phi என்பது Microsoft உருவாக்கிய திறந்த மூல AI மாதிரிகள் தொடர் ஆகும். -Phi தற்போது மிகவும் சக்திவாய்ந்ததும் செலவுகுறைந்த சிறிய மொழி மாதிரியாகும் (SLM), பன்மொழி, காரணம் காண்தல், உரை/chat உருவாக்கம், குறியீடு, படங்கள், ஒலி மற்றும் பிற சந்தர்ப்பங்களில் மிகச் சிறந்த மதிப்பீடுகளுடன் உள்ளது. +Phi தற்போது மிகவும் சக்திவாய்ந்த மற்றும் செலவு குறைந்த சிறிய மொழி மாதிரி (SLM) ஆகும், பன்மொழி, காரணிகை, உரை/செய்தி உருவாக்கம், குறியீடு, படங்கள், ஒலி மற்றும் பிற சூழல்களில் சிறந்த படிவங்கேற்றங்களை கொண்டுள்ளது. -நீங்கள் Phi-ஐ மேகத்திலும் விளிம்புப் சாதனங்களிலும் (edge devices) அமர்த்தலாம், மேலும் குறைந்த கணினி சக்தியுடன் விளைவு AI பயன்பாடுகளை எளிதாக உருவாக்க முடியும். +நீங்கள் Phi ஐ மேகத்திலோ அல்லது எட்ஜ் சாதனங்களிலும் பயன்படுத்தலாம், மற்றும் குறைந்த கணினிசார் திறனுடன் உருவாக்கும் AI பயன்பாடுகளை எளிதில் கட்டமைக்கலாம். -இந்த வளங்களை பயன்படுத்த தொடங்க இந்த படிகளை பின்பற்றவும் : -1. **Repository-ஐ Fork செய்யவும்**: Click [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Repository-ஐ Clone செய்யவும்**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Microsoft AI Discord சமூகத்தில் சேர்ந்து நிபுணர்களையும் மற்ற அபிவிருத்தியாளர்களையும் சந்திக்கவும்**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +இந்த வளங்களைப் பயன்படுத்த ஆரம்பிக்க பின்வரும் படிகளை பின்பற்றுங்கள்: +1. **களஞ்சியத்தை Fork செய்யவும்**: கிளிக் செய்யவும் [![GitHub பங்கை பிரிக்கைகள்](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **களஞ்சியத்தை பிரதியெடுக்கவும்**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**Microsoft AI Discord சமுதாயத்தில் சேரவும், நிபுணர்களும் மற்ற டெவலப்பர்களும் சந்திக்கவும்**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/ta/cover.eb18d1b9605d754b.webp) ### 🌐 பன்மொழி ஆதரவு -#### GitHub Action மூலம் ஆதரவானது (தானியங்கி & எப்போதும் புதுப்பிக்கப்பட்டது) +#### GitHub செயல்பாட்டால் ஆதரிக்கப்படுகிறது (தானாகவும் எப்போதும் புதியதாகவும்) [Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](./README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **உள்ளேகலனாகக் கிளோன் செய்ய விரும்புகிறீர்களா?** +> **உள்ளூர் க்ளோனிங்கை முன்னுரிமை தருகிறீர்களா?** > -> இந்த மறுசீரமைப்பில் 50+ மொழி மொழிபெயர்ப்புகள் உள்ளதால் பதிவிறக்கும் அளவு அதிகமாகிறது. மொழிபெயர்ப்புகள் இல்லாமல் கிளோன் செய்ய sparse checkout-ஐ பயன்படுத்தவும்: +> இந்த களஞ்சியம் 50+ மொழி மொழிபெயர்ப்புகளை கொண்டுள்ளது, இது பதிவிறக்கம் அளவை பெரிதாக்குகிறது. மொழிபெயர்ப்புகள் இல்லாமல் க்ளோன் செய்ய, ஸ்பார்ஸ் சேகரிப்பு பயன்படுத்தவும்: > > **Bash / macOS / Linux:** > ```bash @@ -52,272 +52,278 @@ Phi தற்போது மிகவும் சக்திவாய்ந > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> இது பாடத்திட்டத்தை முடிக்க தேவையான அனைத்தையும் மிகவும் விரைவான பதிவிறக்கத்துடன் தரும். +> இது பாடத்திட்டத்தை முடிக்க தேவையான அனைத்தையும் மிகவும் விரைவான பதிவிறக்கம் மூலம் வழங்கும். ## உள்ளடக்க அட்டவணை - அறிமுகம் - - [Phi குடும்பத்திற்கு வரவேற்கிறோம்](./md/01.Introduction/01/01.PhiFamily.md) - - [உங்கள் சுற்றுப்புறத்தை அமைத்தல்](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [முக்கிய தொழில்நுட்பங்களை புரிந்து கொள்வது](./md/01.Introduction/01/01.Understandingtech.md) + - [Phi குடும்பத்திற்கு வரவேற்பு](./md/01.Introduction/01/01.PhiFamily.md) + - [உங்கள் சூழலை அமைக்கல்](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [முக்கிய தொழில்நுட்பங்களைப் புரிந்து கொள்வது](./md/01.Introduction/01/01.Understandingtech.md) - [Phi மாதிரிகளுக்கான AI பாதுகாப்பு](./md/01.Introduction/01/01.AISafety.md) - - [Phi உடன் பொருந்தும் வன்பொருள் ஆதரவு](./md/01.Introduction/01/01.Hardwaresupport.md) - - [பல்பிளாட்ஃபாரங்களின் Phi மாதிரிகள் மற்றும் கிடைக்கும் நிலை](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Phi ஹார்ட்வேர் ஆதரவு](./md/01.Introduction/01/01.Hardwaresupport.md) + - [பிளாட்ஃபாரங்கள் முழுவதும் Phi மாதிரிகள் மற்றும் கிடைக்கும்தல்கள்](./md/01.Introduction/01/01.Edgeandcloud.md) - [Guidance-ai மற்றும் Phi பயன்படுத்துதல்](./md/01.Introduction/01/01.Guidance.md) - - [GitHub Marketplace மாதிரிகள்](https://github.com/marketplace/models) - - [Azure AI மாதிரி கண்டறிதல்](https://ai.azure.com) + - [GitHub மார்க்கெட்ட்பிளேஸ் மாதிரிகள்](https://github.com/marketplace/models) + - [Azure AI மாதிரி பட்டியல்](https://ai.azure.com) -- வெவ்வேறு சூழல்களில் Phi ஊக்குவிப்பு +- வெவ்வேறு சூழலில் Phi வடிதல் - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub மாதிரிகள்](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry Model Catalog](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Microsoft Foundry மாதிரி பட்டியல்](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [AI கருவி VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - -- Phi குடும்ப ஊக்குவிப்பு - - [iOS இல் Phi ஊக்குவிப்பு](./md/01.Introduction/03/iOS_Inference.md) - - [Android இல் Phi ஊக்குவிப்பு](./md/01.Introduction/03/Android_Inference.md) - - [Jetson இல் Phi ஊக்குவிப்பு](./md/01.Introduction/03/Jetson_Inference.md) - - [AI PC இல் Phi ஊக்குவிப்பு](./md/01.Introduction/03/AIPC_Inference.md) - - [Apple MLX Framework உடன் Phi ஊக்குவிப்பு](./md/01.Introduction/03/MLX_Inference.md) - - [இலங்கை (Local) சர்வரில் Phi ஊக்குவிப்பு](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI Toolkit பயன்படுத்தி தொலைச்சர்வரில் Phi ஊக்குவிப்பு](./md/01.Introduction/03/Remote_Interence.md) - - [Rust உடன் Phi ஊக்குவிப்பு](./md/01.Introduction/03/Rust_Inference.md) - - [இலங்கை (Local) பகுப்பாய்வுடன் Phi - காட்சி](./md/01.Introduction/03/Vision_Inference.md) - - [Kaito AKS, Azure Containers (அதிகாரப்பூர்வ ஆதரவு) உடன் Phi ஊக்குவிப்பு](./md/01.Introduction/03/Kaito_Inference.md) -- [Phi குடும்பம் குறிவைக்கும் அளவீடு](./md/01.Introduction/04/QuantifyingPhi.md) - - [llama.cpp பயன்படுத்தி Phi-3.5 / 4 குறிவைக்கும் படி](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime க்கான Generative AI நீட்டிப்புகள் பயன்படுத்தி Phi-3.5 / 4 குறிவைக்கும் படி](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Intel OpenVINO பயன்படுத்தி Phi-3.5 / 4 குறிவைக்கும் படி](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Apple MLX Framework பயன்படுத்தி Phi-3.5 / 4 குறிவைக்கும் படி](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Foundry உள்ளூர்](./md/01.Introduction/02/07.FoundryLocal.md) + +- Phi குடும்பத்தின் வடிதல் + - [iOS இல் Phi வடிதல்](./md/01.Introduction/03/iOS_Inference.md) + - [Android இல் Phi வடிதல்](./md/01.Introduction/03/Android_Inference.md) + - [Jetson இல் Phi வடிதல்](./md/01.Introduction/03/Jetson_Inference.md) + - [AI PC இல் Phi வடிதல்](./md/01.Introduction/03/AIPC_Inference.md) + - [Apple MLX கட்டமைப்புடன் Phi வடிதல்](./md/01.Introduction/03/MLX_Inference.md) + - [உள்ளூர் சேவையகத்தில் Phi வடிதல்](./md/01.Introduction/03/Local_Server_Inference.md) + - [AI கருவியைப் பயன்படுத்து ரிமோட் சேவையகத்தில் Phi வடிதல்](./md/01.Introduction/03/Remote_Interence.md) + - [Rust உடன் Phi வடிதல்](./md/01.Introduction/03/Rust_Inference.md) + - [உள்ளூரில் Phi--Vision வடிதல்](./md/01.Introduction/03/Vision_Inference.md) + - [Kaito AKS, Azure கொண்டெய்னர்கள்(அதிகாரபூர்வ ஆதரவு) உடன் Phi வடிதல்](./md/01.Introduction/03/Kaito_Inference.md) +- [Phi குடும்பத்தை அளவிடுதல்](./md/01.Introduction/04/QuantifyingPhi.md) + - [llama.cpp பயன்படுத்தி Phi-3.5 /4 அளவிடுதல்](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntimeக்கான உருவாக்கும் AI நீட்சிகளைப் பயன்படுத்தி Phi-3.5 /4 அளவிடுதல்](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Intel OpenVINO பயன்படுத்தி Phi-3.5 /4 அளவிடுதல்](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Apple MLX கட்டமைப்பைப் பயன்படுத்தி Phi-3.5 /4 அளவிடுதல்](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Phi மதிப்பீடு - - [பொறுப்பான AI](./md/01.Introduction/05/ResponsibleAI.md) - - [Microsoft Foundry மதிப்பீட்டுக்கு](./md/01.Introduction/05/AIFoundry.md) - - [Promptflow பயன்படுத்தி மதிப்பீடு](./md/01.Introduction/05/Promptflow.md) + - [பொறுப்பான AI](./md/01.Introduction/05/ResponsibleAI.md) + - [மதிப்பீட்டிற்கு Microsoft Foundry](./md/01.Introduction/05/AIFoundry.md) + - [மதிப்பீட்டிற்கு Promptflow பயன்படுத்துதல்](./md/01.Introduction/05/Promptflow.md) -- Azure AI தேடல் உடன் RAG - - [Phi-4-mini மற்றும் Phi-4-multimodal(RAG) ஐ Azure AI தேடலுடன் பயன்படுத்துவது எப்படி](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) - -- Phi பயன்பாடு அபிவிருத்தி மாதிரிகள் - - உரை மற்றும் chat பயன்பாடுகள் - - Phi-4 மாதிரிகள் - - [📓] [Phi-4-mini ONNX மாடலுடன் உரையாடல்](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 உள்ளூர் ONNX மாடல் .NET உடன் உரையாடல்](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Semantic Kernel பயன்படுத்தி Phi-4 ONNX உடன் .NET Console செயலி உரையாடல்](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 மாதிரிகள் - - [Phi3, ONNX Runtime Web மற்றும் WebGPU பயன்படுத்தி உலாவியில் உள்ளூர் chatbot](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [பல மாடல் - இடையறா Phi-3-மினி மற்றும் OpenAI விச்‍பர்](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - ஒரு அலம்பலை கட்டியமைத்து Phi-3 ஐ MLFlow உடன் பயன்படுத்துவது](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [மாடல் மேம்பாடு - ONNX Runtime வலைத்தளத்துக்கான Phi-3-மினி மாடலை Olive மூலம் எவ்வாறு மேம்படுத்துவது](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [Phi-3 மினி-4k-இன்பிரக்ட்ஒன்க்ஸ் உடன் WinUI3 செயலி](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 பல மாடல் AI சக்தி பெற்ற குறிப்பு செயலி மாதிரி](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [துல்லியமாக அமைய மற்றும் Prompt flow உடன் தனிப்பயன் Phi-3 மாடல்களை இணைத்தல்](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Microsoft Foundry-ல் Prompt flow உடன் தனிப்பயன் Phi-3 மாடல்களை துல்லியமாக அமைய மற்றும் இணைத்தல்](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Microsoft-ன் பொறுப்பான AI 원칙ங்களை சிந்திப்பதில் Microsoft Foundry இல் துல்லியமாக அமைய Phi-3 / Phi-3.5 மாடலை மதிப்பீடு செய்தல்](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-மினி-இன்பிரக்ட் மொழி முன்னறிவிப்பு மாதிரி (சீன/ஆங்கிலம்)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-இன்பிரக்ட் WebGPU RAG சாட்போட்](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU பயன்படுத்தி Phi-3.5-இன்பிரக்ட் ONNX உடன் Prompt flow தீர்வைப் படைப்பது](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Android செயலி உருவாக்க Microsoft Phi-3.5 tflite பயன்படுத்தல்](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET உதாரணம் உள்ளூர் ONNX Phi-3 மாடலை Microsoft.ML.OnnxRuntime பயன்படுத்தி](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Console சேட்டை .NET செயலி Semantic Kernel மற்றும் Phi-3 உடன்](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI பகுப்பாய்வு SDK குறியீடு அடிப்படையிலான மாதிரிகள் - - Phi-4 மாதிரிகள் - - [📓] [Phi-4-பல்சன்மையுடன் திட்ட குறியீட்டை உருவாக்குதல்](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 மாதிரிகள் - - [Microsoft Phi-3 குடும்பத்துடன் உங்கள் சொந்த Visual Studio Code GitHub Copilot சேட்டை கட்டுதல்](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Phi-3.5 மூலம் GitHub மாதிரிகளுடன் Visual Studio Code Chat Copilot முகவரியை தயார் செய்வது](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - மேம்பட்ட தர்க்க மாதிரிகள் - - Phi-4 மாதிரிகள் - - [📓] [Phi-4-மினி-தர்க்கம் அல்லது Phi-4-தர்க்கம் மாதிரிகள்](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Microsoft Olive மூலம் Phi-4-மினி-தர்க்கத்தை துல்லியமாக அமைய செய்தல்](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Apple MLX உடன் Phi-4-மினி-தர்க்கத்தை துல்லியமாக அமைய செய்தல்](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [GitHub மாதிரிகளுடன் Phi-4-மினி-தர்க்கம்](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Microsoft Foundry மாதிரிகளுடன் Phi-4-மினி-தர்க்கம்](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - மாதிரிகள் - - [Phi-4-மினி மாதிரிகள் Hugging Face Spacesல் இடம்பெற்றவை](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-பல்சன்மை மாதிரிகள் Hugginge Face Spacesல் இடம்பெற்றவை](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - காட்சி மாதிரிகள் - - Phi-4 மாதிரிகள் - - [📓] [Phi-4-பல்சன்மையைப் பயன்படுத்தி படங்களை வாசித்து குறியீட்டைக் உருவாக்குதல்](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 மாதிரிகள் - - [📓][Phi-3-காட்சி பட உரையை உரைக்கு மாற்றுதல்](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-காட்சி-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-காட்சி CLIP எம்பெட்டிங்](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [வெளியீடு: Phi-3 மறுசுழற்சி](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-காட்சி - காட்சி மொழி உதவியாளர் - Phi3-காட்சி மற்றும் OpenVINO உதவியுடன்](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 காட்சி Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 காட்சி OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 காட்சி பல-வடிவம் அல்லது பல-பட மாதிரி](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Microsoft.ML.OnnxRuntime .NET மூலம் உள்ளூர் ONNX மாடலான Phi-3 காட்சி](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Microsoft.ML.OnnxRuntime .NET மூலம் மெனு அடிப்படையிலான உள்ளூர் ONNX மாடலான Phi-3 காட்சி](../../md/04.HOL/dotnet/src/LabsPhi304) - - - தர்க்கம்-காட்சி மாதிரிகள் - - Phi-4-தர்க்கம்-காட்சி-15B - - [📓] [Phi-4-தர்க்கம்-காட்சி-15B பயன்படுத்தி ஜெய்வாகிங்கை கண்டுபிடித்தல்](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Phi-4-தர்க்கம்-காட்சி-15B பயன்படுத்தி கணிதம்](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-தர்க்கம்-காட்சி-15B பயன்படுத்தி UIயை கண்டறிதல்](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) +- Azure AI தேடலுடன் RAG + - [Phi-4-mini மற்றும் Phi-4-multimodal(RAG) ஐ Azure AI தேடலுடன் எப்படி பயன்படுத்துவது](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud உள்ளூர் ஹைபிரிட் RAG SQLite FTS5 மற்றும் phi-4-mini உடன்](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) + +- Phi பயன்பாட்டு மேம்பாட்டுக் கூடுகள் + - உரை மற்றும் உரையாடல் பயன்பாடுகள் + - Phi-4 உதாரணங்கள் + - [📓] [Phi-4-mini ONNX மாதிரியுடன் உரையாடல்](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Phi-4 உள்ளூர் ONNX மாதிரியுடன் உரையாடல் .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [செயன்முறை குழாய் வைத்து Phi-4 ONNX உடன் .NET கன்சோல் பயன்பாடு](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - பை-3 / 3.5 மாதிரிகள் + - [Phi3, ONNX Runtime Web மற்றும் WebGPU பயன்படுத்தி உலாவியில் உள்ளூர் அரட்டைப் பொறி](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [OpenVino அரட்டை](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [பல மாதிரிகள் - தொடர்புடைய phi-3-mini மற்றும் OpenAI விசும்பு](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - ஒரு உறை கட்டி Phi-3 உடன் MLFlow பயன்படுத்துவது](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [மாதிரி மேம்படுத்தல் - Olive உடன் ONNX Runtime Webக்கு Phi-3-min மாதிரியை எப்படி மேம்படுத்துவது](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Phi-3 mini-4k-instruct-onnx உடன் WinUI3 பயன்பாடுகள்](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 பல மாதிரி AI வழங்கிய குறிப்பு பயன்பாட்டின் மாதிரி](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [தனி Phi-3 மாதிரிகளை உரைப்பலகை ஓட்டத்தில் சிறப்பாக திருத்தி ஒருங்கிணைப்பது](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Microsoft Foundryல் உரைப்பலகை ஓட்டத்துடன் தனியார் Phi-3 மாதிரிகளை சிறப்பாக திருத்தி ஒருங்கிணைப்பது](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Microsoft Foundryல் Microsoftன் பொறுப்பான AI கொள்கைகளுக்கு முக்கியத்துவம் கொடுத்து சிறப்பாக திருத்தப்பட்ட Phi-3 / Phi-3.5 மாதிரியை மதிப்பாய்வு செய்வது](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct மொழிப் பூர்வீக மாதிரி (சீனா/ஆங்கிலம்)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG அரட்டைப் பொறி](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Phi-3.5-Instruct ONNX உடன் Windows GPU பயன்படுத்தி உரைப்பலகை ஓட்டத் தீர்வு உருவாக்குவது](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Microsoft Phi-3.5 tflite பயன்படுத்தி Android பயன்பாட்டை உருவாக்குவது](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Microsoft.ML.OnnxRuntime பயன்படுத்தி உள்ளூர் ONNX Phi-3 மாதிரியை பயன்படுத்தும் கேள்வி பதில் .NET உதாரணம்](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Semantic Kernel மற்றும் Phi-3 உடன் கான்சோல் அரட்டை .NET பயன்பாடு](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI கணிப்பொறி SDK குறியீடு அடிப்படையிலான மாதிரிகள் + - பை-4 மாதிரிகள் + - [📓] [Phi-4-பலவகை பயன்படுத்தி திட்ட குறியீட்டை உருவாக்குதல்](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - பை-3 / 3.5 மாதிரிகள் + - [Microsoft Phi-3 குடும்பத்துடன் உங்கள் சொந்த Visual Studio Code GitHub Copilot அரட்டை உருவாக்குதல்](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub மாதிரிகளுடன் Phi-3.5 பயன்படுத்தி உங்கள் சொந்த Visual Studio Code அரட்டை Copilot முகவரியை உருவாக்குதல்](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - முன்னேறிய காரணமுடைய மாதிரிகள் + - பை-4 மாதிரிகள் + - [📓] [Phi-4-mini-காரணம்றுக்கும் அல்லது Phi-4-காரணமுடைய மாதிரிகள்](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Microsoft Olive உடன் Phi-4-mini-காரணமுடையதை சிறப்பாக திருத்தல்](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLX உடன் Phi-4-mini-காரணமுடையதை சிறப்பாக திருத்தல்](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [GitHub மாதிரிகளுடன் Phi-4-mini-காரணமுடையது](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Microsoft Foundry மாதிரிகளுடன் Phi-4-mini-காரணமுடையது](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - டெமோக்கள் + - [Hugging Face Spaces இல் Phi-4-mini டெமோக்கள்](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Hugginge Face Spaces இல் Phi-4-பலவகை டெமோக்கள்](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - பார்வை மாதிரிகள் + - பை-4 மாதிரிகள் + - [📓] [படங்களை வாசித்து குறியீடு உருவாக்க Phi-4-பலவகையைப் பயன்படுத்துவது](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - பை-3 / 3.5 மாதிரிகள் + - [📓][Phi-3-பார்வை-பட வாக்கியங்களை உரைக்கு மாற்றுதல்](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-பார்வை-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-பார்வை CLIP இடைநிலை](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [டெமோ: Phi-3 மறுசுழற்சி](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-பார்வை - Visual மொழி உதவியாளர் - Phi3-பார்வை மற்றும் OpenVINO உடன்](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 பார்வை Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 பார்வை OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 பார்வை பல கட்டம் அல்லது பலபட மாதிரி](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Microsoft.ML.OnnxRuntime .NET பயன்படுத்தி Phi-3 பார்வை உள்ளூர் ONNX மாதிரி](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Microsoft.ML.OnnxRuntime .NET பயன்படுத்தி மெனு அடிப்படையிலான Phi-3 பார்வை உள்ளூர் ONNX மாதிரி](../../md/04.HOL/dotnet/src/LabsPhi304) + + - காரணம்-பார்வை மாதிரிகள் + - Phi-4-காரணம்-பார்வை-15B + - [📓] [Jaywalking கண்டறிய Phi-4-காரணம்-பார்வை-15B பயன்படுத்துவது](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [கணிதம் செய்ய Phi-4-காரணம்-பார்வை-15B பயன்படுத்துவது](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [UI கண்டறிய Phi-4-காரணம்-பார்வை-15B பயன்படுத்துவது](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - கணித மாதிரிகள் - - Phi-4-மினி-உயிரின்மிகு-இன்ஸ்ட்ரக்ட் மாதிரிகள் [Phi-4-மினி-உயிரின்மிகு-இன்ஸ்ட்ரக்ட் உடன் கணித வெளிப்பாடு](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-மினி-பிளாஷ்-காரணம்-உபதேச மாதிரிகள் [Phi-4-மினி-பிளாஷ்-காரணம்-உபதேசத்துடன் கணித டெமோ](./md/02.Application/09.Math/MathDemo.ipynb) - - ஆடியோ மாதிரிகள் - - Phi-4 மாதிரிகள் - - [📓] [Phi-4-பல்சன்மை மூலம் ஆடியோ உரைகளை எடுத்தல்](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-பல்சன்மை ஆடியோ மாதிரி](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-பல்சன்மை பேச்சு மொழியாக்க மாதிரி](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET கன்சோல் பயன்பாடு Phi-4-பல்சன்மை ஆடியோ மூலம் ஆடியோ கோப்பை பகுப்பாய்வு செய்து உரையை உருவாக்குதல்](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - ஒலி மாதிரிகள் + - பை-4 மாதிரிகள் + - [📓] [Phi-4-பலவகை பயன்படுத்தி ஒலி உரைகள் எடுப்பது](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-பலவகை ஒலி மாதிரி](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-பலவகை பேச்சு மொழிபெயர்ப்பு மாதிரி](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET கான்சோல் பயன்பாடு Phi-4-பலவகை ஒலியை பயன்படுத்தி ஒலி கோப்பை பகுப்பாய்வு செய்து உரையை உருவாக்குதல்](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE மாதிரிகள் - - Phi-3 / 3.5 மாதிரிகள் - - [📓] [Phi-3.5 நிபுணர்கள் கலவை (MoEs) சமூக ஊடகம் மாதிரி](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search மற்றும் LlamaIndex உடன் ரிட்ரீவல்-வலுப்படுத்தப்பட்ட உருவாக்கம் (RAG) குழாய் உருவாக்கல்](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - பை-3 / 3.5 மாதிரிகள் + - [📓] [Phi-3.5 நிபுணர் கலவை மாதிரிகள் (MoEs) சமூக ஊடக மாதிரி](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI தேடல் மற்றும் LlamaIndex உடன் மீட்டெடுக்கப்பட்ட உருவாக்கம் (RAG) குழாயைக் கட்டுதல்](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - செயல்பாடு அழைப்புப் மாதிரிகள் - - Phi-4 மாதிரிகள் 🆕 - - [📓] [Phi-4-மினியுடன் செயல்பாடு அழைப்பைப் பயன்படுத்துதல்](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Phi-4-மினியுடன் பல முகவர்களை உருவாக்க செயல்பாடு அழைப்பைப் பயன்படுத்துதல்](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Ollama உடன் செயல்பாடு அழைப்பைப் பயன்படுத்துதல்](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [ONNX உடன் செயல்பாடு அழைப்பைப் பயன்படுத்துதல்](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - பல்சன்மை கலவைக் மாதிரிகள் - - Phi-4 மாதிரிகள் 🆕 - - [📓] [கடந்தகால செய்தியாளராக Phi-4-பல்சன்மையைப் பயன்படுத்துதல்](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET கன்சோல் பயன்பாடு Phi-4-பல்சன்மையைப் பயன்படுத்தி படங்களை பகுப்பாய்வு செய்தல்](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Phi மாடல்களை துல்லியமாக அமைத்தல் - - [துல்லியமாக அமைக்கும் கண்காணிப்பு](./md/03.FineTuning/FineTuning_Scenarios.md) - - [துல்லியமாக அமைத்தல் மற்றும் RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Phi-3 ஐ தொழில்துறை நிபுணராக மாற்றுதல்](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [AI கருவி தொகுப்புடன் Phi-3 ஐ துல்லியமாக அமைத்தல்](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [ஆஸ்யூர் இயந்திரக் கற்றல் சேவையுடன் Phi-3 ஐ துல்லியமாக அமைத்தல்](./md/03.FineTuning/Introduce_AzureML.md) - - [Lora உடன் Phi-3 ஐ துல்லியமாக அமைத்தல்](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora உடன் Phi-3 ஐ துல்லியமாக அமைத்தல்](./md/03.FineTuning/FineTuning_Qlora.md) - - [Microsoft Foundry உடன் Phi-3 ஐ துல்லியமாக அமைத்தல்](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Azure ML CLI/SDK உடன் Phi-3 ஐ துல்லியமாக அமைத்தல்](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Microsoft Olive உடன் துல்லியமாக அமைத்தல்](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive கைவினை அறைத் திட்டத்துடன் துல்லியமாக அமைத்தல்](./md/03.FineTuning/olive-lab/readme.md) - - [Weights and Bias உடன் Phi-3-காட்சி துல்லியமாக அமைத்தல்](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Apple MLX கட்டமைப்புடன் Phi-3 ஐ துல்லியமாக அமைத்தல்](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-காட்சி (அதிகாரப்பூர்வ ஆதரவு) துல்லியமாக அமைத்தல்](./md/03.FineTuning/FineTuning_Vision.md) - - [கைட்டோ AKS உடன் Phi-3 நுட்பச் சீரமைப்பு, Azure கன்டெய்னர்ஸ் (அதிகாரப்பூர்வ ஆதரவு)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Phi-3 மற்றும் 3.5 விசனுக்கான நுட்பச் சீரமைப்பு](https://github.com/2U1/Phi3-Vision-Finetune) - -- அனுபவப்படிப்பு ஆய்வு - - [முன்னணி மாதிரிகளை ஆராய்தல்: LLMகள், SLMகள், உள்ளூர் அபிவிருத்தி மற்றும் மேலும்](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP திறன்களைத் திறக்குதல்: Microsoft Olive உடன் நுட்பச் சீரமைப்பு](https://github.com/azure/Ignite_FineTuning_workshop) - -- அகாடமிக் ஆய்வு கட்டுரைகள் மற்றும் வெளியீடுகள் - - [புத்தகங்கள் தான் உங்களுக்குத் தேவையானவை II: phi-1.5 தொழில்நுட்ப அறிக்கை](https://arxiv.org/abs/2309.05463) - - [Phi-3 தொழில்நுட்ப அறிக்கை: உங்கள் போனில் உள்ளூர் மிக திறமையான மொழி மாதிரி](https://arxiv.org/abs/2404.14219) + - செயல்பாடு அழைப்ப மாதிரிகள் + - பை-4 மாதிரிகள் 🆕 + - [📓] [Phi-4-மினி உடன் செயல்பாடு அழைப்பை பயன்படுத்துவது](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Phi-4-மினி மூலம் பல முகவர்கள் உருவாக்க செயல்பாடு அழைப்பைப் பயன்படுத்துவது](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Ollama உடன் செயல்பாடு அழைப்பைப் பயன்படுத்துவது](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [ONNX உடன் செயல்பாடு அழைப்பைப் பயன்படுத்துவது](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - பலவகை கலவை மாதிரிகள் + - பை-4 மாதிரிகள் 🆕 + - [📓] [தொழில்நுட்ப காப்புரை விளக்கப்படுத்தும் தொழில்நுட்ப எழுத்தாளர் ஆக Phi-4-பலவகையைப் பயன்படுத்துவது](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET கான்சோல் பயன்பாடு Phi-4-பலவகையைப் பயன்படுத்தி படங்களைக் கணக்கிடுதல்](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Phi மாதிரிகளை சிறப்பாக திருத்துதல் + - [சிறப்பாக திருத்தும் சூழல்கள்](./md/03.FineTuning/FineTuning_Scenarios.md) + - [சிறப்பாக திருத்தல் மற்றும் RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Phi-3 தொழில்துறை நிபுணராக மாறுவதை சிறப்பாக திருத்துதல்](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [VS Codeக்கான AI கருவிப் பொதியைப் பயன்படுத்தி Phi-3 ஐ சிறப்பாக திருத்துதல்](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure Machine Learning சேவையுடன் Phi-3 ஐ சிறப்பாக திருத்துதல்](./md/03.FineTuning/Introduce_AzureML.md) + - [Lora உடன் Phi-3 ஐ சிறப்பாக திருத்துதல்](./md/03.FineTuning/FineTuning_Lora.md) + - [QLora உடன் Phi-3 ஐ சிறப்பாக திருத்துதல்](./md/03.FineTuning/FineTuning_Qlora.md) + - [Microsoft Foundry உடன் Phi-3 ஐ சிறப்பாக திருத்துதல்](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Azure ML CLI/SDK உடன் Phi-3 ஐ சிறப்பாக திருத்துதல்](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Microsoft Olive உடன் சிறப்பாக திருத்துதல்](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Microsoft Olive செயல்முறை நடைமுறை ஆய்வு க்கான சிறப்பாக திருத்துதல்](./md/03.FineTuning/olive-lab/readme.md) + - [Weights and Bias உடன் Phi-3-பார்வையை சிறப்பாக திருத்துதல்](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [ஆப்பிள் MLX கட்டமைப்புடன் Phi-3 ஐ நுட்பமாக அமைத்தல்](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision ஐ நுட்பமாக அமைத்தல் (அதிகாரப்பூர்வ ஆதரவு)](./md/03.FineTuning/FineTuning_Vision.md) + - [Kaito AKS, Azure Containers உடன் Phi-3 ஐ நுட்பமாக அமைத்தல் (அதிகாரப்பூர்வ ஆதரவு)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Phi-3 மற்றும் 3.5 Vision ஐ நுட்பமாக அமையதிருத்தல்](https://github.com/2U1/Phi3-Vision-Finetune) + +- கைபிடிக்கும் ஆய்வகம் + - [அத்துண்டு மாடல்களை ஆராய்தல்: LLM கள், SLM கள், உள்ளக மேம்பாடு மற்றும் மேலும் பல](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP திறன்களை திறக்குதல்: Microsoft Olive உடன் நுட்பமாக அமைத்தல்](https://github.com/azure/Ignite_FineTuning_workshop) + +- அகடமிக் ஆராய்ச்சி கட்டுரைகள் மற்றும் வெளியீடுகள் + - [Textbooks Are All You Need II: phi-1.5 தொழில்நுட்ப அறிக்கை](https://arxiv.org/abs/2309.05463) + - [Phi-3 தொழில்நுட்ப அறிக்கை: உங்கள் கைபேசியில் உள்ள ஒரு திறமையான மொழி மாதிரியானது](https://arxiv.org/abs/2404.14219) - [Phi-4 தொழில்நுட்ப அறிக்கை](https://arxiv.org/abs/2412.08905) - - [Phi-4-மினி தொழில்நுட்ப அறிக்கை: Mixture-of-LoRAs மூலம் சுருக்கமான ஆனால் சக்திவாய்ந்த பன்முக மொழி மாதிரிகள்](https://arxiv.org/abs/2503.01743) - - [வாகனங்களில் உள்ள செயல்பாடு அழைக்கும் சிறிய மொழி மாதிரிகளை சிறப்படுத்தல்](https://arxiv.org/abs/2501.02342) - - [(ஏன்PHI) பன்முக தேர்வு கேள்வி பதில் தொடர்பான PHI-3 நுட்பச் சீரமைப்பு: முறைமைகள், முடிவுகள் மற்றும் சவால்கள்](https://arxiv.org/abs/2501.01588) - - [Phi-4 காரணமளிக்கும் தொழில்நுட்ப அறிக்கை](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-மினி காரணமளிக்கும் தொழில்நுட்ப அறிக்கை](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Phi-4-Mini தொழில்நுட்ப அறிக்கை: LoRAகளின் கலவையால் நெருக்கமான ஆனால் சக்திவாய்ந்த மல்டிமோடல் மொழி மாதிரிகள்](https://arxiv.org/abs/2503.01743) + - [காரில் செயல்படும் செயல்பாடுகளுக்கான சிறிய மொழி மாதிரிகளை பரிமாற்றுதல்](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) பல்துறை வினா பதில் அளிப்பதற்கான PHI-3 நுட்பமாக அமைத்தல்: முறைகள், முடிவுகள் மற்றும் சவால்கள்](https://arxiv.org/abs/2501.01588) + - [Phi-4-கருத்தாக்கும் தொழில்நுட்ப அறிக்கை](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4 மினி-கருத்தாக்கும் தொழில்நுட்ப அறிக்கை](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Phi மாதிரிகள் பயன்படுத்துதல் +## Phi மாதிரிகளை பயன்படுத்துதல் ### Microsoft Foundry இல் Phi -நீங்கள் Microsoft Phi யைப் பயன்படுத்துவது எப்படி, உங்கள் வேறுபட்ட ஹார்ட்வேர் சாதனங்களில் எ2எ தீர்வுகளை உருவாக்குவது எப்படி என்பதை அறியலாம். Phi என்பதன் அனுபவத்தை பெற, முதலில் மாதிரிகளுடன் விளையாடி, உங்கள் சூழலுக்கு ஏற்ப Phi ஐ தனிப்பயனாக்குங்கள் [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai)ஐப் பயன்படுத்தி மேலும் அறிய, [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) உடன் தொடங்கலாம் +நீங்கள் Microsoft Phi ஐ எப்படி பயன்படுத்துவது மற்றும் உங்கள் வித்தியாசமான ஹார்டுவேர் சாதனங்களில் முதலில் இறுதி தீர்வுகளை எப்படி கட்டுவது என்று கற்றுகொள்ளலாம். Phi ஐ நேரடியாக அனுபவிக்க, முதலில் மாதிரிகளை விளையாடி உங்கள் சூழ்நிலைகளுக்காக Phi ஐ தனிப்பயனாக்க [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) பயன்படுத்தலாம். மேலும் காண [Microsoft Foundry](../../../../../../../md/02.QuickStart/AzureAIFoundry_QuickStart.md ) உடன் தொடங்குங்கள். -**விளையாட்டு நிலை** -ஒவ்வொரு மாதிரிக்கும் தனித்துவமான விளையாட்டு நிலை உள்ளது மாதிரியை சோதிக்க [Azure AI Playground](https://aka.ms/try-phi3). +**விளையாட்டு மைதானம்** +ஒவ்வொரு மாதிரிக்கும் தனித்துவமான பரிசோதனை மைதானம் உள்ளது [Azure AI Playground](https://aka.ms/try-phi3). ### GitHub மாதிரிகளில் Phi -நீங்கள் Microsoft Phi யைப் பயன்படுத்துவது எப்படி, உங்கள் வேறுபட்ட ஹார்ட்வேர் சாதனங்களில் எ2எ தீர்வுகளை உருவாக்குவது எப்படி என்பதை அறியலாம். Phi அனுபவத்தைப் பெற, முதலில் மாதிரியுடன் விளையாடி, உங்கள் சூழல்களுக்கு ஏற்ப Phi ஐ தனிப்பயனாக்குங்கள் [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo)ஐப் பயன்படுத்தி மேலும் அறிய, [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) உடன் தொடங்கலாம் +Microsoft Phi ஐ எப்படி பயன்படுத்துவது மற்றும் உங்கள் வித்தியாசமான ஹார்டுவேரில் இறுதி தீர்வுகளை எப்படி கட்டுவது என்று அறியலாம். Phi யை நேரடியாக அனுபவிக்க, முதலில் மாதிரியை விளையாடி உங்கள் சூழ்நிலைகளுக்காக Phi ஐ தனிப்பயனாக்க [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) பயன்படுத்தலாம். மேலும் காண [GitHub Model Catalog](../../../../../../../md/02.QuickStart/GitHubModel_QuickStart.md ) உடன் தொடங்குங்கள். -**விளையாட்டு நிலை** -ஒவ்வொரு மாதிரிக்கும் தனித்துவமான [விளையாட்டு நிலை உள்ளது](/md/02.QuickStart/GitHubModel_QuickStart.md) மாதிரியை சோதிக்க. +**விளையாட்டு மைதானம்** +ஒவ்வொரு மாதிரிக்கும் [பரிசோதனைக்கு தனித்துவமான விளையாட்டு மைதானம்](/md/02.QuickStart/GitHubModel_QuickStart.md) உள்ளது. ### Hugging Face இல் Phi -நீங்கள் மாதிரியை [Hugging Face](https://huggingface.co/microsoft) இலும் கண்டறியலாம் +நீங்கள் மாதிரியை [Hugging Face](https://huggingface.co/microsoft) இல் காணலாம் -**விளையாட்டு நிலை** -[Hugging Chat விளையாட்டு நிலை](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**விளையாட்டு மைதானம்** + [Hugging Chat விளையாட்டு மைதானம்](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 மற்ற பாடங்கள் + ## 🎒 பிற பாடங்கள் -எங்கள் குழு பிற பாடத்திட்டங்களையும் தயாரிக்கிறது! பாருங்களேன்: +எங்கள் குழு பிற பாடங்களையும் தயாரிக்கிறது! பார்க்கலாம்: ### LangChain -[![LangChain4j உள்ளangs](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js உள்ளangs](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain உள்ளangs](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![புதியவர்களுக்கு LangChain4j](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![புதியவர்களுக்கு LangChain.js](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![புதியவர்களுக்கு LangChain](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / முனையிகள் -[![AZD உள்ளangs](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI உள்ளangs](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP உள்ளangs](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI முனையிகள் உள்ளangs](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / முகவர்கள் +[![புதியவர்களுக்கு AZD](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![புதியவர்களுக்கு Edge AI](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![புதியவர்களுக்கு MCP](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![புதியவர்களுக்கு AI முகவர்கள்](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### உருவாக்கும் AI தொடர் -[![உருவாக்கும் AI உள்ளangs](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![உருவாக்கும் AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![உருவாக்கும் AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![உருவாக்கும் AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### செயற்கை நுண்ணறிவு தொடர் +[![புதியவர்களுக்கு செயற்கை நுண்ணறிவு](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![செயற்கை நுண்ணறிவு (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![செயற்கை நுண்ணறிவு (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![செயற்கை நுண்ணறிவு (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### அடிப்படை கற்றல் -[![ML உள்ளangs](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![தரவு அறிவியல் உள்ளangs](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI உள்ளangs](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![சைபர் பாதுகாப்பு உள்ளangs](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![வலை மேம்பாடு உள்ளangs](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![இணைய பொருட்கள் உள்ளangs](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR மேம்பாடு உள்ளangs](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) + +### முதன்மை கற்றல் +[![புதியவர்களுக்கு இயந்திர கற்றல்](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![புதியவர்களுக்கு தரவு அறிவியல்](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![புதியவர்களுக்கு செயற்கை நுண்ணறிவு](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![புதியவர்களுக்கு சைபர் பாதுகாப்பு](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![புதியவர்களுக்கு வலை மேம்பாடு](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![புதியவர்களுக்கு IoT](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![புதியவர்களுக்கு XR மேம்பாடு](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### கோபைலட் தொடர் -[![AI இணையும் நிரல் கோபைலட்](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![C#/.NET கோபைலட்](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![கோபைலட் சாதனை](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![செயற்கை நுண்ணறிவு ஜோடி நிரலாக்கத்திற்கான கோபைலட்](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![C#/.NET இற்கான கோபைலட்](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![கோபைலட் கலப்பு](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## பொறுப்பான AI +## பொறுப்பான செயற்கை நுண்ணறிவு + +Microsoft நமக்கு AI தயாரிப்புகளை பொறுப்புடன் பயன்படுத்துவதற்கும், நமது கற்றல்களை பகிர்வதற்கும், Transparency Notes மற்றும் Impact Assessments போன்ற கருவிகளின் மூலம் நம்பிக்கையைக் கலந்த கூட்டமைப்புகளை உருவாக்குவதற்கும் உறுதிப்படையாக உள்ளது. இந்தப் பல வளங்களையும் [https://aka.ms/RAI](https://aka.ms/RAI) இல் காணலாம். +Microsoft இன் பொறுப்பான செயற்கை நுண்ணறிவு அணுகுமுறை நியாயம், நம்பகத்தன்மை மற்றும் பாதுகாப்பு, தனியுரிமை மற்றும் பாதுகாப்பு, கைகோர்த்தல், வெளிப்படைத்தன்மை மற்றும் பொறுப்புணர்வு என்பது போன்ற AI கோட்பாடுகளின் அடிப்படையில் உள்ளது. + +பெரிய அளவிலான இயற்பியல் மொழி, படங்கள் மற்றும் பேச்சு மாதிரிகள் - இந்த மாதிரிகளில் பயன்படுத்தப்படுகின்றன - அதுவே நியாயமற்ற, நம்பமுடியாத அல்லது அவமதிப்பித்தல் போன்றவைகள் நடக்கும் வாய்ப்பு உள்ளது, இது சேதங்களை உண்டாக்கலாம். ஆபத்து மற்றும் வரம்புகளைப் பற்றி அறிவதற்கு [Azure OpenAI சேவை Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) என்பதைப் படிக்கவும். -Microsoft நமது வாடிக்கையாளர்களுக்கு நமது AI தயாரிப்புகளை பொறுப்புடன் பயன்படுத்த உதவுகின்றது, நமது கற்றல்களை பகிர்ந்து, Transparency Notes மற்றும் Impact Assessments போன்ற கருவிகளின் மூலம் நம்பிக்கையூட்டும் கூட்டணிகளை உருவாக்குகின்றது. இந்த வளங்கள் அனைத்தும் [https://aka.ms/RAI](https://aka.ms/RAI) இல் காணப்படுகின்றன. -Microsoft இன் பொறுப்பான AI அணுகுமுறை நியாயம், நம்பகத்தன்மை மற்றும் பாதுகாப்பு, தனியுரிமை மற்றும் பாதுகாப்பு, உட்புகுத்தல், தெளிவுத்தன்மை மற்றும் பொறுப்பு ஆகிய AI கோட்பாடுகளின் அடிப்படையில் அமைந்துள்ளது. -பெரிய அளவிலான இயற்கை மொழி, படம், மற்றும் பேச்சு மாதிரிகள் - இந்த மாதிரிகள் உதாரணமாக பயன்படும் - எந்தவொரு முறையிலும் அநியாயமான, நம்பமுடியாத அல்லது உளவுத்தான் புள்ளிகளான தொல்லைகளை உண்டாக்கக்கூடிய நடத்தைகளை காணலாம். எச்சரிக்கை மற்றும் வரையறைகளை அறிய [Azure OpenAI சேவை Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ஐப் பாருங்கள். -இந்த ஆபத்துக்களை குறைக்க பரிந்துரைக்கப்படும் அணுகுமுறை உங்கள் கட்டமைப்பில் ஒரு பாதுகாப்பு அமைப்பைச் சேர்ப்பது, அது தீங்கு விளைவிக்கும் நடத்தை கண்டறிந்து தடுப்பதற்கு உதவும். [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) பயன்பாடுகள் மற்றும் சேவைகளில் பயனர் உருவாக்கிய மற்றும் AI உருவாக்கிய தீங்கு விளைவிக்கும் உள்ளடக்கத்தை கண்டறியக்கூடிய சுயாதீன பாதுகாப்பு அடுக்கு வழங்குகின்றது. Azure AI Content Safety ஆக்சிகளும் படங்களுக்குமான APIகளும் உள்ளன, இது தீங்கு விளைவிக்கும் பொருட்களை கண்டறிய உதவுகிறது. Microsoft Foundryக்குள், Content Safety சேவை ஆபத்து உள்ளடக்கத்தை வேறுபட்ட முறைகளில் கண்டறிய உதவும் மாதிரி குறியீடுகளை பார்வையிட, ஆராய்ந்து, முயற்சிக்க வாய்ப்பு அளிக்கிறது. கீழ்காணும் [துரித ஆரம்ப ஆவணம்](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) சேவையை அணுகுவதற்கான வழிகாட்டுதலாக உள்ளது. +இந்த அபாயங்களை குறைக்கும் பரிந்துரைக்கப்பட்ட அணுகுமுறை உங்கள் கட்டமைப்பில் தீங்கு விளைவிக்கும் நடத்தை கண்டறிந்து தடுக்கும் பாதுகாப்பு அமைப்பை சேர்ப்பது ஆகும். [Azure AI உள்ளடக்க பாதுகாப்பு](https://learn.microsoft.com/azure/ai-services/content-safety/overview) பயன்பாடுகள் மற்றும் சேவைகளில் தீங்கு விளைவிக்கும் பயனர்-உருவாக்கப்பட்ட மற்றும் AI-உருவாக்கப்பட்ட உள்ளடக்கத்தை கண்டறியக்கூடிய சுயாதீன பாதுகாப்பு அடுக்கு வழங்குகிறது. Azure AI உள்ளடக்க பாதுகாப்பில், தீங்கு விளைவிக்கும் பொருளை கண்டறிய நீங்கள் பயன்படுத்தக்கூடிய உரை மற்றும் பட உருவாக்க APIகள் அடக்கம் நுழைவது. Microsoft Foundry-இல், Content Safety சேவை தீங்கு விளைவிக்கும் உள்ளடக்கத்தை பல முறை, பல்வேறு வடிவங்களில் கண்டறிய மாதிரிச் குறியீட்டை பார்க்க, ஆராய மற்றும் முயற்சிக்க உதவுகிறது. பின்வரும் [குவிக்க்ஸ்டார்ட் ஆவணம்](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) சேவைக்கு கோரிக்கைகள் செய்ய வழிகாட்டுகிறது. -மற்றொரு முக்கிய அம்சம் உங்கள் செயலியின் மொத்த செயல்திறனை பரிசீலிப்பது ஆகும். பன்முக மற்றும் பன்மாடல் செயலிகளில், செயல்திறன் என்பது உங்கள் மற்றும் உங்கள் பயனர்கள் எதிர்பார்க்கும் விதமாக அமைப்பு செயல்பட வேண்டும், இதில் தீங்கு விளைவிக்கும் வெளியீடுகளை உருவாக்கக்கூடாது என்பதும் அடங்கும். உங்கள் மொத்த செயலியின் செயல்திறனை [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) மூலம் மதிப்பீடு செய்வது அவசியம். மேலும், உங்கள் தேவைக்கேற்ப [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) உருவாக்கி மதிப்பீடு செய்யும் திறனும் உங்களுக்கு உள்ளது. +மற்றொரு அம்சம் நினைவில் கொள்ள வேண்டியது என்பது மொத்த பயன்பாட்டு செயல்திறன் ஆகும். பன்முறை வடிவ மற்றும் பன்முறை மாதிரிகள் பயன்பாடுகளுடன், செயல்திறன் என்பது நீங்கள் மற்றும் உங்கள் பயனர்கள் எதிர்பார்க்கும் முறையில் அமைப்பு செயல்படுவதாக பொருள்படுத்தப்படுகிறது, தீங்கு விளைவிக்கும் வெளியீடுகளை உருவாக்காமலும். உங்கள் மொத்த பயன்பாட்டின் செயல்திறனை [செயல்திறன் மற்றும் தரம் மற்றும் அபாயம் மற்றும் பாதுகாப்பு மதிப்பீட்டாளர்கள்](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) பயன்படுத்தி மதிப்பிடுவது முக்கியம். [தனிப்பயன் மதிப்பீட்டாளர்களுடன்](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) நீங்கள் உருவாக்கி மதிப்பிடவும் முடியும். -[Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) பயன்படுத்தி உங்கள் AI செயலியை உங்கள் மேம்பாட்டு சூழலில் மதிப்பீடு செய்யலாம். சோதனை தரவுத்தொகை அல்லது இலக்கை கொடுத்தால், உங்கள் உருவாக்கும் AI செயலி உருவாக்கங்கள் கட்டமைக்கப்பட்ட மதிப்பீட்டாளர்களால் அல்லது உங்கள் விருப்பத்திற்கேற்ப உள்ள கட்டமைக்கப்பட்ட மதிப்பீட்டாளர்களால் எண்ணிக்கை அளவில் மதிப்பிடப்படுகின்றன. உங்கள் அமைப்பை மதிப்பிட azure ai evaluation sdk உடன் தொடங்க [துரித தொடக்கம் வழிகாட்டி](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) உள்ளது. மதிப்பீடு ஓட்டத்தை இயக்கியதும், [Microsoft Foundryயில் முடிவுகளை காண்பிக்கவும்](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) முடியும். +[Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) பயன்படுத்தி உங்கள் டெவலப்மெண்ட் சூழலில் உங்கள் AI பயன்பாட்டை மதிப்பிடலாம். ஒரு சோதனை தரவுத்தொகைக்கு அல்லது இலக்கிற்கு உங்கள் உருவாக்க AI பயன்பாட்டின் உருவாக்கங்கள் நேர்மையாக, கொண்டு வரும் மதிப்பீட்டாளர்களால் அல்லது உங்கள் தேர்ந்த தனிப்பயன் மதிப்பீட்டாளர்களால் கொள்கின்றன. உங்கள் அமைப்பை மதிப்பிட Azure AI Evaluation SDK-உடன் ஆரம்பிக்க [குவிக்க்ஸ்டார்ட் வழிகாட்டியை](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) பின்பற்றலாம். மதிப்பீட்டுச் சுற்று ஓட்டிய பிறகு, [Microsoft Foundry-இல் முடிவுகளை காணலாம்](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## வர்த்தகச் சொத்துகள் +## வர்த்தக குறியீடுகள் -இந்த திட்டத்தில் திட்டங்கள், தயாரிப்புகள் அல்லது சேவைகளுக்கான வர்த்தகச் சின்னங்கள் அல்லது லோகோக்கள் இருக்கலாம். Microsoft வர்த்தகச் சின்னங்கள் அல்லது லோகோக்களின் ஒப்புதலான பயன்பாடு [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) படி இருக்கும் மற்றும் பின்பற்ற வேண்டும். இந்த திட்டத்தின் மாற்றிய பதிப்புகளில் Microsoft வர்த்தகச் சின்னங்கள் அல்லது லோகோக்கள் பயன்படுத்தப்படும்போது குழப்பம் ஏற்படக்கூடாது அல்லது Microsoft ஆதரவாக இருக்கிறது என பொருள் கூற கூடாது. மற்றோர் தரப்பின் வர்த்தகச் சின்னங்கள் அல்லது லோகோக்கள் பயன்படுத்தும் போது அந்த தரப்பு கொள்கைகளுக்குட்பட்டது தான் பிராமாணிகம் ஆகும். +இந்த திட்டத்தில் திட்டங்கள், தயாரிப்புகள் அல்லது சேவைகளுக்கான வர்த்தக குறியீடுகள் அல்லது லோகோக்கள் இருக்கலாம். Microsoft வர்த்தக குறியீடுகள் அல்லது லோகோக்களை அதிகாரபூர்வமாக பயன்படுத்துவது [Microsoft வர்த்தக குறியீடு மற்றும் பிராண்ட் வழிகாட்டுதல்கள்](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) உடன் இணக்கம் கொண்டதாக இருக்க வேண்டும். +இந்த திட்டத்தின் மாற்றப்பட்ட பதிப்புகளில் Microsoft வர்த்தக குறியீடுகள் அல்லது லோகோக்களை பயன்படுத்துவது குழப்பத்தை ஏற்படுத்தக்கூடாது அல்லது Microsoft அங்கீகாரம் என்று அர்த்தமூட்டக்கூடாது. மூன்றாம் தரப்பினர் வர்த்தக குறியீடுகள் அல்லது லோகோக்களைப் பயன்படுத்துவது அந்த மூன்றாம் தரப்பினர் கொள்கைகளின் படி இருக்க வேண்டும். ## உதவி பெறுதல் -AI செயலிகளை உருவாக்குவதில் சிக்கல் ஏற்பட்டால் அல்லது சந்தேகம் இருந்தால் கீழ்காணும் ஹோட்டைபோட்டிலில் இணையுங்கள்: +AI பயன்பாடுகளை உருவாக்குவதில் சிக்கல் ஏற்பட்டால் அல்லது கேள்விகள் இருப்பின், சேருங்கள்: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -உற்பத்தியின் கருத்துக்கள் அல்லது பிழைகள் இருந்தால், கீழ்காணும் இடத்தை பார்வையிடுங்கள்: +தயாரிப்பு கருத்துமுடி அல்லது பிழைகள் இருந்தால், பார்வையிடவும்: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**குறிப்பு**: -இந்த ஆவணம் AI மொழிபெயர்ப்பு சேவை [Co-op Translator](https://github.com/Azure/co-op-translator) பயன்படுத்தி மொழிபெயர்க்கப்பட்டுள்ளது. நியாயமான துல்லியத்திற்காக முயற்சிப்பாலும், தானாக-generated மொழிபெயர்ப்புகளில் பிழைகள் அல்லது தவறுகள் இருக்க வாய்ப்பு உள்ளது என்பதைக் கவனத்தில் கொள்ளவும். அசல் ஆவணம் அதன் சொந்த மொழியில் அமைந்திருக்கும் ஆதாரமாக கருதப்படும். முக்கியமான தகவலுக்காக, தொழில்முறை மனித மொழிபெயர்ப்பு பரிந்துரைக்கப்படுகிறது. இந்த மொழிபெயர்ப்பின் பயன்படுத்துவதால் ஏற்படும் எந்த தவறான புரிதல்கள் அல்லது தவறான விளக்கங்களுக்கும் நாங்கள்உறுப்புத் தகுதி எடுப்பதில்லை. +**மறுப்பு**: +இந்த ஆவணம் AI மொழிபெயர்ப்பு சேவை [Co-op Translator](https://github.com/Azure/co-op-translator) பயன்படுத்தி மொழிபெயர்க்கப்பட்டுள்ளது. நாங்கள் துல்லியத்திற்காக முயற்சி செய்துள்ளோம், ஆனால் தானாக செய்யப்படும் மொழிபெயர்ப்புகளில் பிழைகள் அல்லது தவறுகள் இருக்கலாம் என்பதை கவனத்தில் கொள்ளவும். அசல் ஆவணம் அதன் தாய்மொழியில் அதிகாரப்பூர்வ ஆதாரமாக கருதப்பட வேண்டும். முக்கியமான தகவல்களுக்கு, தொழில்நுட்பமான மனித மொழிபெயர்ப்பு பரிந்துரைக்கப்படுகிறது. இந்த மொழிபெயர்ப்பைப் பயன்படுத்துவதால் ஏற்படும் எந்த தவறான புரிதல்கள் அல்லது தவறான விளக்கத்திற்கும் நாங்கள் பொறுப்பில்வில்லை. \ No newline at end of file diff --git a/translations/ta/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/ta/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..8e7635c84 --- /dev/null +++ b/translations/ta/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Microsoft phi-4-mini & SQLite FTS5 (சூனிய மேகம் SLM) உடன் ஹைபிரிட் RAG \n", + "\n", + "> **எழுத்தாளர்:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **முகப்பு:** சிறிய மொழி மாதிரிகள் (SLMs), SQLite FTS5 BM25, அடர்த்தியான எம்பெட்டிங்குகள், பரஸ்பர தரவரிசை கரம் (RRF) \n", + "\n", + "---\n", + "\n", + "## 📌 1. ஊக்கமருத்துவம்: உள்ளூர் SLMகளில் முக்கிய வார்த்தை நினைவுப் பிரச்சனை \n", + "முழுமையாக அடர்த்தியான வெக்டர் எம்பெட்டிங்குகளுக்கு மட்டுமே அ основыபடும் சாதாரண RAG கட்டமைப்புகள் சரியான எண்கணக்குக் குறியீடுகளை (எ.கா., `2,340,000 TL`, ஒப்பந்தக் குறியீடுகள், கணக்கு எண்கள்) மீட்டெடுக்க முறியடைகின்றன. \n", + "மாறாக, குறைக்கப்பட்ட சொற்றொடர் தேடல் (BM25) கருத்தாற்றலின் இணைப்புக்களையும் மறுபடி அமைந்த கேள்விகளையும் தவறவிட்டுக் கொள்கிறது. \n", + "\n", + "இந்த குக்க்புக் ஒரு **உயர்தர மற்றும் நினைவகத்தில் இயங்கும் ஹைபிரிட் மறுபடியாய்த் தேடல் இயந்திரம்** உருவாக்குவதற்கான முறையை காட்டுகிறது, இதில் சேர்க்கப்பட்டுள்ளது: \n", + "1. **அடர்த்தியான வெக்டர்கள்** (கோசைன் ஒத்திருப்பு) \n", + "2. **குறைவான சொற்களின் தேடல்** (SQLite FTS5 BM25) \n", + "3. **பரஸ்பர தரவரிசை கரம் (RRF, $k=60$)** \n", + "4. Microsoft `phi-4-mini` உடன் **அடிப்படையான மேற்கோள் உள்வாங்கல் (`[1]`, `[2]`)** \n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. இரட்டை SQLite திட்டம் (அடர்த்தியான வெக்டார்கள் + மெய்நிகர் FTS5 BM25 அட்டவணை)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. மாதிரி சேர்க்கை மற்றும் செயலாக்க ஒப்புமுறை\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. மைக்ரோசாஃப்ட் phi-4-mini க்கான நிலைத்த(prompt) முன்மொழிவு வடிவமைப்பு\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**மறுப்பு**:\nஇந்த ஆவணம் AI மொழிபெயர்ப்பு சேவை [Co-op Translator](https://github.com/Azure/co-op-translator) பயன்படுத்தி மொழிபெயர்க்கப்பட்டுள்ளது. நாங்கள் துல்லியத்திற்காக முயற்சி செய்துள்ளோம், ஆனால் தானாக செய்யப்படும் மொழிபெயர்ப்புகளில் பிழைகள் அல்லது தவறுகள் இருக்கலாம் என்பதை கவனத்தில் கொள்ளவும். அசல் ஆவணம் அதன் தாய்மொழியில் அதிகாரப்பூர்வ ஆதாரமாக கருதப்பட வேண்டும். முக்கியமான தகவல்களுக்கு, தொழில்நுட்பமான மனித மொழிபெயர்ப்பு பரிந்துரைக்கப்படுகிறது. இந்த மொழிபெயர்ப்பைப் பயன்படுத்துவதால் ஏற்படும் எந்த தவறான புரிதல்கள் அல்லது தவறான விளக்கத்திற்கும் நாங்கள் பொறுப்பில்வில்லை.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/te/.co-op-translator.json b/translations/te/.co-op-translator.json index c396b1d71..6c2cddc82 100644 --- a/translations/te/.co-op-translator.json +++ b/translations/te/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "te" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T08:03:17+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T16:56:52+00:00", "source_file": "README.md", "language_code": "te" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "te" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:04:21+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "te" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-12-21T15:49:48+00:00", diff --git a/translations/te/README.md b/translations/te/README.md index e1754b547..4c69d77d3 100644 --- a/translations/te/README.md +++ b/translations/te/README.md @@ -1,42 +1,42 @@ -# ఫై కుక్బుక్: మైక్రోసాఫ్ట్ యొక్క ఫై మోడల్స్‌తో హ్యాండ్స్-ఆన్ ఉదాహరణలు +# ఫై కుక్ బుక్: Microsoft's Phi మోడల్స్‌తో హ్యాండ్స్-ఆన్ ఉదాహరణలు -[![GitHub కోడ్స్‌పేస్‌ల‌లో నమూనాలను ఓపెన్ చేసి ఉపయోగించండి](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Dev కంటెయినర్ల‌లో ఓపెన్ చేయండి](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![GitHub Codespacesలో నమూనాలు తెరవండి మరియు ఉపయోగించండి](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Dev Containersలో తెరవండి](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub సహకారులు](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub కాంట్రిబ్యూటర్స్](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub సమస్యలు](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub పుల్-రిక్వెస్ట్లు](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub పుల్-రివెస్ట్‌లు](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![PRs స్వాగతం](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub వాచర్లు](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub ఫోర్క్లు](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub వాచర్స్](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub ఫోర్క్స్](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub స్టార్‌లు](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -ఫై అనేది మైక్రోసాఫ్ట్ అభివృద్ధి చేసిన ఓపెన్ సోర్స్ AI మోడల్స్ సిరీస్. +ఫై అనేది Microsoft ద్వారా అభివృద్ధి చేయబడిన ఓపెన్ సోర్స్ AI మోడల్స్ శ్రేణి. -ఫై ప్రస్తుతం అత్యంత శక్తివంతమైన మరియు ఖర్చు-ప్రభావవంతమైన చిన్న భాషా మోడల్ (SLM), బహుభాషా, తర్కం, టెక్స్ట్/చాట్ జనరేషన్, కోడింగ్, చిత్రాలు, ఆడియో మరియు ఇతర సందర్భాలలో చాలా మంచి బెంచ్‌మార్క్‌లతో ఉంది. +ఫై ప్రస్తుతం అత్యంత శక్తివంతమైన మరియు తక్కువ ఖర్చుతో కూడుకున్న చిన్న భాషా మోడల్ (SLM), ఇది బహుభాషా, కారణం, టెక్స్ట్/చాట్ జెనరేషన్, కోడింగ్, చిత్రం, ఆడియో మరియు ఇతర సన్నివేశాలలో అత్యుత్తమ బెంచ్‌మార్క్‌లను కలిగి ఉంది. -మీరు ఫైని క్లౌడ్‌లో లేదా ఎడ్జ్ పరికరాలలో డిప్లోయ్ చేయవచ్చు, మరియు పరిమిత కంప్యూటింగ్ శక్తితో తేలికగా జనరేటివ్ AI అనువర్తనాలను తయారుచేయవచ్చు. +మీరు ఫైని క్లౌడ్ లేదా ఎడ్జ్ పరికరాల్లో పయర్తించవచ్చు, మరియు మీరు పరిమిత కంప్యూటింగ్ శక్తితో సులభంగా జనరేటివ్ AI అనువర్తనాలను నిర్మించవచ్చు. -ఈ వనరులను ఉపయోగించడం ప్రారంభించడానికి ఈ దశలను అనుసరించండి : +ఈ వనరులను ఉపయోగించడం ప్రారంభించడానికి ఈ దశలను అనుసరించండి: 1. **రిపాజిటరీని ఫోర్క్ చేయండి**: క్లిక్ చేయండి [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **రిపాజిటరీని క్లోన్ చేయండి**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Microsoft AI Discord కమ్యూనిటీకి సైన్ అప్ అవ్వండి మరియు నిపుణులు మరియు ఇతర డెవలపర్లను కలుసుకోండి**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Microsoft AI Discord కమ్యూనిటీతో చేరండి మరియు నిపుణులు మరియు ఇతర డెవలపర్లను కలుసుకోండి**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/te/cover.eb18d1b9605d754b.webp) ### 🌐 బహుభాషా మద్దతు -#### GitHub యాక్షన్ ద్వారా మద్దతు (ఆటోమేటెడ్ & ఎప్పుడూ నవీకరించబడింది) +#### GitHub యాక్షన్ ద్వారా మద్దతు (ఆటోమేటెడ్ & ఎప్పుడూ తాజా) -[అరవబిక్](../ar/README.md) | [బెంగాలీ](../bn/README.md) | [బల్గేరియన్](../bg/README.md) | [బర్మీస్ (మయన్మార్)](../my/README.md) | [చైనీస్ (సింప్లిఫైడ్)](../zh-CN/README.md) | [చైనీస్ (పారంపరిక, హాంకాంగ్)](../zh-HK/README.md) | [చైనీస్ (పారంపరిక, మకావు)](../zh-MO/README.md) | [చైనీస్ (పారంపరిక, తైవాన్)](../zh-TW/README.md) | [క్రొయేటియన్](../hr/README.md) | [చెక్](../cs/README.md) | [డానిష్](../da/README.md) | [డచ్](../nl/README.md) | [ఎస్టోనియన్](../et/README.md) | [ఫినిష్](../fi/README.md) | [ఫ్రెంచ్](../fr/README.md) | [జర్మన్](../de/README.md) | [గ్రీకు](../el/README.md) | [హీబ్రూ](../he/README.md) | [హింది](../hi/README.md) | [హంగేరియన్](../hu/README.md) | [ఇండోనేషియన్](../id/README.md) | [ఇటాలియన్](../it/README.md) | [జపానీస్](../ja/README.md) | [కన్నడ](../kn/README.md) | [ఖ్మేర్](../km/README.md) | [కొరియన్](../ko/README.md) | [లిథువేనియన్](../lt/README.md) | [మలయ్](../ms/README.md) | [మలయాళం](../ml/README.md) | [మరాఠీ](../mr/README.md) | [నేపాలీ](../ne/README.md) | [నైజీరియన్ పిడ్జిన్](../pcm/README.md) | [నార్వేజియన్](../no/README.md) | [పర్షియన్ (ఫార్సీ)](../fa/README.md) | [పోలిష్](../pl/README.md) | [పోర్చుగీస్ (బ్రెజిల్)](../pt-BR/README.md) | [పోర్చుగీస్ (పోర్టుగల్)](../pt-PT/README.md) | [పంజాబీ (గుర్ముఖీ)](../pa/README.md) | [రోమేనియన్](../ro/README.md) | [రష్యన్](../ru/README.md) | [సెర్బియన్ (సిరిలిక్)](../sr/README.md) | [స్లోవాక్](../sk/README.md) | [స్లోవేనియన్](../sl/README.md) | [స్పానిష్](../es/README.md) | [స్వాహిలి](../sw/README.md) | [స్వీడిష్](../sv/README.md) | [టాగలాగ్ (ఫిలిపినా)](../tl/README.md) | [తమిళ్](../ta/README.md) | [తెలుగు](./README.md) | [థాయ్](../th/README.md) | [టర్కిష్](../tr/README.md) | [ఉక్రెయిన్](../uk/README.md) | [ఉర్దూ](../ur/README.md) | [వియత్నామీస్](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](./README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **స్థానికంగా క్లోన్ చేయడం ఇష్టమా?** +> **స్థానికంగా క్లోన్ చేయడం ఇష్టం?** > -> ఈ రిపాజిటరీ 50+ భాషా అనువాదాలను కలిగి ఉంది, ఇది డౌన్‌లోడ్ పరిమాణాన్ని గణనీయంగా పెంచుతుంది. అనువాదాలు లేకుండా క్లోన్ చేయడానికి, స్పార్స్ చెకౌట్‌ను ఉపయోగించండి: +> ఈ రిపాజిటరీలో 50+ భాషా అనువాదాలు ఉన్నాయి, ఇవి డౌన్‌లోడ్ పరిమాణాన్ని గణనీయంగా పెంచుతాయి. అనువాదాలు లేకుండా క్లోన్ చేయడానికి, స్పార్స్ చెకౌట్ ఉపయోగించండి: > > **Bash / macOS / Linux:** > ```bash @@ -52,30 +52,30 @@ > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> ఇది మీరు కోర్సును పూర్తి చేయడానికి అవసరమైన అన్ని వస్తువులను వేగంగా డౌన్‌లోడ్ చేసుకోవడానికి సహాయపడుతుంది. +> ఇది కోర్సును పూర్తి చేయడానికి అవసరమైన అన్ని విషయాలను చాలా త్వరగా డౌన్‌లోడ్‌ను అందిస్తుంది. -## అంశాల పట్టిక +## సూచిక - పరిచయం - [ఫై కుటుంబానికి స్వాగతం](./md/01.Introduction/01/01.PhiFamily.md) - - [మీ వాతావరణాన్ని సెటప్ చేసుకోవడం](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [ముఖ్య సాంకేతికతలను అర్థం చేసుకోవడం](./md/01.Introduction/01/01.Understandingtech.md) - - [ఫై మోడల్స్ కోసం AI సేఫ్టీ](./md/01.Introduction/01/01.AISafety.md) + - [మీ పరిసరాన్ని సెటప్ చేయడం](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [ముఖ్య సాంకేతికతలను అవగాహన చేసుకోవడం](./md/01.Introduction/01/01.Understandingtech.md) + - [ఫై మోడల్స్ కోసం AI భద్రత](./md/01.Introduction/01/01.AISafety.md) - [ఫై హార్డ్‌వేర్ మద్దతు](./md/01.Introduction/01/01.Hardwaresupport.md) - - [విభిన్న ప్లాట్‌ఫారమ్‌లలో ఫై మోడల్స్ & అందుబాటు](./md/01.Introduction/01/01.Edgeandcloud.md) + - [ఫై మోడల్స్ & వేదికలు మీద అందుబాటు](./md/01.Introduction/01/01.Edgeandcloud.md) - [Guidance-ai మరియు ఫై ఉపయోగించడం](./md/01.Introduction/01/01.Guidance.md) - - [GitHub మాకెట్‌ప్లేస్ మోడల్స్](https://github.com/marketplace/models) + - [GitHub మార్కెట్‌ప్లేస్ మోడల్స్](https://github.com/marketplace/models) - [Azure AI మోడల్ క్యాటలాగ్](https://ai.azure.com) -- వివిధ వాతావరణాల్లో ఫై ఇన్ఫరెన్స్ +- వివిధ పరిసరాల్లో ఫై ఇన్ఫరెన్స్ - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub మోడల్స్](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry Model క్యాటలాగ్](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Microsoft Foundry మోడల్ క్యాటలాగ్](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [AI టూల్కిట్ VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [AI టూల్‌కిట్ VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) + - [Foundry లోకల్](./md/01.Introduction/02/07.FoundryLocal.md) - ఫై కుటుంబం ఇన్ఫరెన్స్ - [iOSలో ఫై ఇన్ఫరెన్స్](./md/01.Introduction/03/iOS_Inference.md) @@ -83,242 +83,247 @@ - [Jetsonలో ఫై ఇన్ఫరెన్స్](./md/01.Introduction/03/Jetson_Inference.md) - [AI PCలో ఫై ఇన్ఫరెన్స్](./md/01.Introduction/03/AIPC_Inference.md) - [Apple MLX ఫ్రేమ్‌వర్క్‌తో ఫై ఇన్ఫరెన్స్](./md/01.Introduction/03/MLX_Inference.md) - - [స్థానిక సర్వర్‌లో ఫై ఇన్ఫరెన్స్](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI టూల్కిట్ ఉపయోగించి రిమోట్ సర్వర్‌లో ఫై ఇన్ఫరెన్స్](./md/01.Introduction/03/Remote_Interence.md) + - [లోకల్ సర్వర్‌లో ఫై ఇన్ఫరెన్స్](./md/01.Introduction/03/Local_Server_Inference.md) + - [AI టూల్‌కిట్ ఉపయోగించి రిమోట్ సర్వర్‌లో ఫై ఇన్ఫరెన్స్](./md/01.Introduction/03/Remote_Interence.md) - [రస్ట్‌తో ఫై ఇన్ఫరెన్స్](./md/01.Introduction/03/Rust_Inference.md) - - [స్థానికంగా ఫై -- విజన్ ఇన్ఫరెన్స్](./md/01.Introduction/03/Vision_Inference.md) - - [Kaito AKS, Azure కంటెయినర్ల(అధికారిక మద్దతు)తో ఫై ఇన్ఫరెన్స్](./md/01.Introduction/03/Kaito_Inference.md) -- [ఫై కుటుంబం పరిమాణం గణన](./md/01.Introduction/04/QuantifyingPhi.md) - - [llama.cpp ఉపయోగించి ఫై-3.5 / 4 పరిమాణాన్ని తగ్గించడం](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime కోసం Generative AI అనుబంధాలు ఉపయోగించి ఫై-3.5 / 4 పరిమాణం తగ్గించడం](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Intel OpenVINO ఉపయోగించి ఫై-3.5 / 4 పరిమాణం తగ్గించడం](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Apple MLX ఫ్రేమ్‌వర్క్ ఉపయోగించి ఫై-3.5 / 4 పరిమాణం తగ్గించడం](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [లోకల్‌లో ఫై-విజన్ ఇన్ఫరెన్స్](./md/01.Introduction/03/Vision_Inference.md) + - [Kaito AKS, Azure Containers(అధికారిక మద్దతు)తో ఫై ఇన్ఫరెన్స్](./md/01.Introduction/03/Kaito_Inference.md) +- [ఫై కుటుంబం కొలత](./md/01.Introduction/04/QuantifyingPhi.md) + - [llama.cpp ఉపయోగించి ఫై-3.5 / 4 కి కొలత వేయడం](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntime కోసం జనరేటివ్ AI విస్తరణలతో ఫై-3.5 / 4 కి కొలత వేయడం](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Intel OpenVINO ఉపయోగించి ఫై-3.5 / 4 కి కొలత వేయడం](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Apple MLX ఫ్రేమ్‌వర్క్ ఉపయోగించి ఫై-3.5 / 4 కి కొలత వేయడం](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - ఫై మూల్యాంకనం - - [సమాధానదాయక AI](./md/01.Introduction/05/ResponsibleAI.md) - - [మైక్రోసాఫ్ట్ Foundry ఫిర్యాదుల కోసం](./md/01.Introduction/05/AIFoundry.md) - - [మూల్యాంకన కోసం Promptflow ఉపయోగించడం](./md/01.Introduction/05/Promptflow.md) + - [స్పందన AI](./md/01.Introduction/05/ResponsibleAI.md) + - [మూల్యాంకనం కోసం Microsoft Foundry](./md/01.Introduction/05/AIFoundry.md) + - [మూల్యాంకనం కోసం Promptflow ఉపయోగించడం](./md/01.Introduction/05/Promptflow.md) -- Azure AI Search తో RAG - - [Azure AI Search తో Phi-4-mini మరియు Phi-4-multimodal(RAG) ఎలా ఉపయోగించాలి](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- Azure AI సెర్చ్‌తో RAG + - [Azure AI సెర్చ్‌తో Phi-4-mini మరియు Phi-4-мултимోడల్ (RAG) ఎలా ఉపయోగించాలి](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [SQLite FTS5 మరియు phi-4-mini తో జీరో-క్లౌడ్ లోకల్ హైబ్రిడ్ RAG](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - ఫై అప్లికేషన్ అభివృద్ధి నమూనాలు - టెక్స్ట్ & చాట్ అప్లికేషన్లు - - ఫై-4 నమూనాలు + - ఫై-4 నమూనాలు - [📓] [Phi-4-mini ONNX మోడల్‌తో చాట్](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [ఫై-4 స్థానిక ONNX మోడల్‌తో చాట్ .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [సెమెటిక్ కర్నెల్ ఉపయోగించి ఫై-4 ONNX తో చాట్ .NET కన్సోల్ యాప్](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - [Phi-4 లోకల్ ONNX మోడల్ .NETతో చాట్](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [సెమెంటిక్ కర్నల్ ఉపయోగించి Phi-4 ONNX తో .NET కంసోల్ యాప్‌లో చాట్](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - ఫై-3 / 3.5 నమూనాలు - - [ఫై3, ONNX రన్టైమ్ వెబ్ మరియు WebGPU ఉపయోగించి బ్రౌజర్‌లో స్థానిక చాట్‌బాట్](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino చాట్](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [బ్రౌజర్‌లో లోకల్ చాట్బోట్ వాడకం Phi3, ONNX రన్‌టైమ్ వెబ్ మరియు WebGPU ఉపయోగించి](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [ఓపెన్‌వినో చాట్](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [మల్టీ మోడల్ - ఇంటరాక్టివ్ Phi-3-mini మరియు OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - రాపర్ నిర్మాణం మరియు Phi-3 ను MLFlow తో ఉపయోగించడం](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [మోడల్ ఆప్టిమైజేషన్ - ONNX రంటైమ్ వెబ్ కోసం Phi-3-min మోడల్ ని Olive తో ఎలా ఆప్టిమైజ్ చేయాలి](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 యాప్ తో Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 మల్టీ మోడల్ AI పవర్డ్ నోట్స్ యాప్ నమూనా](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [కస్టమ్ Phi-3 మోడల్స్ ను ఫైన్-ట్యూన్ చేసి Prompt flow తో ఇంటిగ్రేట్ చేయడం](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Microsoft Foundry లో Prompt flow తో కస్టమ్ Phi-3 మోడల్స్ ని ఫైన్-ట్యూన్ చేసి ఇంటిగ్రేట్ చేయడం](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Microsoft Foundry లోని Microsoft యొక్క బాధ్యతాయుత AI సిద్ధాంతాలపైన దృష్టి పెట్టి ఫైన్-ట్యూన్ చేసిన Phi-3 / Phi-3.5 మోడల్ ని మూల్యాంకనం చేయడం](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct భాషా భావన నమూనా (చైనా/ఆంగ్లం)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG చాట్ బాట్](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU ఉపయోగించి Phi-3.5-Instruct ONNX తో Prompt flow పరిష్కారం సృష్టించడం](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Microsoft Phi-3.5 tflite ఉపయోగించి Android యాప్ సృష్టించడం](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET ఉదాహరణ స్థానిక ONNX Phi-3 మోడల్ Microsoft.ML.OnnxRuntime ఉపయోగించి](../../md/04.HOL/dotnet/src/LabsPhi301) - - [సెమెంటిక్ కర్నెల్ మరియు Phi-3 తో కಾನ్సోల్ చాట్ .NET యాప్](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI ఇన్ఫరెన్స్ SDK కోడ్ ఆధారిత నమూనాలు - - Phi-4 నమూనాలు - - [📓] [Phi-4-multimodal ఉపయోగించి ప్రాజెక్ట్ కోడ్ జనరేట్ చేయడం](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [మోడల్ ఆప్టిమైజేషన్ - ONNX రన్‌టైమ్ వెబ్ కోసం Phi-3-min మోడల్‌ను Olive తో ఎలా ఆప్టిమైజ్ చేయాలి](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 యాప్ Phi-3 mini-4k-instruct-onnx తో](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 మల్టీ మోడల్ ఏఐ శక్తివంత నోట్స్ యాప్ నమూనా](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [కస్టమ్ Phi-3 మోడల్స్‌ను Prompt flowతో ఫైన్-ట్యూన్ და ఇంటిగ్రేట్ చేయడం](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Microsoft Foundryలో Prompt flowతో కస్టమ్ Phi-3 మోడల్స్‌ను ఫైన్-ట్యూన్ и ఇంటిగ్రేట్ చేయడం](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Microsoft Foundryలో ఫైన్-ట్యూన్ చేసిన Phi-3 / Phi-3.5 మోడల్‌ను Microsoft's బాధ్యతాయుత AI సూత్రాలపై కేంద్రీకృతం చేస్తూ అంచనా వేయండి](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct భాషా పలుకుబడి నమూనా (చైనీస్/ఇంగ్లిష్)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG చాట్బోట్](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Windows GPU ఉపయోగించి Phi-3.5-Instruct ONNXతో Prompt flow పరిష్కారాన్ని సృష్టించడం](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Android యాప్ సృష్టించడానికి Microsoft Phi-3.5 tflite ఉపయోగించడం](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Q&A .NET ఉదాహరణ, Microsoft.ML.OnnxRuntime ఉపయోగించి లోకల్ ONNX Phi-3 మోడల్ ఉపయోగించి](../../md/04.HOL/dotnet/src/LabsPhi301) + - [సెమాంటిక్ కర్నల్ మరియు Phi-3తో కన్సోల్ చాట్ .NET యాప్](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI ఇన్ఫరెన్స్ SDK కోడ్ ఆధారిత నమూనాలు + - Phi-4 నమూనాలు + - [📓] [Phi-4-multimodal ఉపయోగించి ప్రాజెక్ట్ కోడ్ సృష్టించడం](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 నమూనాలు - - [Microsoft Phi-3 ఫ్యామిలీ తో మీ స్వంత విజువల్ స్టూడియో కోడ్ GitHub Copilot చాట్ నిర్మించండి](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub మోడల్స్ తో Phi-3.5 తో మీ స్వంత Visual Studio Code చాట్ Copilot ఏజెంట్ సృష్టించడం](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - అధునాతన తర్కం నమూనాలు - - Phi-4 నమూనాలు - - [📓] [Phi-4-mini-తర్కం లేదా Phi-4-తర్కం నమూనాలు](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Microsoft Olive తో Phi-4-mini-తర్కం ఫైన్-ట్యూనింగ్](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Apple MLX తో Phi-4-mini-తర్కం ఫైన్-ట్యూనింగ్](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [GitHub మోడల్స్ తో Phi-4-mini-తర్కం](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Microsoft Foundry మోడల్స్ తో Phi-4-mini-తర్కం](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [Microsoft Phi-3 కుటుంబంతో మీ స్వంత Visual Studio Code GitHub Copilot చాట్ నిర్మించండి](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub మోడల్స్‌తో Phi-3.5 ఉపయోగించి మీ స్వంత Visual Studio Code చాట్ కోపిలట్ ఏజెంట్ సృష్టించండి](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - అభివృద్ధి(reasoning) నమూనాలు + - Phi-4 నమూనాలు + - [📓] [Phi-4-mini-reasoning లేదా Phi-4-reasoning నమూనాలు](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Microsoft Olive తో Phi-4-mini-reasoning ఫైన్-ట్యూనింగ్](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLXతో Phi-4-mini-reasoning ఫైన్-ట్యూనింగ్](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [GitHub మోడల్స్ తో Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Microsoft Foundry మోడల్స్ తో Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - డెమోలు - - [Phi-4-mini డెమోలు Hugging Face Spacesలో హోస్ట్ చేయబడ్డాయి](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-మల్టిమోడల్ డెమోలు Hugginge Face Spacesలో హోస్ట్ చేయబడ్డాయి](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [Hugging Face Spacesలో హోస్ట్ చేసిన Phi-4-mini డెమోలు](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Hugginge Face Spacesలో హోస్ట్ చేసిన Phi-4-multimodal డెమోలు](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - విజన్ నమూనాలు - - Phi-4 నమూనాలు - - [📓] [Phi-4-multimodal ఉపయోగించి చిత్రాలు చదవడం మరియు కోడ్ జనరేట్ చేయడం](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-4 నమూనాలు + - [📓] [Phi-4-multimodal ఉపయోగించి చిత్రాలను చదవడం మరియు కోడ్ సృష్టించడం](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 నమూనాలు - - [📓][Phi-3-vision-చిత్రం టెక్స్ట్ నుండి టెక్స్ట్](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-చిత్రం టెక్స్ట్ నుంచి టెక్స్ట్](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP ఎంబెడ్డింగ్](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision CLIP ఇంబెడ్డింగ్](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [డెమో: Phi-3 రీసైక్లింగ్](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - విజువల్ భాష సహాయకుడు - Phi3-విజన్ మరియు OpenVINO తో](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3-vision - విజువల్ లాంగ్వేజ్ అసిస్టెంట్ - Phi3-విజన్ మరియు OpenVINOతో](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 విజన్ Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 విజన్ OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - [📓][Phi-3.5 విజన్ మల్టీ-ఫ్రేమ్ లేదా మల్టీ-ఇమేజ్ నమూనా](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Microsoft.ML.OnnxRuntime .NET ఉపయోగించి స్థానిక Phi-3 విజన్ ONNX మోడల్](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Microsoft.ML.OnnxRuntime .NET ఉపయోగించి మెనూ ఆధారిత Phi-3 విజన్ స్థానిక ONNX మోడల్](../../md/04.HOL/dotnet/src/LabsPhi304) + - [Microsoft.ML.OnnxRuntime .NET ఉపయోగించి Phi-3 విజన్ లోకల్ ONNX మోడల్](../../md/04.HOL/dotnet/src/LabsPhi303) + - [మెను ఆధారిత Phi-3 విజన్ లోకల్ ONNX మోడల్ Microsoft.ML.OnnxRuntime .NET ఉపయోగించి](../../md/04.HOL/dotnet/src/LabsPhi304) - - తర్కం-విజన్ నమూనాలు - - Phi-4-తర్కం-విజన్-15B - - [📓] [Phi-4-తర్కం-విజన్-15B ఉపయోగించి జేవాకింగ్ గుర్తించడం](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Phi-4-తర్కం-విజన్-15B ఉపయోగించి గణితం](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-తర్కం-విజన్-15B ఉపయోగించి UI గుర్తించడం](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - రీజనింగ్-విజన్ నమూనాలు + - Phi-4-రిజనింగ్-విజన్-15B + - [📓] [Phi-4-రిజనింగ్-విజన్-15B ఉపయోగించి జేవాకింగ్ గుర్తించడం](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-రిజనింగ్-విజన్-15B ఉపయోగించి గణితం](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Phi-4-రిజనింగ్-విజన్-15B ఉపయోగించి UI గుర్తించడం](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - గణితం నమూనాలు - - Phi-4-మినీ-ఫ్లాష్-తర్కం-ఇన్‌స్ట్రక్ట్ నమూనాలు [Phi-4-మినీ-ఫ్లాష్-తర్కం-ఇన్‌స్ట్రక్ట్ తో గణిత డెమో](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-మినీ-ఫ్లాష్-రిజనింగ్-ఇన్‌స్ట్రక్ట్ నమూనాలు [Phi-4-మినీ-ఫ్లాష్-రిజనింగ్-ఇన్‌స్ట్రక్ట్ తో గణితం డెమో](./md/02.Application/09.Math/MathDemo.ipynb) - ఆడియో నమూనాలు - - Phi-4 నమూనాలు - - [📓] [Phi-4-multimodal ఉపయోగించి ఆడియో ట్రాన్స్క్రిప్ట్ సేకరణ](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - Phi-4 నమూనాలు + - [📓] [Phi-4-multimodal ఉపయోగించి ఆడియో ట్రాన్స్క్రిప్ట్స్ తీసేసుకోవడం](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Phi-4-multimodal ఆడియో నమూనా](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Phi-4-multimodal స్పీచ్ అనువాద నమూనా](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET కాంసోల్ అప్లికేషన్ Phi-4-multimodal ఆడియోను ఉపయోగించి ఆడియో ఫైల్ విశ్లేషణ మరియు ట్రాన్స్క్రిప్ట్ సృష్టించడం](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [.NET కన్సోల్ అప్లికేషన్ Phi-4-multimodal ఆడియో ఉపయోగించి ఆడియో ఫైల్ విశ్లేషించి ట్రాన్స్క్రిప్ట్ రూపొందించడం](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE నమూనాలు - Phi-3 / 3.5 నమూనాలు - - [📓] [Phi-3.5 మిశ్రమ నిపుణుల మోడల్స్ (MoEs) సోషల్ మీడియా నమూనా](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI సెర్చ్ మరియు LlamaIndex తో రిట్రీవల్-ఆగ్మెంటెడ్ జనరేషన్ (RAG) పైప్‌లైన్ నిర్మాణం](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 మిక్చర్ ఆఫ్ ఎక్స్పర్ట్స్ మోడల్స్ (MoEs) సోషల్ మీడియా నమూనా](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI సెర్చ్, మరియు LlamaIndex తో రిట్రీవల్-ఆగ్మెంటెడ్ జనరేషన్ (RAG) పైప్‌లైన్ నిర్మించడం](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - ఫంక్షన్ కాలింగ్ నమూనాలు - Phi-4 నమూనాలు 🆕 - [📓] [Phi-4-mini తో ఫంక్షన్ కాలింగ్ ఉపయోగించడం](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Phi-4-mini తో మల్టీ-ఏజెంట్స్ సృష్టించడానికి ఫంక్షన్ కాలింగ్ ఉపయోగించడం](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Phi-4-mini తో బహుళ ఏజెంట్స్ సృష్టించడానికి ఫంక్షన్ కాలింగ్ ఉపయోగించడం](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Ollama తో ఫంక్షన్ కాలింగ్ ఉపయోగించడం](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [ONNX తో ఫంక్షన్ కాలింగ్ ఉపయోగించడం](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - మల్టిమోడల్ మిక్సింగ్ నమూనాలు + - మల్టీమోడల్ మిక్సింగ్ నమూనాలు - Phi-4 నమూనాలు 🆕 - - [📓] [టెక్నాలజీ జర్నలిస్ట్ గా Phi-4-multimodal ఉపయోగించడం](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET కాంసోల్ అప్లికేషన్ Phi-4-multimodal ఉపయోగించి చిత్రాల విశ్లేషణ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- ఫైన్-ట్యూనింగ్ Phi నమూనాలు - - [ఫైన్-ట్యూనింగ్ దృశ్యాలు](./md/03.FineTuning/FineTuning_Scenarios.md) - - [ఫైన్-ట్యూనింగ్ vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Phi-3 ను పరిశ్రమ నిపుణుడిగా మార్చేందుకు ఫైన్-ట్యూనింగ్](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [VS కోడ్ కోసం AI టూల్‌కిట్ తో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Azure మెషిన్ లర్నింగ్ సర్వీస్ తో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/Introduce_AzureML.md) - - [Lora తో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora తో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_Qlora.md) - - [Microsoft Foundry తో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Azure ML CLI/SDK తో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Microsoft Olive తో ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive Hands-On ల్యాబ్ తో ఫైన్-ట్యూనింగ్](./md/03.FineTuning/olive-lab/readme.md) - - [Weights and Bias తో Phi-3-vision ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Apple MLX ఫ్రేమ్‌వర్క్ తో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_MLX.md) + - [📓] [ఫై-4 మల్టీమోడల్ టెక్నాలజీ జర్నలిస్ట్ గా ఉపయోగించడం](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET కన్సోల్ అప్లికేషన్ Phi-4 మల్టీమోడల్ ఉపయోగించి చిత్రాలను విశ్లేషించడం](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- ఫైన్-ట్యూనింగ్ ఫై నమూనాలు + - [ఫైన్-ట్యూనింగ్ సన్నివేశాలు](./md/03.FineTuning/FineTuning_Scenarios.md) + - [ఫైన్-ట్యూనింగ్ వర్సెస్ RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [ఫైన్-ట్యూనింగ్ ద్వారా ఫై-3ని ఒక పరిశ్రమ నిపుణుడిగా మార్చడం](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [AI టూల్‌కిట్ ఉపయోగించి Phi-3 ఫైన్-ట్యూనింగ్ for VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure మషీన్ లెర్నింగ్ సర్వీస్ తో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/Introduce_AzureML.md) + - [Loraతో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_Lora.md) + - [QLoraతో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_Qlora.md) + - [Microsoft Foundryతో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Azure ML CLI/SDKతో Phi-3 ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Microsoft Oliveతో ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Microsoft Olive Hands-On Labతో ఫైన్-ట్యూనింగ్](./md/03.FineTuning/olive-lab/readme.md) + - [Weights and Bias తో Phi-3 విజన్ ఫైన్-ట్యూనింగ్](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Apple MLX ఫ్రేమ్‌వర్క్‌తో Phi-3 ను ఫైన్-ట్యూన్ చేయడం](./md/03.FineTuning/FineTuning_MLX.md) - [Phi-3-vision ఫైన్-ట్యూనింగ్ (అధికారిక మద్దతు)](./md/03.FineTuning/FineTuning_Vision.md) - - [KaiTo AKS , Azure కంటైనర్లతో Phi-3 ఫైన్-ట్యూనింగ్ (అధికారిక మద్దతు)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Phi-3 మరియు 3.5 విజన్ ఫైన్-ట్యూనింగ్](https://github.com/2U1/Phi3-Vision-Finetune) + - [Kaito AKS, Azure కంటైనర్లతో Phi-3 ను ఫైన్-ట్యూన్ చేయడం (అధికారిక మద్దతు)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Phi-3 మరియు 3.5 Vision ఫైన్-ట్యూనింగ్](https://github.com/2U1/Phi3-Vision-Finetune) -- ప్రాక్టికల్ ల్యాబ్ - - [కటింగ్-ఎడ్జ్ మోడల్స్ అన్వేషణ: LLMs, SLMs, స్థానిక అభివృద్ధి మరియు మరిన్ని](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP సామర్థ్యాన్ని విడుదల చేయడం: Microsoft Olive తో ఫైన్-ట్యూనింగ్](https://github.com/azure/Ignite_FineTuning_workshop) +- ప్రయోగాల శాల + - [అత్యాధునిక నమూనాలను అన్వేషించడం: LLMs, SLMs, స్థానిక అభివృద్ధి మరియు మరిన్ని](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP సామర్థ్యాన్ని తెరవడం: Microsoft Olive తో ఫైన్-ట్యూనింగ్](https://github.com/azure/Ignite_FineTuning_workshop) -- అకడమిక్ పరిశోధన పత్రాలు మరియు ప్రచురణలు - - [పాఠ్యపుస్తకాలు చేనే కావాలి II: phi-1.5 సాంకేతిక నివేదిక](https://arxiv.org/abs/2309.05463) - - [Phi-3 సాంకేతిక నివేదిక: మీ ఫోన్‌లో స్థానికంగా ఉన్న అత్యున్నత సామర్థ్య భాషా మోడల్](https://arxiv.org/abs/2404.14219) +- అకాడమిక్ పరిశోధనా పత్రాలు మరియు ప్రచురణలు + - [పాఠ్య పుస్తకాలు అన్ని కావాలి II: phi-1.5 సాంకేతిక నివేదిక](https://arxiv.org/abs/2309.05463) + - [Phi-3 సాంకేతిక నివేదిక: మీ ఫోన్‌లో స్థానికంగా అధిక సామర్థ్యమైన భాషా నమూనా](https://arxiv.org/abs/2404.14219) - [Phi-4 సాంకేతిక నివేదిక](https://arxiv.org/abs/2412.08905) - - [Phi-4-మినీ సాంకేతిక నివేదిక: మిశ్రమ-ఆఫ్-లోరాలతో కంపాక్ట్ అయిన శక్తివంతమైన బహుముఖ భాషా మోడల్స్](https://arxiv.org/abs/2503.01743) - - [వాహనంలో ఫంక్షన్-కాల్ కోసం చిన్న భాషా మోడల్స్ ఆప్టిమైజేషన్](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) బహుళ-ఛాయిస్ ప్రశ్నలకు జవాబుకు PHI-3 ఫైన్-ట్యూనింగ్: విధానం, ఫలితాలు, మరియు సవాళ్లు](https://arxiv.org/abs/2501.01588) - - [Phi-4-ఆలోచన సాంకేతిక నివేదిక](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-మినీ-ఆలోచన సాంకేతిక నివేదిక](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Phi-4-Mini సాంకేతిక నివేదిక: మిశ్రమ-LoRAs ద్వారా సన్నబడ్డ కానీ శక్తివంతమైన బహుళమోడ్ భాషా నమూనాలు](https://arxiv.org/abs/2503.01743) + - [వాహనంలో ఫంక్షన్-కాలింగ్ కోసం చిన్న భాషా నమూనాల оптимైజేషన్](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) ఒకाधिक-ఎంపిక ప్రశ్నల సమాధాన కోసం PHI-3 ఫైన్-ట్యూనింగ్: విధానం, ఫలితాలు, మరియు సవాళ్లు](https://arxiv.org/abs/2501.01588) + - [Phi-4-రిజనింగ్ సాంకేతిక నివేదిక](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-మినీ-రిజనింగ్ సాంకేతిక నివేదిక](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Phi మోడల్స్ ఉపయోగించడం +## Phi నమూనాలను ఉపయోగించడం -### మైక్రోసాఫ్ట్ ఫౌండ్రీలో Phi +### Microsoft Foundry లో Phi -మీ వివిధ హార్డ్‌వేర్ పరికరాల్లో Microsoft Phi ను ఎలా ఉపయోగించాలో, E2E పరిష్కారాలను ఎలా నిర్మించాలో మీరు తెలుసుకోవచ్చు. Phi ని ప్రత్యక్షంగా అనుభవించేందుకు, మొదట మోడల్స్‌తో ఆడండి మరియు మీ సందర్భాల కోసం Phi ను అనుకూలీకరించండి [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) ఉపయోగించి. మీరు [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) ద్వారా ప్రారంభం చేసుకోవచ్చు. +మీరు Microsoft Phi ఎలా ఉపయోగించాలో మరియు మీ వివిధ హార్డ్‌వేర్ పరికరాల్లో E2E పరిష్కారాలను ఎలా నిర్మించాలో నేర్చుకోవచ్చు. Phi ను ప్రత్యక్షంగా అనుభవించడానికి, నమూనాలతో ఆడడం మరియు మీ పరిస్తితుల కోసం Phi ను అనుకూలీకరించడం మొదలుపెట్టండి [Microsoft Foundry Azure AI మోడల్ క్యాటలాగ్](https://aka.ms/phi3-azure-ai) ఉపయోగించండి, మీరు [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) తో ప్రారంభించడంపై మరింత తెలుసుకోండి -**ప్లేగ్రౌండ్** -ప్రతి మోడల్‌కు మోడల్ పరీక్షించడానికి ప్రత్యేక ప్లేగ్రౌండ్ ఉంది [Azure AI Playground](https://aka.ms/try-phi3). +**ప్లేగ్రౌండ్** +ప్రతి నమూనాకు నమూనాని పరీక్షించడానికి ప్రత్యేకమైన ప్లేగ్రౌండ్ ఉంది [Azure AI Playground](https://aka.ms/try-phi3). -### GitHub మోడల్స్‌లో Phi +### GitHub Models లో Phi -మీ వివిధ హార్డ్‌వేర్ పరికరాల్లో Microsoft Phi ను ఎలా ఉపయోగించాలో, E2E పరిష్కారాలను ఎలా నిర్మించాలో మీరు తెలుసుకోవచ్చు. Phi ని ప్రత్యక్షంగా అనుభవించేందుకు, మొదట మోడల్‌తో ఆడండి మరియు మీ సందర్భాల కోసం Phi ను అనుకూలీకరించండి [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ఉపయోగించి. మీరు [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) ద్వారా ప్రారంభం చేసుకోవచ్చు. +మీరు Microsoft Phi ఎలా ఉపయోగించాలో మరియు మీ వివిధ హార్డ్‌వేర్ పరికరాల్లో E2E పరిష్కారాలను ఎలా నిర్మించాలో నేర్చుకోవచ్చు. Phi ను ప్రత్యక్షంగా అనుభవించడానికి, నమూనాతో ఆడడం మరియు మీ పరిస్తితుల కోసం Phi ను అనుకూలీకరించడం మొదలుపెట్టండి [GitHub Model క్యాటలాగ్](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ఉపయోగించండి, మీరు [GitHub Model క్యాటలాగ్](/md/02.QuickStart/GitHubModel_QuickStart.md) తో ప్రారంభించడంపై మరింత తెలుసుకోండి -**ప్లేగ్రౌండ్** -ప్రతి మోడల్‌కు ప్రత్యేక [ప్లేగ్రౌండ్ లో మోడల్ పరీక్షించడానికి](/md/02.QuickStart/GitHubModel_QuickStart.md) ఉంటుంది. +**ప్లేగ్రౌండ్** +ప్రతి నమూనాకు [ఉదాహరణ కోసం ఒక ప్లేగ్రౌండ్](/md/02.QuickStart/GitHubModel_QuickStart.md) ఉంది. ### Hugging Face లో Phi -మీకు మోడల్ [Hugging Face](https://huggingface.co/microsoft) లో కూడా లభిస్తుంది. +మీరు నమూనాను [Hugging Face](https://huggingface.co/microsoft) లో కూడా కనుగొనవచ్చు -**ప్లేగ్రౌండ్** -[Hugging Chat ప్లేగ్రౌండ్](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**ప్లేగ్రౌండ్** + [Hugging Chat ప్లేగ్రౌండ్](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 ఇతర కోర్సులు + ## 🎒 ఇతర కోర్సులు -మా బృందం ఇతర కోర్సులు కూడా తయారు చేస్తోంది! పరిశీలించండి: +మా బృందం ఇతర కోర్సులు ఉత్పత్తి చేస్తోంది! చూడండి: ### LangChain -[![LangChain4j బిగినర్స్ కోసం](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js బిగినర్స్ కోసం](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain బిగినర్స్ కోసం](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain4j ప్రారంభికులకు](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js ప్రారంభికులకు](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain ప్రారంభికులకు](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agents -[![AZD బిగినర్స్ కోసం](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI బిగినర్స్ కోసం](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP బిగినర్స్ కోసం](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents బిగినర్స్ కోసం](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / ఏజెంట్లు +[![AZD ప్రారంభికులకు](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI ప్రారంభికులకు](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP ప్రారంభికులకు](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI ఏజెంట్లు ప్రారంభికులకు](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Generative AI సిరీస్ -[![Generative AI బిగినర్స్ కోసం](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### జనరేటివ్ AI సిరీస్ +[![జనరేటివ్ AI ప్రారంభికులకు](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![జనరేటివ్ AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![జనరేటివ్ AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![జనరేటివ్ AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - + ### కోర్ లెర్నింగ్ -[![ML బిగినర్స్ కోసం](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![డేటా సైన్స్ బిగినర్స్ కోసం](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI బిగినర్స్ కోసం](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![సైబర్‌సెక్యూరిటీ బిగినర్స్ కోసం](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![వెబ్ డెవ్ బిగినర్స్ కోసం](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT బిగినర్స్ కోసం](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR అభివృద్ధి బిగినర్స్ కోసం](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![ML ప్రారంభికులకు](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![డేటా సైన్స్ ప్రారంభికులకు](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI ప్రారంభికులకు](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![సైబర్‌సెక్యూరిటీ ప్రారంభికులకు](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![వెబ్ డెవలప్‌మెంట్ ప్రారంభికులకు](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT ప్రారంభికులకు](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR అభివృద్ధి ప్రారంభికులకు](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### కోపైలట్ సిరీస్ -[![AI పేరడ ప్రోగ్రామింగ్ కోసం కోపైలట్](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![C#/.NET కోపైలట్](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![కోపైలట్ అడ్వెంచర్](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![AI జంట ప్రోగ్రామింగ్ కోసం Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![C#/.NET కోసం Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot సాహసం](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## బాధ్యతాయుత AI +## బాధ్యతాయుతమైన AI + +Microsoft మా కస్టమర్లు మా AI ఉత్పత్తులను బాధ్యతగా ఉపయోగించడంలో సహాయం చేయటానికి, మా అనుభవాలను పంచుకోవటానికి, మరియు Transparency Notes మరియు Impact Assessments వంటి సాధనాల ద్వారా నమ్మకాల ఆధారిత భాగస్వామ్యాలను నిర్మించడానికి కట్టుబడి ఉంది. ఈ వనరులలో చాలా వాటిని మీరు [https://aka.ms/RAI](https://aka.ms/RAI) వద్ద కనుగొనవచ్చు. +Microsoft బాధ్యతాయుతమైన AI పద్ధతులు మా AI సూత్రాలు – న్యాయం, విశ్వాసక్యత మరియు సురక్షత, గోప్యత మరియు భద్రత, సమగ్రత, పారదర్శకత, మరియు బాధ్యతాయుతత – పై ఆధారపడి ఉంటాయి. + +ఈ నమూనా వంటి పెద్ద-స్థాయి సహజ భాష, చిత్రం, మరియు గాత్ర నమూనాలు అన్యాయంగా, అసమ్మతంగా లేదా అవమానకరంగానే ప్రవర్తించే అవకాశం ఉంది, ఇది ద్రోహాలను కలిగించవచ్చు. దయచేసి [Azure OpenAI సర్వీస్ Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ను సంప్రదించి ప్రమాదాలు మరియు పరిమితుల గురించి సమాచారం పొందండి. -మైక్రోసాఫ్ట్ మా కస్టమర్లకు మా AI ఉత్పత్తులను బాధ్యతాయుతంగా ఉపయోగించుకోవడంలో సహాయం చేయడానికి प्रतिबद्धం, మా అభ్యాసాలను పంచుకోవడం, మరియు ట్రాన్స్పరెన్సీ నోట్స్ మరియు ఇంపెక్ట్ అస్సెస్‌మెంట్‌ల వంటి పరికరాల ద్వారా విశ్వాస లేని భాగస్వామ్యాలను నిర్మించడం. ఈ వనరులలో చాలావరకు [https://aka.ms/RAI](https://aka.ms/RAI) లో కనుగొనవచ్చు. -మైక్రోసాఫ్ట్ బాధ్యతాయుత AI విధానం మా AI సూత్రాల ఆధారంగా ఉంటుంది: న్యాయం, విశ్వసనీయత మరియు భద్రత, గోప్యత మరియు భద్రత, సమగ్రత, పారదర్శకత మరియు బాధ్యత. -ఈ నమూనాలో ఉపయోగించిన పెద్ద స్కేల్ సహజ భాష, చిత్రం, మరియు మాట మోడల్స్ - అవి అన్యాయం, అప్రామాణికత లేదా అవమానకరమైన రూపంలో ప్రవర్తించవచ్చు, మరియు దాని వల్ల హానులు రావచ్చు. ప్రాథమిక జాగ్రత్తలు మరియు పరిమితుల గురించి తెలియకుండానే మీరు వాడకూడదు, దయచేసి [Azure OpenAI సేవా పారదర్శకత నోట్](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) ని సంప్రదించండి. -ఈ ప్రమాదాలను నివారించడానికి సూచించబడిన విధానం మీ ఆర్కిటెక్చర్‌లో హానికరమైన ప్రవర్తనను గుర్తించి నివారించగల సేఫ్టీ సిస్టమ్‌ను చేర్చడం. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ఒక స్వతంత్ర రక్షణ పొరను అందిస్తుంది, ఇది అప్లికేషన్లు మరియు సేవలలో హానికరమైన వినియోగదారు-సృష్టించిన మరియు AI-సృష్టించిన కంటెంట్‌ను గుర్తించగలదు. Azure AI Content Safety టెక్స్ట్ మరియు చిత్రం APIల‌ను కలిగి ఉంది, ఇవి హానికరమైన పదార్థాన్ని గుర్తించడానికి అనుమతిస్తాయి. Microsoft Foundryలో, Content Safety సేవ ద్వారా మీరు వివిధ మోడాలిటీలలో హానికరమైన కంటెంట్‌ను గుర్తించడానికి నమూనా కోడ్‌ను వీక్షించడానికి, అన్వేషించడానికి మరియు ప్రయత్నించడానికి వీలుగా ఉంటుంది. ఈ కింది [quickstart డాక్యుమెంటేషన్](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) మీరు సేవకు అభ్యర్థనలను చేయడంలో గైడ్ చేస్తుంది. +ఈ ప్రమాదాలను తగ్గించడానికి సూచించబడిన పద్ధతి మీ ఆర్కిటెక్చర్‌లో హానికరమైన ప్రవర్తనను గుర్తించి నిరోధించగల సేఫ్టీ సిస్టమ్‌ను చేర్చడమే. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) అనేది స్వతంత్ర రక్షణా పొరను అందిస్తుంది, ఇది అప్లికేషన్ల మరియు సేవలలో హానికరమైన వాడుకరి-ఉత్పన్నము మరియు AI-ఉత్పన్నము కంటెంట్‌ను గుర్తించగలదు. Azure AI Content Safety టెక్స్ట్ మరియు ఇమేజ్ API లను కలిగి ఉంది, ఇవి హానికరమైన పదార్థాన్ని గుర్తించడానికి అనుమతిస్తాయి. Microsoft Foundry లో, Content Safety సేవ మీకు వివిధ మోడ్‌ల మధ్య హానికరమైన కంటెంట్‌ను గుర్తించడానికి నమూనా కోడ్‌ను వీక్షించడానికి, అన్వేషించడానికి మరియు ప్రయత్నించడానికి అవకాశాన్ని ఇస్తుంది. ఈ క్రింది [quickstart డాక్యుమెంటేషన్](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ఈ సేవకు అభ్యర్థనలు చేసే విధానంలో మీకు మార్గనిర్దేశం చేస్తుంది. -మరొక అంశం వెయ్యాలి అంటే మొత్తం అప్లికేషన్ పనితీరు. బహుళ-మోడల్ మరియు బహుళ-మోడల్స్ అప్లికేషన్లతో, మేము పనితీరును మీరూ మరియు మీ వినియోగదారులు ఆశించే విధంగా వ్యవహరించడం అని భావిస్తాము, హానికరమైన అవుట్పుట్‌లు ఉత్పత్తి చేయకుండా ఉండటం సహా. మీరు మీ మొత్త అప్లికేషన్ పనితీరును [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) ఉపయోగించి అంచనా వేయడం ముఖ్యము. మీరు [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) సృష్టించి, వాటితో అంచనా వేయగలుగుతారు. +మరొక విషయం పరిగణనలోకి తీసుకోవలసినదైనది మొత్తం అప్లికేషన్ ప్రదర్శన. బహుముఖ మరియు బహుమోడల్ అప్లికేషన్లతో, మేము ప్రదర్శన అంటే మీరు మరియు మీ వినియోగదారులు అనుకున్న విధంగా సిస్టమ్ పనిచేస్తుంది, హానికరమైన ఔట్‌పుట్‌లు ఉత్పత్తి కాకుండా ఉంటాయి అని భావిస్తాము. [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) ఉపయోగించి మీ మొత్తం అప్లికేషన్ యొక్క ప్రదర్శనను అంచనా వేయడం ముఖ్యం. మీరు కూడా [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) తో సృష్టించి, అంచనా వేయగల సామర్థ్యం కలిగి ఉన్నారు. -మీ AI అప్లికేషన్‌ను మీ అభివృద్ధి పర్యావరణంలో [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ఉపయోగించి అంచనా వేయవచ్చు. ఒక పరీక్ష డేటాసెట్ లేదా లక్ష్యము ఇచ్చినప్పుడల్లా, మీ జనరేటివ్ AI అప్లికేషన్ పుట్టిపడినది అంతర్జాతీయ ప్రమాణాల లేదా మీరు ఎంచుకున్న కస్టమ్ ఈవాల్యుటర్ల ద్వారా గణాంకపరంగా కొలవబడుతుంది. మీ సిస్టమ్‌ను అంచనా వేయడానికి Azure AI Evaluation SDKతో ప్రారంభించడానికి, మీరు [quickstart గైడ్](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)‌ను అనుసరించవచ్చు. మీరు ఒక ఈవాల్యుయేషన్ ను అమలు చేసిన వెంటనే, మీరు [Microsoft Foundryలో ఫలితాలను విజువలైజ్](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) చేయవచ్చు. +మీరు [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ఉపయోగించి మీ అభివృద్ధి పరిసరంలో మీ AI అప్లికేషన్‌ను అంచనా వేయవచ్చు. ఒక పరీక్ష డేటాసెట్ లేదా ఒక లక్ష్యం ఇవ్వబడినప్పుడు, మీ జనరేటివ్ AI అప్లికేషన్ ఉత్పత్తులను మీరు ఎంచుకున్న బిల్ట్-ఇన్ లేదా అనుకూల అంచనా కారులకు సంఖ్యాబద్ధంగా కొలవబడతాయి. మీ సిస్టమ్‌ను అంచనా లోకి తీసుకోవడానికి Azure AI Evaluation SDK తో ప్రారంభం కాని విధానం కోసం మీరు [quickstart గైడ్](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) ను అనుసరించవచ్చు. ఒక సారి మీరు అంచనా నడుపుతుంటే, మీరు ఫలితాలను [Microsoft Foundry లో వీక్షించవచ్చు](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## ట్రేడ్మార్కులు -ఈ ప్రాజెక్ట్ ప్రాజెక్టులు, ఉత్పత్తులు లేదా సేవలకు ట్రేడ్మార్కులు లేదా లోగోలు కలిగి ఉండవచ్చు. Microsoft ట్రేడ్మార్కులు లేదా లోగోల అధికారిక వినియోగం [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) కు అనుగుణంగా ఉండాలి మరియు దాని ప్రకారం మాత్రమే ఉండాలి. -మార్పులు చేసిన సంస్కరణలలో Microsoft ట్రేడ్మార్కులు లేదా లోగోలు వాడటం Microsoft స్పాన్సర్‌షిప్‌ను కలిగించినట్లు లేదా గందరగోళం కలిగించే విధంగా ఉండకూడదు. మూడవ పక్ష ట్రేడ్మార్కులు లేదా లోగోల ఏదైనా వినియోగం ఆ మూడవ పక్ష పాలసీలకు అనుగుణంగా ఉండాలి. +ఈ ప్రాజెక్ట్ ప్రాజెక్టులు, ఉత్పత్తులు లేదా సేవల కోసం ట్రేడ్మార్కులు లేదా లోగోలు ఉండవచ్చు. Microsoft ట్రేడ్మార్కులు లేదా లోగోలను అధీకృతంగా ఉపయోగించడం [Microsoft యొక్క ట్రేడ్మార్క్ & బ్రాండ్ మార్గదర్శకాలు](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) ప్రకారం ఉండాలి మరియు వాటిని అనుసరించాలి. +ఈ ప్రాజెక్ట్ యొక్క మోడిఫైడ్ వెర్షన్‌లలో Microsoft ట్రేడ్మార్కులు లేదా లోగోలను ఉపయోగించడం గందరగోళాన్ని కలిగించకూడదు లేదా Microsoft అనుబంధతను సూచించకూడదు. మూడవ పార్టీ ట్రేడ్మార్కులు లేదా లోగోలను ఉపయోగించడం ఆ మూడవ పార్టీ విధానాలకు లోబడి ఉంటుంది. -## సహాయం పొందుట +## సహాయం పొందడం -మీరు ఏదైనా ఇబ్బంది పడితే లేదా AI అప్లికేషన్లు నిర్మించడంపై ఏవైనా ప్రశ్నలు ఉంటే, చేరండి: +మీరు అడ్డంకిలో పడినట్టైతే లేదా AI యాప్‌ల నిర్మాణంపై ఏవైనా ప్రశ్నలు ఉంటే, చేరండి: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -మీకు ఉత్పత్తి ప్రసక్తి సూచనలు లేదా లోపాలు ఉన్నట్లయితే సందర్శించండి: +మీరు ఉత్పత్తి ఫీడ్‌బ్యాక్ లేదా దోషాలు ఎదుర్కొని ఉంటే సందర్శించండి: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**అస్పష్టత**: -ఈ డాక్యుమెంట్ [Co-op Translator](https://github.com/Azure/co-op-translator) అనే AI అనువాద సేవ ఉపయోగించి అనువదించబడింది. మనం ఖచ్చితత్వానికి ప్రయత్నిస్తుంటే కూడా, ఆటోమేటిక్ అనువాదాల్లో పొరపాట్లు లేదా లోపాలు ఉండవచ్చు. అసలు డాక్యుమెంట్ తన స్వదేశీ భాషలో వున్నదే అధికారం కలిగిన మూలం గా పరిగణించాలి. కీలకమైన సమాచారం కోసం, ప్రొఫెషనల్ మానవ అనువాదం సిఫార్సు చేయబడుతుంది. ఈ అనువాదం ఉపయోగించుట వలన వచ్చిన ఏదైనా అపార్థాలు లేదా తప్పుదోషాల కోసం మేము బాధ్యులు క కాదు. +**అస్వీకరణ**: +ఈ పత్రం AI అనువాద సేవ [Co-op Translator](https://github.com/Azure/co-op-translator) ఉపయోగించి అనువదించబడింది. మేము ఖచ్చితత్వానికి ప్రయత్నిస్తున్నప్పటికీ, ఆటోమేటెడ్ అనువాదాలు తప్పులు లేదా అసమగ్రతలను కలిగి ఉండవచ్చు. దాని స్వదేశ భాషలో ఉన్న అసలు పత్రాన్ని అధికారం కలిగిన మూలంగా పరిగణించాలి. కీలకమైన సమాచారం కోసం, ప్రొఫెషనల్ మానవ అనువాదాన్ని సిఫారసు చేస్తాము. ఈ అనువాదం ఉపయోగం వల్ల కలిగే ఏవైనా అపార్థాలు లేదా తప్పుదారులు కోసం మేము బాధ్యత వహించము. \ No newline at end of file diff --git a/translations/te/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/te/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..ad62a3a8b --- /dev/null +++ b/translations/te/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 మైక్రోసాఫ్ట్ phi-4-mini & SQLite FTS5 (జీరో-క్లౌడ్ SLM) తో హైబ్రిడ్ RAG\n", + "\n", + "> **రచయిత:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **ఫోకస్:** చిన్న భాషా నమూనాలు (SLMs), SQLite FTS5 BM25, డెన్స్ ఎంబెడ్డింగ్స్, రెసిప్రోకల్ ర్యాంక్ ఫ్యూజన్ (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. ప్రేరణ: స్థానిక SLMలలో కీవర్డ్ రికాల్ సమస్య\n", + "కేవలం డెన్స్ వెక్టార్ ఎంబెడ్డింగ్స్ పైన ఆధారపడే ప్రామాణిక RAG నిర్మాణాలు సాధారణంగా ఖచ్చిత సంఖ్యా టోకన్లు (ఉదా: `2,340,000 TL`, ఒప్పంద కోడ్లు, ఖాతా సంఖ్యలు) ని తిరిగి పొందడంలో విఫలమవుతాయి. \n", + "వేరేవైపు, స్పార్స్ లెక్సికల్ సెర్చ్ (BM25) అర్థ సంబంధి సమానార్థకాలు మరియు పదబంధ ప్రశ్నలను మిస్ అవుతుంది.\n", + "\n", + "ఈ కుక్‌బుక్ **హై-స్పీడ్, ఇన్-మెమరీ హైబ్రిడ్ రిట్రీవల్ ఇంజిన్** ను ఎలా అమలు చేయాలో చూపిస్తుంది, ఇది కలిపినది:\n", + "1. **డెన్స్ వెక్టార్లు** (కోసైన్ సమానత్వం)\n", + "2. **స్పార్స్ లెక్సికల్ సెర్చ్** (SQLite FTS5 BM25)\n", + "3. **రెసిప్రోకల్ ర్యాంక్ ఫ్యూజన్ (RRF, $k=60$)**\n", + "4. Microsoft `phi-4-mini`తో **గ్రౌండెడ్ సిటేషన్ జ‌న‌రేషన్ (`[1]`, `[2]`)**.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. ద్వంద్వ SQLite స్కీమా (సాంద్ర వెక్టార్లు + వర్చువల్ FTS5 BM25 పట్టిక)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. నమూనా ఇంజెస్టివ్ & ఎగ్జిక్యూషన్ బెంచ్‌మార్క్\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Microsoft phi-4-mini కోసం గ్రౌండెడ్ ప్రాంప్ట్ రూపకల్పన\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**అస్వీకరణ**:\nఈ పత్రం AI అనువాద సేవ [Co-op Translator](https://github.com/Azure/co-op-translator) ఉపయోగించి అనువదించబడింది. మేము ఖచ్చితత్వానికి ప్రయత్నిస్తున్నప్పటికీ, ఆటోమేటెడ్ అనువాదాలు తప్పులు లేదా అసమగ్రతలను కలిగి ఉండవచ్చు. దాని స్వదేశ భాషలో ఉన్న అసలు పత్రాన్ని అధికారం కలిగిన మూలంగా పరిగణించాలి. కీలకమైన సమాచారం కోసం, ప్రొఫెషనల్ మానవ అనువాదాన్ని సిఫారసు చేస్తాము. ఈ అనువాదం ఉపయోగం వల్ల కలిగే ఏవైనా అపార్థాలు లేదా తప్పుదారులు కోసం మేము బాధ్యత వహించము.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/th/.co-op-translator.json b/translations/th/.co-op-translator.json index 8d30951f9..069c97abe 100644 --- a/translations/th/.co-op-translator.json +++ b/translations/th/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "th" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:58:45+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:14:41+00:00", "source_file": "README.md", "language_code": "th" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "th" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:19:01+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "th" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:02+00:00", diff --git a/translations/th/README.md b/translations/th/README.md index 7821256de..9f751a2b5 100644 --- a/translations/th/README.md +++ b/translations/th/README.md @@ -1,42 +1,42 @@ # Phi Cookbook: ตัวอย่างใช้งานจริงกับโมเดล Phi ของ Microsoft -[![เปิดและใช้ตัวอย่างใน GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![เปิดและใช้งานตัวอย่างใน GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![เปิดใน Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![ผู้ร่วมพัฒนา GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![ปัญหา GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![คำขอดึง GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![ยินดีต้อนรับคำขอดึง](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![ผู้ร่วมพัฒนาบน GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![ปัญหาบน GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![คำร้องขอ pull บน GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![ยินดีต้อนรับคำร้องขอ PR](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![ผู้ติดตาม GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![โฟร์ก GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![ดาว GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![ผู้ติดตามบน GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![ผู้แยกโปรเจคบน GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![ดาวบน GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi คือชุดโมเดล AI แบบโอเพนซอร์สที่พัฒนาโดย Microsoft +Phi คือชุดโมเดล AI แบบโอเพ่นซอร์สที่พัฒนาโดย Microsoft -Phi ปัจจุบันเป็นโมเดลภาษาขนาดเล็ก (SLM) ที่ทรงพลังและคุ้มค่าที่สุด มีการวัดผลที่ดีมากในหลายภาษา การให้เหตุผล การสร้างข้อความ/แชท โค้ดดิ้ง รูปภาพ เสียง และสถานการณ์อื่นๆ +Phi เป็นโมเดลภาษาเล็ก (SLM) ที่ทรงพลังและคุ้มค่าที่สุดในขณะนี้ พร้อมด้วยเกณฑ์มาตรฐานที่ดีมากในหลายภาษา การใช้เหตุผล การสร้างข้อความ/แชท การเขียนโปรแกรม ภาพ เสียง และสถานการณ์อื่นๆ -คุณสามารถปรับใช้ Phi บนคลาวด์หรืออุปกรณ์ที่ EDGE และสามารถสร้างแอปพลิเคชัน AI เชิงสร้างสรรค์ได้ง่ายๆ ด้วยพลังการคำนวณที่จำกัด +คุณสามารถปรับใช้ Phi บนคลาวด์หรืออุปกรณ์เอดจ์ และสามารถสร้างแอปพลิเคชัน AI สร้างสรรค์ได้อย่างง่ายดายแม้มีพลังคอมพิวเตอร์จำกัด -ทำตามขั้นตอนเหล่านี้เพื่อเริ่มใช้งานแหล่งข้อมูลนี้: -1. **Fork ที่เก็บนี้:** คลิก [![โฟร์ก GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **โคลนที่เก็บนี้:** `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**เข้าร่วมชุมชน Microsoft AI Discord และพบปะผู้เชี่ยวชาญและนักพัฒนาร่วมกัน**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +ทำตามขั้นตอนเหล่านี้เพื่อเริ่มต้นใช้งานแหล่งข้อมูลเหล่านี้: +1. **ฟอร์ครีโพซิทอรี**: คลิก [![ผู้แยกโปรเจคบน GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **โคลนรีโพซิทอรี**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**เข้าร่วมชุมชน Microsoft AI Discord และพบกับผู้เชี่ยวชาญรวมถึงนักพัฒนาร่วม**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/th/cover.eb18d1b9605d754b.webp) ### 🌐 รองรับหลายภาษา -#### รองรับผ่าน GitHub Action (อัตโนมัติและอัปเดตเสมอ) +#### รองรับโดย GitHub Action (อัตโนมัติ & อัปเดตเสมอ) -[อาหรับ](../ar/README.md) | [เบงกาลี](../bn/README.md) | [บัลแกเรีย](../bg/README.md) | [พม่า (เมียนมา)](../my/README.md) | [จีน (ตัวย่อ)](../zh-CN/README.md) | [จีน (ตัวเต็ม, ฮ่องกง)](../zh-HK/README.md) | [จีน (ตัวเต็ม, มาเก๊า)](../zh-MO/README.md) | [จีน (ตัวเต็ม, ไต้หวัน)](../zh-TW/README.md) | [โครเอเชีย](../hr/README.md) | [เช็ก](../cs/README.md) | [เดนมาร์ก](../da/README.md) | [ดัตช์](../nl/README.md) | [เอสโตเนีย](../et/README.md) | [ฟินแลนด์](../fi/README.md) | [ฝรั่งเศส](../fr/README.md) | [เยอรมัน](../de/README.md) | [กรีก](../el/README.md) | [ฮีบรู](../he/README.md) | [ฮินดี](../hi/README.md) | [ฮังการี](../hu/README.md) | [อินโดนีเซีย](../id/README.md) | [อิตาลี](../it/README.md) | [ญี่ปุ่น](../ja/README.md) | [กันนาดา](../kn/README.md) | [เขมร](../km/README.md) | [เกาหลี](../ko/README.md) | [ลิทัวเนีย](../lt/README.md) | [มาเลย์](../ms/README.md) | [มาลายาลัม](../ml/README.md) | [มาราธี](../mr/README.md) | [เนปาล](../ne/README.md) | [ไพจิ้นไนจีเรีย](../pcm/README.md) | [นอร์เวย์](../no/README.md) | [เปอร์เซีย (ฟาร์ซี)](../fa/README.md) | [โปแลนด์](../pl/README.md) | [โปรตุเกส (บราซิล)](../pt-BR/README.md) | [โปรตุเกส (โปรตุเกส)](../pt-PT/README.md) | [ปัญจาบี (กูรมุกฮี)](../pa/README.md) | [โรมาเนีย](../ro/README.md) | [รัสเซีย](../ru/README.md) | [เซอร์เบีย (คีริลลิก)](../sr/README.md) | [สโลวัก](../sk/README.md) | [สโลวีเนีย](../sl/README.md) | [สเปน](../es/README.md) | [สวาฮิลี](../sw/README.md) | [สวีเดน](../sv/README.md) | [ตากาล็อก (ฟิลิปปินส์)](../tl/README.md) | [ทมิฬ](../ta/README.md) | [เทลูกู](../te/README.md) | [ไทย](./README.md) | [ตุรกี](../tr/README.md) | [ยูเครน](../uk/README.md) | [อูรดู](../ur/README.md) | [เวียดนาม](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](./README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **ต้องการโคลนแบบโลคัล?** +> **ต้องการโคลนแบบท้องถิ่นใช่ไหม?** > -> ที่เก็บนี้มีการแปลภาษามากกว่า 50 ภาษา ซึ่งทำให้ขนาดดาวน์โหลดเพิ่มขึ้นอย่างมาก หากต้องการโคลนโดยไม่รวมการแปล ให้ใช้ sparse checkout: +> รีโพซิทอรีนี้มีการแปลมากกว่า 50 ภาษา ซึ่งเพิ่มขนาดการดาวน์โหลดอย่างมาก หากต้องการโคลนโดยไม่รวมการแปล ให้ใช้ sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,47 +52,47 @@ Phi ปัจจุบันเป็นโมเดลภาษาขนาด > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> นี่จะมอบทุกอย่างที่คุณต้องการเพื่อทำคอร์สให้เสร็จเร็วขึ้นมาก +> วิธีนี้จะให้ทุกสิ่งที่คุณต้องการเพื่อทำคอร์สนี้ให้เสร็จเร็วขึ้นมาก ## สารบัญ - บทนำ - [ยินดีต้อนรับสู่ครอบครัว Phi](./md/01.Introduction/01/01.PhiFamily.md) - - [การตั้งค่าสิ่งแวดล้อมของคุณ](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [ตั้งค่าสภาพแวดล้อมของคุณ](./md/01.Introduction/01/01.EnvironmentSetup.md) - [ทำความเข้าใจเทคโนโลยีหลัก](./md/01.Introduction/01/01.Understandingtech.md) - - [ความปลอดภัย AI สำหรับโมเดล Phi](./md/01.Introduction/01/01.AISafety.md) - - [การรองรับฮาร์ดแวร์ Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [โมเดล Phi และการใช้งานบนแพลตฟอร์มต่างๆ](./md/01.Introduction/01/01.Edgeandcloud.md) + - [ความปลอดภัยของ AI สำหรับโมเดล Phi](./md/01.Introduction/01/01.AISafety.md) + - [การสนับสนุนฮาร์ดแวร์ Phi](./md/01.Introduction/01/01.Hardwaresupport.md) + - [โมเดล Phi & ความพร้อมใช้งานบนแพลตฟอร์มต่างๆ](./md/01.Introduction/01/01.Edgeandcloud.md) - [การใช้ Guidance-ai และ Phi](./md/01.Introduction/01/01.Guidance.md) - - [โมเดล GitHub Marketplace](https://github.com/marketplace/models) - - [แคตตาล็อกโมเดล Azure AI](https://ai.azure.com) + - [โมเดลใน GitHub Marketplace](https://github.com/marketplace/models) + - [แค็ตตาล็อกโมเดล AI ของ Azure](https://ai.azure.com) -- การรัน Phi ในสภาพแวดล้อมต่างๆ +- การสืบค้น Phi ในสภาพแวดล้อมต่างๆ - [Hugging face](./md/01.Introduction/02/01.HF.md) - [โมเดล GitHub](./md/01.Introduction/02/02.GitHubModel.md) - - [แคตตาล็อกโมเดล Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Microsoft Foundry Model Catalog](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- การรัน Phi Family - - [การรัน Phi บน iOS](./md/01.Introduction/03/iOS_Inference.md) - - [การรัน Phi บน Android](./md/01.Introduction/03/Android_Inference.md) - - [การรัน Phi บน Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [การรัน Phi บน AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [การรัน Phi กับ Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [การรัน Phi บน Local Server](./md/01.Introduction/03/Local_Server_Inference.md) - - [การรัน Phi บน Remote Server โดยใช้ AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [การรัน Phi กับ Rust](./md/01.Introduction/03/Rust_Inference.md) - - [การรัน Phi--Vision ในเครื่อง](./md/01.Introduction/03/Vision_Inference.md) - - [การรัน Phi กับ Kaito AKS, Azure Containers(รองรับอย่างเป็นทางการ)](./md/01.Introduction/03/Kaito_Inference.md) -- [การ Quantify Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) - - [การ Quantize Phi-3.5 / 4 โดยใช้ llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [การ Quantize Phi-3.5 / 4 โดยใช้ Generative AI extensions สำหรับ onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [การ Quantize Phi-3.5 / 4 โดยใช้ Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [การ Quantize Phi-3.5 / 4 โดยใช้ Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) +- การสืบค้นครอบครัว Phi + - [การสืบค้น Phi บน iOS](./md/01.Introduction/03/iOS_Inference.md) + - [การสืบค้น Phi บน Android](./md/01.Introduction/03/Android_Inference.md) + - [การสืบค้น Phi บน Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [การสืบค้น Phi บน AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [การสืบค้น Phi ด้วย Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [การสืบค้น Phi บนเซิร์ฟเวอร์ท้องถิ่น](./md/01.Introduction/03/Local_Server_Inference.md) + - [การสืบค้น Phi บนเซิร์ฟเวอร์ระยะไกลโดยใช้ AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [การสืบค้น Phi ด้วย Rust](./md/01.Introduction/03/Rust_Inference.md) + - [การสืบค้น Phi-Vision บนพื้นที่ท้องถิ่น](./md/01.Introduction/03/Vision_Inference.md) + - [การสืบค้น Phi ด้วย Kaito AKS, Azure Containers (รองรับอย่างเป็นทางการ)](./md/01.Introduction/03/Kaito_Inference.md) +- [การควอนไทซ์ครอบครัว Phi](./md/01.Introduction/04/QuantifyingPhi.md) + - [การควอนไทซ์ Phi-3.5 / 4 ด้วย llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [การควอนไทซ์ Phi-3.5 / 4 ด้วยส่วนขยาย Generative AI สำหรับ onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [การควอนไทซ์ Phi-3.5 / 4 ด้วย Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [การควอนไทซ์ Phi-3.5 / 4 ด้วย Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - การประเมิน Phi - [AI ที่รับผิดชอบ](./md/01.Introduction/05/ResponsibleAI.md) @@ -100,54 +100,56 @@ Phi ปัจจุบันเป็นโมเดลภาษาขนาด - [การใช้ Promptflow สำหรับการประเมิน](./md/01.Introduction/05/Promptflow.md) - RAG กับ Azure AI Search - - [วิธีใช้ Phi-4-mini และ Phi-4-multimodal(RAG) กับ Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [วิธีใช้ Phi-4-mini และ Phi-4-multimodal (RAG) กับ Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud Local Hybrid RAG กับ SQLite FTS5 และ phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - ตัวอย่างการพัฒนาแอปพลิเคชัน Phi - - แอปพลิเคชันข้อความและแชท - - Phi-4 ตัวอย่าง + - แอปข้อความและแชท + - ตัวอย่าง Phi-4 - [📓] [แชทกับโมเดล Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [แชทกับโมเดล Phi-4 local ONNX แบบ .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [แชทแอปพลิเคชันคอนโซล .NET กับ Phi-4 ONNX โดยใช้ Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 ตัวอย่าง - - [แชทบอทโลคัลในเว็บเบราว์เซอร์โดยใช้ Phi3, ONNX Runtime Web และ WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [โมเดลหลายตัว - การโต้ตอบ Phi-3-mini และ OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [แชทกับโมเดล Phi-4 ONNX แบบท้องถิ่น .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [แอปคอนโซลแชท .NET กับ Phi-4 ONNX ใช้ Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - ตัวอย่าง Phi-3 / 3.5 + - [แชทบอทในเครื่องในเบราว์เซอร์โดยใช้ Phi3, ONNX Runtime Web และ WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [แชท OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [โมเดลหลายตัว - Phi-3-mini โต้ตอบได้และ OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - การสร้าง wrapper และใช้ Phi-3 กับ MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [การปรับแต่งโมเดล - วิธีปรับแต่งโมเดล Phi-3-min สำหรับ ONNX Runtime Web ด้วย Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [การปรับแต่งโมเดล - วิธีการปรับแต่งโมเดล Phi-3-min สำหรับ ONNX Runtime Web ด้วย Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [แอป WinUI3 กับ Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - - [ตัวอย่างแอปโน้ตที่ขับเคลื่อนด้วย AI แบบหลายโมเดล WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) + -[ตัวอย่างแอปบันทึกข้อความ AI หลายโมเดลใน WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) - [ปรับแต่งและรวมโมเดล Phi-3 ที่กำหนดเองกับ Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - [ปรับแต่งและรวมโมเดล Phi-3 ที่กำหนดเองกับ Prompt flow ใน Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [ประเมินโมเดล Phi-3 / Phi-3.5 ที่ปรับแต่งใน Microsoft Foundry มุ่งเน้นไปที่หลักการ Responsible AI ของ Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [ตัวอย่างการทำนายภาษา Phi-3.5-mini-instruct (จีน/อังกฤษ)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [ใช้ Windows GPU สร้างโซลูชัน Prompt flow กับ Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [ใช้ Microsoft Phi-3.5 tflite สร้างแอป Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [ตัวอย่าง Q&A .NET การใช้โมเดล ONNX Phi-3 ท้องถิ่นโดยใช้ Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [ประเมินโมเดล Phi-3 / Phi-3.5 ที่ปรับแต่งใน Microsoft Foundry โดยเน้นหลักจริยธรรม AI ของ Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [ตัวอย่างการทำนายภาษาของ Phi-3.5-mini-instruct (จีน/อังกฤษ)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU แชทบอท RAG](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [การใช้ GPU บน Windows เพื่อสร้างโซลูชัน Prompt flow ด้วย Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [การใช้ Microsoft Phi-3.5 tflite เพื่อสร้างแอป Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [ตัวอย่าง Q&A .NET โดยใช้โมเดล ONNX Phi-3 ในเครื่องด้วย Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - [แอปแชทคอนโซล .NET กับ Semantic Kernel และ Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - ตัวอย่างโค้ด SDK การอนุมาน Azure AI - - ตัวอย่าง Phi-4 - - [📓] [สร้างโค้ดโปรเจกต์โดยใช้ Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - ตัวอย่างโค้ด Azure AI Inference SDK + - ตัวอย่าง Phi-4 + - [📓] [สร้างโค้ดโปรเจกต์ด้วย Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - ตัวอย่าง Phi-3 / 3.5 - - [สร้าง Visual Studio Code GitHub Copilot Chat ของคุณเองกับครอบครัว Microsoft Phi-3](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [สร้าง Visual Studio Code Chat Copilot Agent ของคุณเองกับ Phi-3.5 โดย GitHub Models](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [สร้าง Visual Studio Code GitHub Copilot Chat ของคุณเองกับ Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [สร้างตัวแทน Chat Copilot ใน Visual Studio Code ของคุณเองด้วย Phi-3.5 โดย GitHub Models](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - ตัวอย่างเหตุผลขั้นสูง - - ตัวอย่าง Phi-4 + - ตัวอย่างการคิดขั้นสูง + - ตัวอย่าง Phi-4 - [📓] [ตัวอย่าง Phi-4-mini-reasoning หรือ Phi-4-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - [📓] [การปรับแต่ง Phi-4-mini-reasoning ด้วย Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [การปรับแต่ง Phi-4-mini-reasoning ด้วย Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-reasoning กับ GitHub Models](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-reasoning กับ Microsoft Foundry Models](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - ตัวอย่างสาธิต - - [ตัวอย่าง Phi-4-mini โฮสต์บน Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [ตัวอย่าง Phi-4-multimodal โฮสต์บน Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - ตัวอย่างการมองเห็น - - ตัวอย่าง Phi-4 + - [Phi-4-mini สาธิตบน Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal สาธิตบน Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - ตัวอย่าง Vision + - ตัวอย่าง Phi-4 - [📓] [ใช้ Phi-4-multimodal อ่านภาพและสร้างโค้ด](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - ตัวอย่าง Phi-3 / 3.5 - - [📓][Phi-3-vision-แปลงข้อความจากภาพเป็นข้อความ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-ข้อความจากภาพเป็นข้อความ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [สาธิต: Phi-3 รีไซเคิล](https://github.com/jennifermarsman/PhiRecycling/) @@ -155,70 +157,71 @@ Phi ปัจจุบันเป็นโมเดลภาษาขนาด - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - [📓][ตัวอย่าง Phi-3.5 Vision หลายเฟรมหรือหลายภาพ](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision โมเดล ONNX ท้องถิ่นโดยใช้ Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [เมนูพื้นฐาน Phi-3 Vision โมเดล ONNX ท้องถิ่นโดยใช้ Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [Phi-3 Vision โมเดล ONNX ในเครื่องโดยใช้ Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Phi-3 Vision โมเดล ONNX ในเครื่องแบบเมนูโดยใช้ Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - ตัวอย่าง Reasoning-Vision - - Phi-4-Reasoning-Vision-15B - - [📓] [ใช้ Phi-4-Reasoning-Vision-15B ตรวจจับการเดินลัดถนน](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - Phi-4-Reasoning-Vision-15B + - [📓] [ใช้ Phi-4-Reasoning-Vision-15B เพื่อตรวจจับการข้ามถนนอย่างผิดกฎหมาย](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [ใช้ Phi-4-Reasoning-Vision-15B ทำคณิตศาสตร์](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [ใช้ Phi-4-Reasoning-Vision-15B ตรวจจับ UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [ใช้ Phi-4-Reasoning-Vision-15B เพื่อตรวจจับ UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - ตัวอย่างคณิตศาสตร์ - - ตัวอย่าง Phi-4-Mini-Flash-Reasoning-Instruct [สาธิตคณิตศาสตร์กับ Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - ตัวอย่าง Phi-4-Mini-Flash-Reasoning-Instruct [สาธิตคณิตศาสตร์ด้วย Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - ตัวอย่างเสียง - - ตัวอย่าง Phi-4 - - [📓] [การถอดเสียงเสียงด้วย Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - ตัวอย่าง Phi-4 + - [📓] [การถอดเสียงเสียงโดยใช้ Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [ตัวอย่างเสียง Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [ตัวอย่างการแปลเสียงพูด Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [แอปพลิเคชันคอนโซล .NET ที่ใช้ Phi-4-multimodal Audio เพื่อวิเคราะห์ไฟล์เสียงและสร้างถอดความ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [แอปคอนโซล .NET ใช้ Phi-4-multimodal เพื่อวิเคราะห์ไฟล์เสียงและสร้างถอดเสียง](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - ตัวอย่าง MOE - ตัวอย่าง Phi-3 / 3.5 - - [📓] [ตัวอย่างโมเดล Mixture of Experts (MoEs) Phi-3.5 สำหรับโซเชียลมีเดีย](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [การสร้าง Retrieval-Augmented Generation (RAG) Pipeline กับ NVIDIA NIM Phi-3 MOE, Azure AI Search, และ LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [ตัวอย่าง Phi-3.5 Mixture of Experts Models (MoEs) ในสื่อสังคมออนไลน์](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [การสร้าง Retrieval-Augmented Generation (RAG) Pipeline ด้วย NVIDIA NIM Phi-3 MOE, Azure AI Search และ LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - ตัวอย่างการเรียกใช้ฟังก์ชัน + - ตัวอย่าง Function Calling - ตัวอย่าง Phi-4 🆕 - [📓] [การใช้ Function Calling กับ Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [การใช้ Function Calling เพื่อสร้าง multi-agents กับ Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [การใช้ Function Calling เพื่อสร้างตัวแทนหลายตัวด้วย Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [การใช้ Function Calling กับ Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [การใช้ Function Calling กับ ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - ตัวอย่างการผสมผสานมัลติโมดอล + - ตัวอย่างผสมมัลติโมดัล - ตัวอย่าง Phi-4 🆕 - - [📓] [การใช้ Phi-4-multimodal เป็นนักข่าวเทคโนโลยี](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [แอปคอนโซล .NET ที่ใช้ Phi-4-multimodal วิเคราะห์ภาพ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- การปรับแต่งโมเดล Phi - - [สถานการณ์การปรับแต่งโมเดล](./md/03.FineTuning/FineTuning_Scenarios.md) - - [การปรับแต่งโมเดล vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [ให้ Phi-3 กลายเป็นผู้เชี่ยวชาญในอุตสาหกรรม](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [การปรับแต่ง Phi-3 ด้วย AI Toolkit สำหรับ VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [การปรับแต่ง Phi-3 ด้วย Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [การปรับแต่ง Phi-3 ด้วย Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [การปรับแต่ง Phi-3 ด้วย QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [การปรับแต่ง Phi-3 ด้วย Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [การปรับแต่ง Phi-3 ด้วย Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [การปรับแต่งด้วย Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [ห้องปฏิบัติการปรับแต่งด้วย Microsoft Olive Hands-On](./md/03.FineTuning/olive-lab/readme.md) - - [การปรับแต่ง Phi-3-vision ด้วย Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [📓] [ใช้ Phi-4-multimodal เป็นนักข่าวเทคโนโลยี](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [แอปคอนโซล .NET ใช้ Phi-4-multimodal เพื่อวิเคราะห์ภาพ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- การปรับแต่ง Phi + - [สถานการณ์การปรับแต่ง](./md/03.FineTuning/FineTuning_Scenarios.md) + - [การปรับแต่ง vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [ปรับแต่งให้ Phi-3 เป็นผู้เชี่ยวชาญในอุตสาหกรรม](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [ปรับแต่ง Phi-3 ด้วย AI Toolkit สำหรับ VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [ปรับแต่ง Phi-3 ด้วย Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [ปรับแต่ง Phi-3 ด้วย Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [ปรับแต่ง Phi-3 ด้วย QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [ปรับแต่ง Phi-3 ด้วย Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [ปรับแต่ง Phi-3 ด้วย Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [ปรับแต่งด้วย Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [ปรับแต่งด้วย Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) + - [ปรับแต่ง Phi-3-vision ด้วย Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [การปรับแต่ง Phi-3 ด้วย Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - [การปรับแต่ง Phi-3-vision (สนับสนุนอย่างเป็นทางการ)](./md/03.FineTuning/FineTuning_Vision.md) - - [การปรับแต่ง Phi-3 ด้วย Kaito AKS , Azure Containers (การสนับสนุนอย่างเป็นทางการ)](./md/03.FineTuning/FineTuning_Kaito.md) + - [การปรับแต่ง Phi-3 ด้วย Kaito AKS, Azure Containers (สนับสนุนอย่างเป็นทางการ)](./md/03.FineTuning/FineTuning_Kaito.md) - [การปรับแต่ง Phi-3 และ 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) -- ห้องปฏิบัติการเชิงปฏิบัติการ - - [สำรวจโมเดลล้ำสมัย: LLMs, SLMs, การพัฒนาท้องถิ่น และอื่น ๆ](https://github.com/microsoft/aitour-exploring-cutting-edge-models) +- ห้องปฏิบัติการ + - [สำรวจโมเดลล้ำสมัย: LLMs, SLMs, การพัฒนาท้องถิ่น และอื่นๆ](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - [ปลดล็อกศักยภาพ NLP: การปรับแต่งด้วย Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- งานวิจัยทางวิชาการและสิ่งพิมพ์ - - [หนังสือเรียนคือสิ่งที่คุณต้องการ II: รายงานทางเทคนิค phi-1.5](https://arxiv.org/abs/2309.05463) - - [รายงานทางเทคนิค Phi-3: โมเดลภาษาที่มีความสามารถสูงบนโทรศัพท์ของคุณ](https://arxiv.org/abs/2404.14219) +- บทความวิจัยทางวิชาการและสิ่งตีพิมพ์ + - [Textbooks Are All You Need II: รายงานทางเทคนิค phi-1.5](https://arxiv.org/abs/2309.05463) + - [รายงานทางเทคนิค Phi-3: โมเดลภาษาแบบมีศักยภาพสูงบนโทรศัพท์ของคุณ](https://arxiv.org/abs/2404.14219) - [รายงานทางเทคนิค Phi-4](https://arxiv.org/abs/2412.08905) - - [รายงานทางเทคนิค Phi-4-Mini: โมเดลภาษามัลติโมเดลที่กะทัดรัดแต่ทรงพลังผ่าน Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [การปรับแต่งโมเดลภาษาขนาดเล็กสำหรับฟังก์ชันเรียกใช้งานในรถยนต์](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) การปรับแต่ง PHI-3 สำหรับการตอบคำถามแบบตัวเลือกหลายข้อ: ระเบียบวิธี ผลลัพธ์ และความท้าทาย](https://arxiv.org/abs/2501.01588) + - [รายงานทางเทคนิค Phi-4-Mini: โมเดลภาษามัลติโมเดลขนาดกะทัดรัดแต่ทรงพลังผ่าน Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [การเพิ่มประสิทธิภาพโมเดลภาษาเล็กสำหรับการเรียกใช้ฟังก์ชันในยานพาหนะ](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) การปรับแต่ง PHI-3 สำหรับการตอบคำถามแบบเลือกตอบหลายตัวเลือก: วิธีการ ผลลัพธ์ และความท้าทาย](https://arxiv.org/abs/2501.01588) - [รายงานทางเทคนิค Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [รายงานทางเทคนิค Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) @@ -226,28 +229,28 @@ Phi ปัจจุบันเป็นโมเดลภาษาขนาด ### Phi บน Microsoft Foundry -คุณสามารถเรียนรู้วิธีใช้ Microsoft Phi และวิธีสร้างโซลูชันแบบ E2E ในอุปกรณ์ฮาร์ดแวร์ที่แตกต่างกันของคุณ เพื่อสัมผัสประสบการณ์ Phi ด้วยตัวเอง ให้เริ่มต้นด้วยการลองเล่นกับโมเดลและปรับแต่ง Phi สำหรับสถานการณ์ของคุณโดยใช้ [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) คุณสามารถเรียนรู้เพิ่มเติมได้ที่ การเริ่มต้นใช้งานกับ [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +คุณสามารถเรียนรู้การใช้ Microsoft Phi และวิธีสร้างโซลูชัน E2E บนอุปกรณ์ฮาร์ดแวร์ต่างๆ ของคุณ เพื่อสัมผัสประสบการณ์ Phi ด้วยตัวเอง เริ่มต้นได้โดยทดลองเล่นกับโมเดลและปรับแต่ง Phi ให้เหมาะกับสถานการณ์ของคุณโดยใช้ [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) คุณสามารถเรียนรู้เพิ่มเติมได้ที่ การเริ่มต้นใช้งาน [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**สนามเล่น** -แต่ละโมเดลมีสนามเล่นเฉพาะเพื่อทดสอบโมเดล [Azure AI Playground](https://aka.ms/try-phi3) +**สนามทดลอง** +โมเดลแต่ละตัวมีสนามทดลองเฉพาะสำหรับทดสอบโมเดล [Azure AI Playground](https://aka.ms/try-phi3). ### Phi บน GitHub Models -คุณสามารถเรียนรู้วิธีใช้ Microsoft Phi และวิธีสร้างโซลูชันแบบ E2E ในอุปกรณ์ฮาร์ดแวร์ของคุณต่าง ๆ เพื่อสัมผัสประสบการณ์ Phi ด้วยตัวเอง ให้เริ่มต้นด้วยการเล่นกับโมเดลและปรับแต่ง Phi สำหรับสถานการณ์ของคุณโดยใช้ [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) คุณสามารถเรียนรู้เพิ่มเติมได้ที่ การเริ่มต้นใช้งานกับ [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +คุณสามารถเรียนรู้การใช้ Microsoft Phi และวิธีสร้างโซลูชัน E2E บนอุปกรณ์ฮาร์ดแวร์ต่างๆ ของคุณ เพื่อสัมผัสประสบการณ์ Phi ด้วยตัวเอง เริ่มต้นได้โดยทดลองเล่นกับโมเดลและปรับแต่ง Phi ให้เหมาะกับสถานการณ์ของคุณโดยใช้ [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) คุณสามารถเรียนรู้เพิ่มเติมได้ที่ การเริ่มต้นใช้งาน [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**สนามเล่น** -แต่ละโมเดลมี [สนามเล่นเพื่อทดสอบโมเดล](/md/02.QuickStart/GitHubModel_QuickStart.md) +**สนามทดลอง** +โมเดลแต่ละตัวมี [สนามทดลองสำหรับทดสอบโมเดล](/md/02.QuickStart/GitHubModel_QuickStart.md) โดยเฉพาะ ### Phi บน Hugging Face -คุณยังสามารถค้นหาโมเดลบน [Hugging Face](https://huggingface.co/microsoft) +คุณยังสามารถหาโมเดลบน [Hugging Face](https://huggingface.co/microsoft) -**สนามเล่น** - [สนามเล่น Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**สนามทดลอง** + [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) - ## 🎒 หลักสูตรอื่นๆ + ## 🎒 คอร์สอื่นๆ -ทีมของเราผลิตหลักสูตรอื่นๆ! ลองดู: +ทีมงานของเราผลิตคอร์สอื่นๆ ด้วย! ตรวจสอบได้ที่: ### LangChain @@ -264,46 +267,49 @@ Phi ปัจจุบันเป็นโมเดลภาษาขนาด --- -### ชุดการเรียนรู้ Generative AI -[![Generative AI สำหรับผู้เริ่มต้น](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### ชุดวิชาปัญญาประดิษฐ์เชิงสร้างสรรค์ +[![ปัญญาประดิษฐ์เชิงสร้างสรรค์สำหรับผู้เริ่มต้น](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![ปัญญาประดิษฐ์เชิงสร้างสรรค์ (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![ปัญญาประดิษฐ์เชิงสร้างสรรค์ (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![ปัญญาประดิษฐ์เชิงสร้างสรรค์ (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- -### การเรียนรู้แกนหลัก -[![ML สำหรับผู้เริ่มต้น](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +### การเรียนรู้พื้นฐาน +[![Machine Learning สำหรับผู้เริ่มต้น](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Data Science สำหรับผู้เริ่มต้น](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![AI สำหรับผู้เริ่มต้น](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity สำหรับผู้เริ่มต้น](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![ความปลอดภัยไซเบอร์สำหรับผู้เริ่มต้น](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Web Dev สำหรับผู้เริ่มต้น](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![IoT สำหรับผู้เริ่มต้น](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![XR Development สำหรับผู้เริ่มต้น](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### ชุด Copilot -[![Copilot สำหรับการเขียนโปรแกรมร่วมกับ AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +### ชุดวิชา Copilot +[![Copilot สำหรับการเขียนโปรแกรมคู่กับ AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot สำหรับ C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## ปัญญาประดิษฐ์ที่รับผิดชอบ +## AI ที่รับผิดชอบ + +Microsoft มุ่งมั่นที่จะช่วยลูกค้าของเราใช้ผลิตภัณฑ์ AI อย่างรับผิดชอบ แบ่งปันสิ่งที่เราเรียนรู้ และสร้างความร่วมมือที่มีพื้นฐานจากความไว้วางใจผ่านเครื่องมือต่างๆ เช่น หมายเหตุความโปร่งใสและการประเมินผลกระทบ แหล่งข้อมูลหลายอย่างเหล่านี้สามารถพบได้ที่ [https://aka.ms/RAI](https://aka.ms/RAI) +แนวทางของ Microsoft ในการใช้ AI อย่างรับผิดชอบตั้งอยู่บนหลักการ AI ของเราคือ ความยุติธรรม ความน่าเชื่อถือและความปลอดภัย ความเป็นส่วนตัวและความปลอดภัย ความครอบคลุม ความโปร่งใส และความรับผิดชอบ + +โมเดลภาษา, รูปภาพ และเสียงขนาดใหญ่ - เช่นที่ใช้ในตัวอย่างนี้ - อาจมีพฤติกรรมที่ไม่เป็นธรรม ไม่เชื่อถือได้ หรือไม่เหมาะสม ซึ่งอาจก่อให้เกิดอันตรายได้ โปรดดู [หมายเหตุความโปร่งใสของบริการ Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) เพื่อรับทราบเกี่ยวกับความเสี่ยงและข้อจำกัด -ไมโครซอฟท์มุ่งมั่นที่จะช่วยลูกค้าใช้ผลิตภัณฑ์ AI ของเราอย่างรับผิดชอบ แบ่งปันบทเรียนของเรา และสร้างพันธมิตรที่มีพื้นฐานความไว้วางใจผ่านเครื่องมือต่าง ๆ เช่น Transparency Notes และ Impact Assessments แหล่งข้อมูลเหล่านี้หลายรายการสามารถพบได้ที่ [https://aka.ms/RAI](https://aka.ms/RAI) -แนวทางของไมโครซอฟท์สำหรับ AI ที่รับผิดชอบมีรากฐานจากหลักการ AI ของเราเกี่ยวกับความเป็นธรรม ความน่าเชื่อถือและความปลอดภัย ความเป็นส่วนตัวและความปลอดภัย ความครอบคลุม ความโปร่งใส และความรับผิดชอบ -โมเดลภาษา ข้อความ และเสียงขนาดใหญ่ - เช่น โมเดลที่ใช้ในตัวอย่างนี้ - อาจมีพฤติกรรมที่ไม่เป็นธรรม ไม่น่าเชื่อถือ หรือน่าสะเทือนใจจนก่อให้เกิดความเสียหาย โปรดศึกษาเอกสาร [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) เพื่อรับทราบเกี่ยวกับความเสี่ยงและข้อจำกัดต่าง ๆ -แนวทางที่แนะนำในการลดความเสี่ยงเหล่านี้คือการรวมระบบความปลอดภัยไว้ในสถาปัตยกรรมของคุณที่สามารถตรวจจับและป้องกันพฤติกรรมที่เป็นอันตรายได้ [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ให้ชั้นการป้องกันอิสระที่สามารถตรวจจับเนื้อหาที่เป็นอันตรายที่สร้างขึ้นโดยผู้ใช้และ AI ในแอปพลิเคชันและบริการต่างๆ Azure AI Content Safety รวมถึง API ข้อความและภาพที่ช่วยให้คุณตรวจจับเนื้อหาที่เป็นอันตราย ใน Microsoft Foundry บริการ Content Safety ช่วยให้คุณดู สำรวจ และลองใช้ตัวอย่างโค้ดสำหรับการตรวจจับเนื้อหาที่เป็นอันตรายในหลายรูปแบบ เอกสาร [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ต่อไปนี้จะแนะนำคุณผ่านการส่งคำขอไปยังบริการนี้ +วิธีที่แนะนำในการลดความเสี่ยงเหล่านี้คือการรวมระบบความปลอดภัยในสถาปัตยกรรมของคุณที่สามารถตรวจจับและป้องกันพฤติกรรมที่เป็นอันตรายได้ [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ให้ชั้นการป้องกันอิสระ สามารถตรวจจับเนื้อหาที่เป็นอันตรายซึ่งผู้ใช้และ AI สร้างขึ้นในแอปพลิเคชันและบริการ Azure AI Content Safety รวมถึง API ข้อความและภาพที่ช่วยให้คุณตรวจจับเนื้อหาที่เป็นอันตราย ภายใน Microsoft Foundry บริการ Content Safety ช่วยให้คุณดู สำรวจ และทดลองโค้ดตัวอย่างสำหรับการตรวจจับเนื้อหาที่เป็นอันตรายในหลายรูปแบบ เอกสาร [quickstart](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ต่อไปนี้จะแนะนำคุณในการส่งคำขอต่อบริการ -อีกด้านหนึ่งที่ต้องพิจารณาคือประสิทธิภาพโดยรวมของแอปพลิเคชัน ด้วยแอปพลิเคชันแบบมัลติ-โหมดและมัลติ-โมเดล เราถือว่าประสิทธิภาพหมายถึงระบบทำงานตามที่คุณและผู้ใช้ของคุณคาดหวัง รวมถึงการไม่สร้างผลลัพธ์ที่เป็นอันตราย สิ่งสำคัญคือต้องประเมินประสิทธิภาพของแอปพลิเคชันโดยรวมของคุณโดยใช้ [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) นอกจากนี้คุณยังสามารถสร้างและประเมินผลด้วย [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) +อีกแง่มุมที่ต้องพิจารณาคือประสิทธิภาพโดยรวมของแอปพลิเคชัน สำหรับแอปพลิเคชันแบบมัลติโมดัลและหลายโมเดล เราพิจารณาว่าประสิทธิภาพหมายถึงระบบทำงานตามที่คุณและผู้ใช้ของคุณคาดหวัง รวมถึงไม่สร้างผลลัพธ์ที่เป็นอันตราย การประเมินประสิทธิภาพของแอปพลิเคชันโดยรวมของคุณด้วย [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) เป็นสิ่งสำคัญ คุณยังสามารถสร้างและประเมินด้วย [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) -คุณสามารถประเมินแอปพลิเคชัน AI ของคุณในสภาพแวดล้อมการพัฒนาด้วย [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) โดยใช้ชุดข้อมูลทดสอบหรือเป้าหมาย การสร้าง AI ของแอปพลิเคชันของคุณจะถูกวัดเชิงปริมาณด้วยตัวประเมินในตัวหรือผู้ประเมินที่คุณกำหนดเอง ในการเริ่มต้นใช้งาน azure ai evaluation sdk เพื่อประเมินระบบของคุณ คุณสามารถติดตาม [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) เมื่อคุณดำเนินการรันการประเมินแล้ว คุณสามารถ [visualize the results in Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) +คุณสามารถประเมินแอป AI ของคุณในสภาพแวดล้อมการพัฒนาด้วย [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) ไม่ว่าจะเป็นชุดข้อมูลทดสอบหรือเป้าหมาย ผลงานสร้าง AI ของคุณจะถูกวัดเชิงปริมาณด้วยตัวประเมินในตัวหรือ custom evaluators ที่คุณเลือก ในการเริ่มต้นกับ azure ai evaluation sdk เพื่อประเมินระบบของคุณ คุณสามารถติดตาม [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) เมื่อคุณทำการประเมินแล้ว คุณสามารถ [แสดงผลลัพธ์ใน Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results) ได้ ## เครื่องหมายการค้า -โปรเจกต์นี้อาจมีเครื่องหมายการค้าหรือโลโก้สำหรับโปรเจกต์ ผลิตภัณฑ์ หรือบริการ ใช้งานที่ได้รับอนุญาตของเครื่องหมายการค้าหรือโลโก้ของ Microsoft ต้องเป็นไปตามและปฏิบัติตาม [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) การใช้เครื่องหมายการค้าหรือโลโก้ของ Microsoft ในเวอร์ชันที่ดัดแปลงของโปรเจกต์นี้ต้องไม่ก่อให้เกิดความสับสนหรือสื่อว่ามีการสนับสนุนจาก Microsoft การใช้เครื่องหมายการค้าหรือโลโก้ของบุคคลที่สามใดๆ อยู่ภายใต้นโยบายของบุคคลที่สามเหล่านั้น +โปรเจกต์นี้อาจมีเครื่องหมายการค้าหรือโลโก้สำหรับโปรเจกต์ ผลิตภัณฑ์ หรือบริการ การใช้เครื่องหมายการค้าหรือโลโก้ของ Microsoft อย่างถูกต้องต้องเป็นไปตาม [นโยบายเครื่องหมายการค้า & แบรนด์ของ Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) +การใช้เครื่องหมายการค้าหรือโลโก้ของ Microsoft ในเวอร์ชันที่ดัดแปลงของโปรเจกต์นี้ต้องไม่ก่อให้เกิดความสับสนหรือนัยว่ามีการสนับสนุนจาก Microsoft การใช้เครื่องหมายการค้าหรือโลโก้ของบุคคลที่สามให้เป็นไปตามนโยบายของบุคคลที่สามเหล่านั้น ## ขอความช่วยเหลือ @@ -311,13 +317,13 @@ Phi ปัจจุบันเป็นโมเดลภาษาขนาด [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -หากคุณมีข้อเสนอแนะเกี่ยวกับผลิตภัณฑ์หรือข้อผิดพลาดระหว่างการสร้างเยี่ยมชมได้ที่: +หากคุณมีคำติชมผลิตภัณฑ์หรือข้อผิดพลาดระหว่างการสร้าง เยี่ยมชมได้ที่: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**ข้อจำกัดความรับผิดชอบ**: -เอกสารนี้ได้รับการแปลโดยใช้บริการแปลภาษา AI [Co-op Translator](https://github.com/Azure/co-op-translator) แม้เราจะพยายามให้เกิดความถูกต้องแต่กรุณาทราบว่าการแปลอัตโนมัติอาจมีข้อผิดพลาดหรือความไม่ถูกต้อง เอกสารต้นฉบับในภาษาต้นทางควรถือเป็นแหล่งข้อมูลที่เชื่อถือได้ สำหรับข้อมูลที่สำคัญ ควรใช้การแปลโดยผู้เชี่ยวชาญด้านมนุษย์ เราไม่รับผิดชอบต่อความเข้าใจผิดหรือการตีความผิดที่เกิดจากการใช้การแปลนี้ +**ปฏิเสธความรับผิดชอบ**: +เอกสารนี้ได้รับการแปลโดยใช้บริการแปลภาษา AI [Co-op Translator](https://github.com/Azure/co-op-translator) ขณะที่เราพยายามให้ความถูกต้อง โปรดทราบว่าการแปลโดยอัตโนมัติอาจมีข้อผิดพลาดหรือความไม่ถูกต้อง เอกสารต้นฉบับในภาษาต้นทางควรถูกพิจารณาเป็นแหล่งข้อมูลที่เชื่อถือได้ สำหรับข้อมูลที่สำคัญ แนะนำให้ใช้การแปลโดยมนุษย์มืออาชีพ เราไม่รับผิดชอบต่อความเข้าใจผิดหรือการตีความที่ผิดพลาดที่เกิดขึ้นจากการใช้การแปลนี้ \ No newline at end of file diff --git a/translations/th/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/th/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..935799a81 --- /dev/null +++ b/translations/th/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrid RAG กับ Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **ผู้เขียน:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **เน้น:** Small Language Models (SLMs), SQLite FTS5 BM25, Dense Embeddings, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. แรงจูงใจ: ปัญหาการค้นหาคีย์เวิร์ดใน Local SLMs\n", + "สถาปัตยกรรม RAG มาตรฐานที่พึ่งพาเพียงการฝังเวกเตอร์หนาแน่นมักล้มเหลวในการดึงโทเค็นตัวเลขที่แม่นยำ (เช่น `2,340,000 TL`, รหัสสัญญา, หมายเลขบัญชี) \n", + "ในทางกลับกัน การค้นหาคำแบบกระจัดกระจาย (BM25) มักพลาดคำพ้องความหมายและคำถามที่สลับประโยค\n", + "\n", + "คู่มือนี้แสดงวิธีการสร้าง **เอนจิน Hybrid Retrieval ความเร็วสูงในเมมโมรี** ที่รวม:\n", + "1. **เวกเตอร์หนาแน่น** (Cosine Similarity)\n", + "2. **การค้นหาคำแบบกระจัดกระจาย** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **การสร้างการอ้างอิงที่ยืนยันแหล่งที่มา (`[1]`, `[2]`)** ด้วย Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. สคีมา SQLite คู่ (เวกเตอร์หนาแน่น + ตารางเสมือน FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. ตัวอย่างการรับเข้าและเกณฑ์มาตรฐานการดำเนินการ\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. การสร้างคำสั่งแบบมีหลักฐานสำหรับ Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**ปฏิเสธความรับผิดชอบ**:\nเอกสารนี้ได้รับการแปลโดยใช้บริการแปลภาษา AI [Co-op Translator](https://github.com/Azure/co-op-translator) ขณะที่เราพยายามให้ความถูกต้อง โปรดทราบว่าการแปลโดยอัตโนมัติอาจมีข้อผิดพลาดหรือความไม่ถูกต้อง เอกสารต้นฉบับในภาษาต้นทางควรถูกพิจารณาเป็นแหล่งข้อมูลที่เชื่อถือได้ สำหรับข้อมูลที่สำคัญ แนะนำให้ใช้การแปลโดยมนุษย์มืออาชีพ เราไม่รับผิดชอบต่อความเข้าใจผิดหรือการตีความที่ผิดพลาดที่เกิดขึ้นจากการใช้การแปลนี้\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/tl/.co-op-translator.json b/translations/tl/.co-op-translator.json index 1e84e31b9..74864b6fb 100644 --- a/translations/tl/.co-op-translator.json +++ b/translations/tl/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "tl" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:42:10+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T17:31:35+00:00", "source_file": "README.md", "language_code": "tl" }, @@ -28,7 +28,7 @@ "original_hash": "3bb9f5c926673593287eddc3741226cb", "source_file": "md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md", "language_code": "tl", - "failure_date": "2026-08-26T15:33:23+00:00", + "failure_date": "2026-08-26T17:32:21+00:00", "status": "failed", "error_type": "TranslationIncompleteError", "error_message": "Markdown translation remained incomplete for chunk 1 of 'UsingORTGenAIQuantifyingPhi.md', and the chunk could not be split further.", @@ -66,6 +66,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "tl" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T17:33:16+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "tl" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:44+00:00", diff --git a/translations/tl/README.md b/translations/tl/README.md index 6528756d3..18c177420 100644 --- a/translations/tl/README.md +++ b/translations/tl/README.md @@ -1,42 +1,42 @@ -# Phi Cookbook: Mga Hands-On na Halimbawa gamit ang mga Phi Models ng Microsoft +# Phi Cookbook: Mga Hands-On na Halimbawa gamit ang Phi Models ng Microsoft -[![Open and use the samples in GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Open in Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![Buksan at gamitin ang mga sample sa GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![Buksan sa Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub contributors](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub issues](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub pull-requests](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Mga kontribyutor ng GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![Mga isyu ng GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![Mga pull-request ng GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![Malugod ang mga PR](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub watchers](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub stars](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![Mga tagamasid sa GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![Mga fork ng GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![Mga bituin sa GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Ang Phi ay isang serye ng bukas na pinagmulan na mga AI modelong binuo ng Microsoft. +Ang Phi ay isang serye ng mga open source AI models na binuo ng Microsoft. -Ang Phi ay kasalukuyang ang pinaka-makapangyarihan at cost-effective na maliit na language model (SLM), na may napakahusay na mga benchmark sa multi-wika, pag-iisip, text/chat generation, coding, mga larawan, audio at iba pang mga sitwasyon. +Ang Phi ay kasalukuyang pinakamakapangyarihan at cost-effective na maliit na language model (SLM), na may napakagandang benchmarks sa multi-wika, pangangatwiran, paggawa ng teksto/chat, pag-coding, mga larawan, audio at iba pang mga senaryo. -Maaari mong i-deploy ang Phi sa cloud o sa mga edge device, at madali kang makakagawa ng mga generative AI application na may limitadong computing power. +Maaari mong i-deploy ang Phi sa cloud o sa mga edge devices, at madali kang makakagawa ng mga generative AI applications gamit ang limitadong computing power. -Sundin ang mga hakbang na ito upang makapagsimula gamit ang mga resource na ito : -1. **I-fork ang Repository**: Pindutin ang [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +Sundin ang mga hakbang na ito upang makapagsimula gamit ang mga resources na ito: +1. **I-fork ang Repository**: I-click ang [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **I-clone ang Repository**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Sumali sa Microsoft AI Discord Community at makipagkita sa mga eksperto at kapwa mga developer**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Sumali sa Microsoft AI Discord Community at makipagkilala sa mga eksperto at kapwa developer**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/tl/cover.eb18d1b9605d754b.webp) -### 🌐 Suporta sa Maramihang Wika +### 🌐 Suporta sa Maraming Wika -#### Sinusuportahan sa pamamagitan ng GitHub Action (Automated at Palaging Napapanahon) +#### Sinusuportahan sa pamamagitan ng GitHub Action (Automated at Palaging Naka-update) [Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](./README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **Mas gusto mo bang Mag-clone nang Lokal?** +> **Mas gusto mong i-clone Lokal?** > -> Ang repositoryong ito ay may higit sa 50 na pagsasalin ng wika na malaki ang pinapataas sa laki ng i-download. Upang mag-clone nang walang mga pagsasalin, gamitin ang sparse checkout: +> Kasama sa repository na ito ang mahigit 50 na mga pagsasalin ng wika na nagpapalaki nang malaki sa laki ng pag-download. Upang mag-clone nang walang mga pagsasalin, gamitin ang sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,20 +52,20 @@ Sundin ang mga hakbang na ito upang makapagsimula gamit ang mga resource na ito > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Ito ay magbibigay sa iyo ng lahat ng kailangan mo upang makumpleto ang kurso nang mas mabilis ang pag-download. +> Bibigyan ka nito ng lahat ng kailangan mo upang matapos ang kurso nang mas mabilis ang pag-download. -## Talaan ng Nilalaman +## Tala ng mga Nilalaman - Panimula - - [Maligayang pagdating sa Pamilyang Phi](./md/01.Introduction/01/01.PhiFamily.md) - - [Pagsasaayos ng iyong kapaligiran](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [Maligayang pagdating sa Phi Family](./md/01.Introduction/01/01.PhiFamily.md) + - [Pag-set up ng iyong kapaligiran](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Pag-unawa sa mga Pangunahing Teknolohiya](./md/01.Introduction/01/01.Understandingtech.md) - - [AI Kaligtasan para sa mga Phi Models](./md/01.Introduction/01/01.AISafety.md) - - [Suporta ng Hardware ng Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi Models at Pagkakaroon sa iba't ibang platform](./md/01.Introduction/01/01.Edgeandcloud.md) + - [AI Safety para sa Phi Models](./md/01.Introduction/01/01.AISafety.md) + - [Suporta sa Hardware ng Phi](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Mga Phi Model at Availability sa iba't ibang platform](./md/01.Introduction/01/01.Edgeandcloud.md) - [Paggamit ng Guidance-ai at Phi](./md/01.Introduction/01/01.Guidance.md) - - [GitHub Marketplace Models](https://github.com/marketplace/models) + - [Mga Modelo sa GitHub Marketplace](https://github.com/marketplace/models) - [Azure AI Model Catalog](https://ai.azure.com) - Pagsusuri ng Phi sa iba't ibang kapaligiran @@ -77,7 +77,7 @@ Sundin ang mga hakbang na ito upang makapagsimula gamit ang mga resource na ito - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Pagsusuri ng Pamilyang Phi +- Pagsusuri sa Phi Family - [Pagsusuri ng Phi sa iOS](./md/01.Introduction/03/iOS_Inference.md) - [Pagsusuri ng Phi sa Android](./md/01.Introduction/03/Android_Inference.md) - [Pagsusuri ng Phi sa Jetson](./md/01.Introduction/03/Jetson_Inference.md) @@ -88,112 +88,114 @@ Sundin ang mga hakbang na ito upang makapagsimula gamit ang mga resource na ito - [Pagsusuri ng Phi gamit ang Rust](./md/01.Introduction/03/Rust_Inference.md) - [Pagsusuri ng Phi--Vision sa Lokal](./md/01.Introduction/03/Vision_Inference.md) - [Pagsusuri ng Phi gamit ang Kaito AKS, Azure Containers (opisyal na suporta)](./md/01.Introduction/03/Kaito_Inference.md) -- [Pag-quantify sa Pamilyang Phi](./md/01.Introduction/04/QuantifyingPhi.md) +- [Pag-quantify ng Phi Family](./md/01.Introduction/04/QuantifyingPhi.md) - [Pag-quantize ng Phi-3.5 / 4 gamit ang llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - [Pag-quantize ng Phi-3.5 / 4 gamit ang Generative AI extensions para sa onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [Pag-quantize ng Phi-3.5 / 4 gamit ang Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - [Pag-quantize ng Phi-3.5 / 4 gamit ang Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Pagsusuri ng Phi - - [Responsible AI](./md/01.Introduction/05/ResponsibleAI.md) + - [Responsableng AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry para sa Pagsusuri](./md/01.Introduction/05/AIFoundry.md) - [Paggamit ng Promptflow para sa Pagsusuri](./md/01.Introduction/05/Promptflow.md) - RAG gamit ang Azure AI Search - [Paano gamitin ang Phi-4-mini at Phi-4-multimodal (RAG) gamit ang Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud Lokal na Hybrid RAG gamit ang SQLite FTS5 at phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Mga halimbawa ng pagbuo ng Phi application - - Mga Text at Chat Application - - Phi-4 Samples +- Mga sample ng pagpapaunlad ng Phi application + - Mga Text at Chat na Aplikasyon + - Mga Sample ng Phi-4 - [📓] [Makipag-chat gamit ang Phi-4-mini ONNX Model](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Makipag-chat gamit ang Phi-4 lokal na ONNX Model .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - [Chat .NET Console App gamit ang Phi-4 ONNX gamit ang Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 Samples + + - Phi-3 / 3.5 Mga Sample - [Lokal na Chatbot sa browser gamit ang Phi3, ONNX Runtime Web at WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - Interactive Phi-3-mini and OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Paggawa ng wrapper at paggamit ng Phi-3 kasama ang MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Pag-optimize ng Modelo - Paano i-optimize ang Phi-3-min na modelo para sa ONNX Runtime Web gamit ang Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Multi Model - Interactive Phi-3-mini at OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Pagbuo ng wrapper at paggamit ng Phi-3 gamit ang MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Pag-optimize ng Modelo - Paano i-optimize ang Phi-3-min model para sa ONNX Runtime Web gamit ang Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [WinUI3 App gamit ang Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) -[WinUI3 Multi Model AI Powered Notes App Sample](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Fine-tune at Integrate ang custom Phi-3 models gamit ang Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Fine-tune at Integrate ang custom Phi-3 models gamit ang Prompt flow sa Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Suriin ang Fine-tuned na Phi-3 / Phi-3.5 Model sa Microsoft Foundry na nakatuon sa Mga Prinsipyo ng Responsable AI ng Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct na halimbawa ng prediksyon ng wika (Tsino/Ingles)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Fine-tune at Integrate ang custom na Phi-3 models gamit ang Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Fine-tune at Integrate ang custom na Phi-3 models gamit ang Prompt flow sa Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Pagsusuri ng Fine-tuned Phi-3 / Phi-3.5 Model sa Microsoft Foundry na Nakatuon sa Responsableng AI Principles ng Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct halimbawa ng prediksyon ng wika (Chinese/English)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Paggamit ng Windows GPU upang gumawa ng Prompt flow na solusyon gamit ang Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Paggamit ng Windows GPU para gumawa ng Prompt flow solution gamit ang Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Paggamit ng Microsoft Phi-3.5 tflite para gumawa ng Android app](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET Halimbawa gamit ang local ONNX Phi-3 model gamit ang Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Q&A .NET Halimbawa gamit ang lokal na ONNX Phi-3 model gamit ang Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - [Console chat .NET app gamit ang Semantic Kernel at Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK Code Based Samples - - Phi-4 Samples - - [📓] [Gumawa ng project code gamit ang Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 Samples + - Azure AI Inference SDK Code Based Mga Sample + - Phi-4 Mga Sample + - [📓] [Generate ng project code gamit ang Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Phi-3 / 3.5 Mga Sample - [Bumuo ng sarili mong Visual Studio Code GitHub Copilot Chat gamit ang Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Gumawa ng sarili mong Visual Studio Code Chat Copilot Agent gamit ang Phi-3.5 mula sa mga GitHub Models](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Gumawa ng sarili mong Visual Studio Code Chat Copilot Agent gamit ang Phi-3.5 mula sa GitHub Models](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - Advanced Reasoning Samples - - Phi-4 Samples - - [📓] [Phi-4-mini-reasoning o Phi-4-reasoning Samples](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Fine-tuning ng Phi-4-mini-reasoning gamit ang Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Fine-tuning ng Phi-4-mini-reasoning gamit ang Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - Advanced Reasoning Mga Sample + - Phi-4 Mga Sample + - [📓] [Phi-4-mini-reasoning o Phi-4-reasoning Mga Sample](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Fine-tuning Phi-4-mini-reasoning gamit ang Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Fine-tuning Phi-4-mini-reasoning gamit ang Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-reasoning gamit ang GitHub Models](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-reasoning gamit ang Microsoft Foundry Models](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Demos - [Phi-4-mini demos na naka-host sa Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal demos na naka-host sa Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Vision Samples - - Phi-4 Samples - - [📓] [Gamitin ang Phi-4-multimodal upang basahin ang mga larawan at gumawa ng code](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 Samples - - [📓][Phi-3-vision-Larawan teksto to teksto](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-4-multimodal demos na naka-host sa Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Vision Mga Sample + - Phi-4 Mga Sample + - [📓] [Gamitin ang Phi-4-multimodal para magbasa ng mga larawan at gumawa ng code](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-3 / 3.5 Mga Sample + - [📓][Phi-3-vision-Image text to text](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [DEMO: Phi-3 Recycling](https://github.com/jennifermarsman/PhiRecycling/) - [Phi-3-vision - Visual language assistant - gamit ang Phi3-Vision at OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision multi-frame o multi-image na halimbawa](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [📓][Phi-3.5 Vision multi-frame o multi-image sample](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - [Phi-3 Vision Lokal na ONNX Model gamit ang Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - [Menu based Phi-3 Vision Lokal na ONNX Model gamit ang Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - Reasoning-Vision Samples + - Reasoning-Vision Mga Sample - Phi-4-Reasoning-Vision-15B - - [📓] [Paggamit ng Phi-4-Reasoning-Vision-15B upang makita ang jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Paggamit ng Phi-4-Reasoning-Vision-15B para sa math](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Paggamit ng Phi-4-Reasoning-Vision-15B upang makita ang UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [Paggamit ng Phi-4-Reasoning-Vision-15B upang matukoy ang jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Paggamit ng Phi-4-Reasoning-Vision-15B sa math](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Paggamit ng Phi-4-Reasoning-Vision-15B upang matukoy ang UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - Math Samples - - Phi-4-Mini-Flash-Reasoning-Instruct Samples [Math Demo gamit ang Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Math Mga Sample + - Phi-4-Mini-Flash-Reasoning-Instruct Mga Sample [Math Demo gamit ang Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - Audio Samples - - Phi-4 Samples - - [📓] [Pagkuhà ng audio transcripts gamit ang Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - Audio Mga Sample + - Phi-4 Mga Sample + - [📓] [Pagkuha ng audio transcripts gamit ang Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Phi-4-multimodal Audio Sample](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Phi-4-multimodal Speech Translation Sample](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - [.NET console application gamit ang Phi-4-multimodal Audio para suriin ang isang audio file at gumawa ng transcript](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - MOE Samples - - Phi-3 / 3.5 Samples + - MOE Mga Sample + - Phi-3 / 3.5 Mga Sample - [📓] [Phi-3.5 Mixture of Experts Models (MoEs) Social Media Sample](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Paggawa ng Retrieval-Augmented Generation (RAG) Pipeline gamit ang NVIDIA NIM Phi-3 MOE, Azure AI Search, at LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Pagbuo ng Retrieval-Augmented Generation (RAG) Pipeline gamit ang NVIDIA NIM Phi-3 MOE, Azure AI Search, at LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Function Calling Samples - - Phi-4 Samples 🆕 - - [📓] [Paggamit ng Function Calling kasama ang Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Paggamit ng Function Calling upang gumawa ng multi-agents gamit ang Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Paggamit ng Function Calling kasama ang Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Paggamit ng Function Calling kasama ang ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Multimodal Mixing Samples - - Phi-4 Samples 🆕 - - [📓] [Paggamit ng Phi-4-multimodal bilang isang Technology journalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - Function Calling Mga Sample + - Phi-4 Mga Sample 🆕 + - [📓] [Paggamit ng Function Calling Sa Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Paggamit ng Function Calling para gumawa ng multi-agents Gamit ang Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Paggamit ng Function Calling gamit ang Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Paggamit ng Function Calling gamit ang ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Multimodal Mixing Mga Sample + - Phi-4 Mga Sample 🆕 + - [📓] [Paggamit ng Phi-4-multimodal bilang Technology journalist](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [.NET console application gamit ang Phi-4-multimodal para suriin ang mga larawan](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Fine-tuning Phi Samples - - [Mga Scenario ng Fine-tuning](./md/03.FineTuning/FineTuning_Scenarios.md) +- Fine-tuning Phi Mga Sample + - [Fine-tuning Scenarios](./md/03.FineTuning/FineTuning_Scenarios.md) - [Fine-tuning vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Fine-tuning Hayaan ang Phi-3 maging isang eksperto sa industriya](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Fine-tuning Hayaan ang Phi-3 maging isang industry expert](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - [Fine-tuning Phi-3 gamit ang AI Toolkit para sa VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [Fine-tuning Phi-3 gamit ang Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - [Fine-tuning Phi-3 gamit ang Lora](./md/03.FineTuning/FineTuning_Lora.md) @@ -203,83 +205,84 @@ Sundin ang mga hakbang na ito upang makapagsimula gamit ang mga resource na ito - [Fine-tuning gamit ang Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Fine-tuning gamit ang Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) - [Fine-tuning Phi-3-vision gamit ang Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [Fine-tuning Phi-3 gamit ang Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - [Fine-tuning Phi-3-vision (opisyal na suporta)](./md/03.FineTuning/FineTuning_Vision.md) - [Fine-Tuning Phi-3 gamit ang Kaito AKS, Azure Containers (opisyal na Suporta)](./md/03.FineTuning/FineTuning_Kaito.md) - [Fine-Tuning Phi-3 at 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Hands on Lab - - [Pagsusuri ng mga pinaka-advanced na modelo: LLMs, SLMs, lokal na pag-unlad at iba pa](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Pagsaliksik ng mga pinaka-advanced na modelo: LLMs, SLMs, lokal na pag-develop at iba pa](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - [Pagbubukas ng Potensyal ng NLP: Fine-Tuning gamit ang Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Mga Papel sa Pananaliksik at Publikasyon ng Akademya - - [Textbooks Are All You Need II: phi-1.5 technical report](https://arxiv.org/abs/2309.05463) - - [Phi-3 Technical Report: Isang Mataas na Kakayahang Modelong Pangwika nang Lokal sa Iyong Telepono](https://arxiv.org/abs/2404.14219) - - [Phi-4 Technical Report](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini Technical Report: Compact ngunit Makapangyarihang Multimodal Language Models sa pamamagitan ng Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Pag-optimize ng Maliliit na Modelo ng Wika para sa In-Vehicle Function-Calling](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Fine-Tuning PHI-3 para sa Multiple-Choice Question Answering: Metodolohiya, Mga Resulta, at Mga Hamon](https://arxiv.org/abs/2501.01588) - - [Phi-4-reasoning Technical Report](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini-reasoning Technical Report](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) +- Mga Papel at Publikasyon sa Akademikong Pananaliksik + - [Mga Aklat ang Tanging Kailangan II: phi-1.5 teknikal na ulat](https://arxiv.org/abs/2309.05463) + - [Teknikal na Ulat ng Phi-3: Isang Lubos na May-Kakayahang Wika ng Modelo Nang Lokal sa Iyong Telepono](https://arxiv.org/abs/2404.14219) + - [Teknikal na Ulat ng Phi-4](https://arxiv.org/abs/2412.08905) + - [Teknikal na Ulat ng Phi-4-Mini: Compact ngunit Makapangyarihang Multimodal Language Models via Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Pag-optimize ng Maliit na Mga Modelo ng Wika para sa In-Vehicle Function-Calling](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Fine-Tuning PHI-3 para sa Multiple-Choice Question Answering: Metodo, Resulta, at Mga Hamon](https://arxiv.org/abs/2501.01588) + - [Teknikal na Ulat ng Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Teknikal na Ulat ng Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Paggamit ng Mga Modelo ng Phi +## Paggamit ng Phi Models ### Phi sa Microsoft Foundry -Matutunan mo kung paano gamitin ang Microsoft Phi at kung paano bumuo ng mga solusyong E2E sa iba't ibang hardware devices mo. Para maranasan ang Phi, simulan sa paglalaro gamit ang mga modelo at i-customize ang Phi para sa iyong mga senaryo gamit ang [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Maaari kang matuto nang higit pa sa Pagsisimula gamit ang [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md). +Maaari kang matuto kung paano gamitin ang Microsoft Phi at kung paano gumawa ng E2E solutions sa iyong iba't ibang hardware devices. Para maranasan ang Phi para sa iyong sarili, magsimula sa pamamagitan ng paglalaro gamit ang mga modelo at pag-customize ng Phi para sa iyong mga senaryo gamit ang [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) maaari kang matuto nang higit pa sa Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) **Playground** -Bawat modelo ay may dedikadong playground para subukan ang modelo sa [Azure AI Playground](https://aka.ms/try-phi3). +Bawat modelo ay may nakalaang playground para subukan ang modelo [Azure AI Playground](https://aka.ms/try-phi3). ### Phi sa GitHub Models -Matutunan mo kung paano gamitin ang Microsoft Phi at kung paano bumuo ng mga solusyong E2E sa iba't ibang hardware devices mo. Para maranasan ang Phi, simulan sa paglalaro gamit ang modelo at i-customize ang Phi para sa iyong mga senaryo gamit ang [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Maaari kang matuto nang higit pa sa Pagsisimula gamit ang [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md). +Maaari kang matuto kung paano gamitin ang Microsoft Phi at kung paano gumawa ng E2E solutions sa iyong iba't ibang hardware devices. Para maranasan ang Phi para sa iyong sarili, magsimula sa pamamagitan ng paglalaro gamit ang modelo at pag-customize ng Phi para sa iyong mga senaryo gamit ang [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) maaari kang matuto nang higit pa sa Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) **Playground** -Bawat modelo ay may dedikadong [playground para subukan ang modelo](/md/02.QuickStart/GitHubModel_QuickStart.md). +Bawat modelo ay may nakalaang [playground para subukan ang modelo](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi sa Hugging Face Maaari mo ring makita ang modelo sa [Hugging Face](https://huggingface.co/microsoft) **Playground** -[Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) + [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) ## 🎒 Iba Pang Mga Kurso -Ang aming koponan ay gumagawa rin ng iba pang mga kurso! Tingnan ang: +Gumagawa ang aming koponan ng iba pang mga kurso! Tingnan: ### LangChain -[![LangChain4j para sa mga Nagsisimula](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js para sa mga Nagsisimula](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain para sa mga Nagsisimula](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain4j para sa mga Baguhan](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js para sa mga Baguhan](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain para sa mga Baguhan](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- ### Azure / Edge / MCP / Agents -[![AZD para sa mga Nagsisimula](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI para sa mga Nagsisimula](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP para sa mga Nagsisimula](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents para sa mga Nagsisimula](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AZD para sa mga Baguhan](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI para sa mga Baguhan](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP para sa mga Baguhan](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents para sa mga Baguhan](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- ### Generative AI Series -[![Generative AI para sa mga Nagsisimula](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative AI para sa mga Baguhan](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) [![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) [![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- -### Pangunahing Pagkatuto -[![ML para sa mga Nagsisimula](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Agham sa Datos para sa mga Nagsisimula](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI para sa mga Nagsisimula](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity para sa mga Nagsisimula](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web Dev para sa mga Nagsisimula](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT para sa mga Nagsisimula](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR Development para sa mga Nagsisimula](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +### Core Learning +[![ML para sa mga Baguhan](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science para sa mga Baguhan](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI para sa mga Baguhan](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersecurity para sa mga Baguhan](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web Dev para sa mga Baguhan](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT para sa mga Baguhan](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR Development para sa mga Baguhan](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- @@ -289,36 +292,38 @@ Ang aming koponan ay gumagawa rin ng iba pang mga kurso! Tingnan ang: [![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Responsable AI +## Responsableng AI + +Nakatuon ang Microsoft na tulungan ang aming mga customer na gamitin ang aming mga produkto ng AI nang responsable, ibahagi ang aming mga natutunan, at bumuo ng mga pagtutulungan na nakabase sa tiwala sa pamamagitan ng mga tool tulad ng Transparency Notes at Impact Assessments. Marami sa mga resources na ito ay matatagpuan sa [https://aka.ms/RAI](https://aka.ms/RAI). +Nakabatay ang lapit ng Microsoft sa responsableng AI sa aming mga prinsipyo ng AI na katarungan, pagiging maaasahan at kaligtasan, privacy at seguridad, inklusibidad, transparency, at pananagutan. + +Ang malalawak na natural na mga modelo ng wika, imahe, at pagsasalita - tulad ng mga ginamit sa sample na ito - ay maaaring kumilos sa mga paraan na hindi patas, hindi maaasahan, o nakakasakit, na nagdudulot ng pinsala. Mangyaring sumangguni sa [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) upang maging maalam tungkol sa mga panganib at limitasyon. -Ang Microsoft ay nakatuon sa pagtulong sa aming mga customer na gamitin ang aming mga produktong AI nang responsable, ibahagi ang aming mga natutunan, at bumuo ng mga partnership na nakabatay sa pagtitiwala sa pamamagitan ng mga tool tulad ng Transparency Notes at Impact Assessments. Marami sa mga mapagkukunang ito ay makikita sa [https://aka.ms/RAI](https://aka.ms/RAI). -Ang paraan ng Microsoft sa responsable AI ay nakabase sa aming mga prinsipyong AI ng patas, maaasahan at ligtas, privacy at seguridad, pagiging inklusibo, transparency, at pananagutan. -Ang mga malalaking modelo ng natural na wika, imahe, at pagsasalita - tulad ng mga ginamit sa sample na ito - ay maaaring kumilos sa mga paraan na hindi patas, hindi maaasahan, o nakakasakit, na maaaring magdulot ng pinsala. Mangyaring kumonsulta sa [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) upang malaman ang mga panganib at limitasyon. -Ang inirerekomendang paraan upang mabawasan ang mga panganib na ito ay ang magsama ng isang sistema ng kaligtasan sa iyong arkitektura na maaaring makadetect at makapigil ng mapanganib na pag-uugali. Nagbibigay ang [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ng isang independiyenteng layer ng proteksyon, na kayang matukoy ang mapanganib na nilalaman na ginawa ng user at AI sa mga aplikasyon at serbisyo. Kasama sa Azure AI Content Safety ang mga text at image API na nagpapahintulot sa iyo na matukoy ang mga materyal na mapanganib. Sa loob ng Microsoft Foundry, ang serbisyo ng Content Safety ay nagbibigay-daan sa iyo upang tingnan, tuklasin at subukan ang sample code para sa pagtuklas ng mapanganib na nilalaman sa iba't ibang modalidad. Ang sumusunod na [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ay gagabay sa iyo sa paggawa ng mga request sa serbisyo. +Ang inirerekomendang paraan upang mabawasan ang mga panganib na ito ay ang pagsama ng isang safety system sa iyong arkitektura na makakapag-detect at makakapigilan ng nakakapinsalang pag-uugali. Nagbibigay ang [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ng isang independiyenteng layer ng proteksyon, na kayang matukoy ang nakakapinsalang nilalaman na ginawa ng gumagamit at AI sa mga aplikasyon at serbisyo. Kasama sa Azure AI Content Safety ang mga text at image API na nagpapahintulot sa iyo na matukoy ang materyal na nakakapinsala. Sa loob ng Microsoft Foundry, pinapayagan ka ng Content Safety service na tingnan, tuklasin, at subukan ang sample code para sa pagtuklas ng nakakapinsalang nilalaman sa iba't ibang modality. Ang sumusunod na [quickstart documentation](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) ay gagabay sa iyo sa paggawa ng mga request sa serbisyo. -Isa pang aspeto na dapat isaalang-alang ay ang kabuuang pagganap ng aplikasyon. Sa mga multi-modal at multi-model na aplikasyon, itinuturing namin ang pagganap bilang ang sistema ay gumaganap ayon sa inaasahan mo at ng iyong mga user, kabilang na ang hindi paggawa ng mapanganib na outputs. Mahalaga na suriin ang pagganap ng iyong pangkalahatang aplikasyon gamit ang [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Mayroon ka ring kakayahan na lumikha at magsuri gamit ang [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Isa pang aspeto na dapat isaalang-alang ay ang kabuuang performance ng aplikasyon. Sa mga multi-modal at multi-model na aplikasyon, itinuturing naming ang performance bilang ang sistema ay gumaganap ayon sa inaasahan mo at ng iyong mga gumagamit, kabilang ang hindi paglikha ng nakakapinsalang output. Mahalaga na suriin ang performance ng iyong kabuuang aplikasyon gamit ang [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Mayroon ka ring kakayahan na gumawa at magsuri gamit ang [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Maaari mong suriin ang iyong AI aplikasyon sa iyong development environment gamit ang [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Kapag mayroong test dataset o target, ang mga generative AI application generation mo ay nasusukat nang quantitatively gamit ang built-in evaluators o custom evaluators na iyong pinili. Upang makapagsimula gamit ang azure ai evaluation sdk para suriin ang iyong sistema, maaari mong sundan ang [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Kapag naisagawa mo na ang evaluation run, maaari mo nang [makita ang resulta sa Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Maaari mong suriin ang iyong AI aplikasyon sa iyong development environment gamit ang [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Sa pamamagitan ng test dataset o target, nasusukat nang quantitatively ang mga generasyon ng iyong generative AI application gamit ang built-in evaluators o custom evaluators na iyong pinili. Para makapagsimula gamit ang azure ai evaluation sdk upang suriin ang iyong sistema, maaari kang sumunod sa [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Kapag naisagawa mo na ang evaluation run, maaari mong [vizwalisa ang mga resulta sa Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Mga Trademark -Maaaring naglalaman ang proyektong ito ng mga trademark o logo para sa mga proyekto, produkto, o serbisyo. Ang awtorisadong paggamit ng mga trademark o logo ng Microsoft ay sakop at dapat sumunod sa [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Ang paggamit ng mga trademark o logo ng Microsoft sa mga binagong bersyon ng proyektong ito ay hindi dapat magdulot ng kalituhan o magpahiwatig ng sponsorship ng Microsoft. Anumang paggamit ng trademark o logo ng third-party ay alinsunod sa mga patakaran ng third-party na iyon. +Maaaring naglalaman ang proyektong ito ng mga trademark o logo para sa mga proyekto, produkto, o serbisyo. Ang awtorisadong paggamit ng mga Microsoft trademark o logo ay sakop sa at dapat sumunod sa [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Ang paggamit ng Microsoft trademark o logo sa mga binagong bersyon ng proyektong ito ay hindi dapat magdulot ng kalituhan o magpahiwatig ng sponsorship ng Microsoft. Ang anumang paggamit ng third-party trademarks o logo ay sakop sa patakaran ng mga third-party na iyon. -## Paghingi ng Tulong +## Pagkuha ng Tulong -Kung ikaw ay nahihirapan o may mga tanong tungkol sa paggawa ng mga AI na apps, sumali sa: +Kung ikaw ay nahihirapan o may mga tanong tungkol sa paggawa ng AI apps, sumali sa: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Kung mayroon kang feedback sa produkto o mga error habang nagde-develop, bisitahin ang: +Kung mayroon kang feedback sa produkto o mga error habang gumagawa bisitahin: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Disclaimer**: -Ang dokumentong ito ay isinalin gamit ang AI translation service na [Co-op Translator](https://github.com/Azure/co-op-translator). Bagamat nagsusumikap kami para sa katumpakan, pakiusap tandaan na ang mga awtomatikong pagsasalin ay maaaring maglaman ng mga pagkakamali o di-tumpak na impormasyon. Ang orihinal na dokumento sa orihinal nitong wika ang dapat ituring na pangunahing sanggunian. Para sa mahahalagang impormasyon, inirerekomenda ang propesyonal na pagsasaling-tao. Hindi kami mananagot sa anumang hindi pagkakaunawaan o maling interpretasyon na nagmumula sa paggamit ng pagsasaling ito. +**Pagtatanggi**: +Ang dokumentong ito ay isinalin gamit ang serbisyo ng AI translation na [Co-op Translator](https://github.com/Azure/co-op-translator). Bagama't nagsusumikap kami para sa katumpakan, pakatandaan na ang awtomatikong pagsasalin ay maaaring maglaman ng mga pagkakamali o hindi pagkakatugma. Ang orihinal na dokumento sa orihinal nitong wika ang dapat ituring na pangunahing sanggunian. Para sa mahahalagang impormasyon, inirerekomenda ang propesyonal na pagsasalin ng tao. Hindi kami mananagot sa anumang maling pagkakaintindi o maling interpretasyon na nagmula sa paggamit ng pagsasaling ito. \ No newline at end of file diff --git a/translations/tl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/tl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..fdf490e93 --- /dev/null +++ b/translations/tl/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Hybrid RAG gamit ang Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **May-akda:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Pokús:** Maliit na Language Models (SLMs), SQLite FTS5 BM25, Dense Embeddings, Reciprocal Rank Fusion (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motibasyon: Ang Dilemma ng Keyword Recall sa Lokal na SLMs\n", + "Ang mga karaniwang RAG na arkitektura na nagtitiwala lamang sa dense vector embeddings ay madalas nabibigo sa pagkuha ng eksaktong numerikal na mga token (hal., `2,340,000 TL`, mga code ng kontrata, mga numero ng account). \n", + "Sa kabilang banda, ang sparse lexical search (BM25) ay hindi nakakakuha ng mga semantikong kasingkahulugan at mga paraphrased na tanong.\n", + "\n", + "Ipinapakita ng cookbook na ito kung paano magpatupad ng **mabilis na, in-memory Hybrid Retrieval engine** na pinagsasama ang:\n", + "1. **Dense Vectors** (Cosine Similarity)\n", + "2. **Sparse Lexical Search** (SQLite FTS5 BM25)\n", + "3. **Reciprocal Rank Fusion (RRF, $k=60$)**\n", + "4. **Grounded Citation Generation (`[1]`, `[2]`)** gamit ang Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Dual na SQLite Schema (Dense Vectors + Virtual FTS5 BM25 Table)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Pagsusukat ng Pagsipsip at Pagpapatupad ng Halimbawa\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Pormat ng Grounded Prompt para sa Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Pagtatanggi**:\nAng dokumentong ito ay isinalin gamit ang serbisyo ng AI translation na [Co-op Translator](https://github.com/Azure/co-op-translator). Bagama't nagsusumikap kami para sa katumpakan, pakatandaan na ang awtomatikong pagsasalin ay maaaring maglaman ng mga pagkakamali o hindi pagkakatugma. Ang orihinal na dokumento sa orihinal nitong wika ang dapat ituring na pangunahing sanggunian. Para sa mahahalagang impormasyon, inirerekomenda ang propesyonal na pagsasalin ng tao. Hindi kami mananagot sa anumang maling pagkakaintindi o maling interpretasyon na nagmula sa paggamit ng pagsasaling ito.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/tr/.co-op-translator.json b/translations/tr/.co-op-translator.json index ca9dcfe0d..781a101eb 100644 --- a/translations/tr/.co-op-translator.json +++ b/translations/tr/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "tr" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:52:25+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T19:06:19+00:00", "source_file": "README.md", "language_code": "tr" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "tr" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T19:07:16+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "tr" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:53+00:00", diff --git a/translations/tr/README.md b/translations/tr/README.md index edd947c91..3941aa67a 100644 --- a/translations/tr/README.md +++ b/translations/tr/README.md @@ -1,42 +1,42 @@ -# Phi Yemek Kitabı: Microsoft'un Phi Modelleri ile Uygulamalı Örnekler +# Phi Yemek Kitabı: Microsoft'un Phi Modelleri ile Pratik Örnekler -[![GitHub Codespaces'ta örnekleri açıp kullanın](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![GitHub Codespaces'te örnekleri açın ve kullanın](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Dev Containers'da Aç](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub katkıcılar](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub katkıda bulunanlar](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub sorunlar](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub çekme talepleri](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PR'ler Hoş Geldiniz](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub çekme istekleri](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PR'lar Hoşgeldiniz](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub izleyenler](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub izleyiciler](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub çatallar](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub yıldızlar](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi, Microsoft tarafından geliştirilen açık kaynaklı yapay zeka modelleri serisidir. +Phi, Microsoft tarafından geliştirilmiş açık kaynaklı bir AI model serisidir. -Phi şu anda çok dilli, muhakeme, metin/sohbet üretimi, kodlama, görüntüler, ses ve diğer senaryolarda çok iyi performans gösteren en güçlü ve maliyet etkili küçük dil modeli (SLM) olarak öne çıkmaktadır. +Phi şu anda çoklu dil, muhakeme, metin/sohbet üretimi, kodlama, görüntüler, ses ve diğer senaryolarda çok iyi kıyaslamalara sahip, en güçlü ve maliyet açısından en etkin küçük dil modeli (SLM) durumundadır. -Phi'yi buluta veya uç cihazlara dağıtabilir ve sınırlı hesaplama gücüyle kolayca üretken yapay zeka uygulamaları geliştirebilirsiniz. +Phi'yi buluta veya uç cihazlara dağıtabilir ve sınırlı hesaplama gücüyle kolayca üretken AI uygulamaları geliştirebilirsiniz. Bu kaynakları kullanmaya başlamak için şu adımları izleyin: -1. **Depoyu Çatalla**: Tıklayın [![GitHub çatallar](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Depoyu Klonla**: `git clone https://github.com/microsoft/PhiCookBook.git` +1. **Depoyu Çatallayın**: Tıklayın [![GitHub çatallar](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Depoyu Klonlayın**: `git clone https://github.com/microsoft/PhiCookBook.git` 3. [**Microsoft AI Discord Topluluğuna Katılın ve uzmanlar ile diğer geliştiricilerle tanışın**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![kapak](../../translated_images/tr/cover.eb18d1b9605d754b.webp) ### 🌐 Çok Dilli Destek -#### GitHub Action ile Desteklenmektedir (Otomatik ve Her Zaman Güncel) +#### GitHub Action ile Desteklenir (Otomatik & Her Zaman Güncel) -[Arapça](../ar/README.md) | [Bengalce](../bn/README.md) | [Bulgarca](../bg/README.md) | [Burma (Myanmar)](../my/README.md) | [Çince (Basitleştirilmiş)](../zh-CN/README.md) | [Çince (Geleneksel, Hong Kong)](../zh-HK/README.md) | [Çince (Geleneksel, Makao)](../zh-MO/README.md) | [Çince (Geleneksel, Tayvan)](../zh-TW/README.md) | [Hırvatça](../hr/README.md) | [Çekçe](../cs/README.md) | [Danca](../da/README.md) | [Hollandaca](../nl/README.md) | [Estonca](../et/README.md) | [Fince](../fi/README.md) | [Fransızca](../fr/README.md) | [Almanca](../de/README.md) | [Yunanca](../el/README.md) | [İbranice](../he/README.md) | [Hintçe](../hi/README.md) | [Macarca](../hu/README.md) | [Endonezce](../id/README.md) | [İtalyanca](../it/README.md) | [Japonca](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korece](../ko/README.md) | [Litvanca](../lt/README.md) | [Malayca](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalce](../ne/README.md) | [Nijerya Pidgin](../pcm/README.md) | [Norveççe](../no/README.md) | [Farsça (Farsi)](../fa/README.md) | [Lehçe](../pl/README.md) | [Portekizce (Brezilya)](../pt-BR/README.md) | [Portekizce (Portekiz)](../pt-PT/README.md) | [Pencapça (Gurmukhi)](../pa/README.md) | [Rumence](../ro/README.md) | [Rusça](../ru/README.md) | [Sırpça (Kiril)](../sr/README.md) | [Slovakça](../sk/README.md) | [Slovence](../sl/README.md) | [İspanyolca](../es/README.md) | [Svahili](../sw/README.md) | [İsveççe](../sv/README.md) | [Tagalog (Filipinler)](../tl/README.md) | [Tamilce](../ta/README.md) | [Telugu](../te/README.md) | [Tayca](../th/README.md) | [Türkçe](./README.md) | [Ukraynaca](../uk/README.md) | [Urduca](../ur/README.md) | [Viyetnamca](../vi/README.md) +[Arapça](../ar/README.md) | [Bengalce](../bn/README.md) | [Bulgarca](../bg/README.md) | [Birmanca (Myanmar)](../my/README.md) | [Çince (Basitleştirilmiş)](../zh-CN/README.md) | [Çince (Geleneksel, Hong Kong)](../zh-HK/README.md) | [Çince (Geleneksel, Makao)](../zh-MO/README.md) | [Çince (Geleneksel, Tayvan)](../zh-TW/README.md) | [Hırvatça](../hr/README.md) | [Çekçe](../cs/README.md) | [Danca](../da/README.md) | [Felemenkçe](../nl/README.md) | [Estonca](../et/README.md) | [Fince](../fi/README.md) | [Fransızca](../fr/README.md) | [Almanca](../de/README.md) | [Yunanca](../el/README.md) | [İbranice](../he/README.md) | [Hintçe](../hi/README.md) | [Macarca](../hu/README.md) | [Endonezyaca](../id/README.md) | [İtalyanca](../it/README.md) | [Japonca](../ja/README.md) | [Kannada](../kn/README.md) | [Khmerce](../km/README.md) | [Korece](../ko/README.md) | [Litvanca](../lt/README.md) | [Malayca](../ms/README.md) | [Malayalamca](../ml/README.md) | [Marathi](../mr/README.md) | [Nepalce](../ne/README.md) | [Nijerya Pidgin](../pcm/README.md) | [Norveççe](../no/README.md) | [Farsça (Persian)](../fa/README.md) | [Lehçe](../pl/README.md) | [Portekizce (Brezilya)](../pt-BR/README.md) | [Portekizce (Portekiz)](../pt-PT/README.md) | [Pencapça (Gurmukhi)](../pa/README.md) | [Rumence](../ro/README.md) | [Rusça](../ru/README.md) | [Sırpça (Kiril)](../sr/README.md) | [Slovakça](../sk/README.md) | [Slovence](../sl/README.md) | [İspanyolca](../es/README.md) | [Svahilice](../sw/README.md) | [İsveççe](../sv/README.md) | [Tagalog (Filipince)](../tl/README.md) | [Tamilce](../ta/README.md) | [Telugu](../te/README.md) | [Tayca](../th/README.md) | [Türkçe](./README.md) | [Ukraynaca](../uk/README.md) | [Urduca](../ur/README.md) | [Vietnamca](../vi/README.md) -> **Yerel Klonlamayı mı Tercih Ediyorsunuz?** +> **Yerel olarak klonlamayı mı tercih ediyorsunuz?** > -> Bu depo 50'den fazla dil çevirisi içerir ve bu da indirme boyutunu önemli ölçüde artırır. Çeviriler olmadan klonlamak için sparse checkout kullanın: +> Bu depo 50'den fazla dil çevirisi içerir, bu da indirme boyutunu önemli ölçüde artırır. Çeviriler olmadan klonlamak için, seyrek çekme (sparse checkout) kullanın: > > **Bash / macOS / Linux:** > ```bash @@ -52,200 +52,203 @@ Bu kaynakları kullanmaya başlamak için şu adımları izleyin: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Bu, kursu tamamlamak için ihtiyacınız olan her şeyi çok daha hızlı bir indirme ile sunar. +> Bu, kursu tamamlamak için ihtiyacınız olan her şeyi çok daha hızlı bir indirme ile sağlar. ## İçindekiler - Giriş - - [Phi Ailesine Hoş Geldiniz](./md/01.Introduction/01/01.PhiFamily.md) + - [Phi Ailesine Hoşgeldiniz](./md/01.Introduction/01/01.PhiFamily.md) - [Ortamınızı Kurma](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Ana Teknolojileri Anlama](./md/01.Introduction/01/01.Understandingtech.md) - - [Phi Modelleri için Yapay Zeka Güvenliği](./md/01.Introduction/01/01.AISafety.md) + - [Temel Teknolojileri Anlama](./md/01.Introduction/01/01.Understandingtech.md) + - [Phi Modelleri için AI Güvenliği](./md/01.Introduction/01/01.AISafety.md) - [Phi Donanım Desteği](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi Modelleri ve Platformlar Arasındaki Kullanılabilirlik](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Phi Modelleri ve Platformlar Arası Kullanılabilirlik](./md/01.Introduction/01/01.Edgeandcloud.md) - [Guidance-ai ve Phi Kullanımı](./md/01.Introduction/01/01.Guidance.md) - [GitHub Marketplace Modelleri](https://github.com/marketplace/models) - [Azure AI Model Kataloğu](https://ai.azure.com) -- Farklı Ortamlarda Phi Çıkarımı +- Farklı ortamda Phi çıkarımı - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub Modelleri](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry Model Kataloğu](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry Yerel](./md/01.Introduction/02/07.FoundryLocal.md) - -- Phi Ailesi Çıkarımı - - [iOS'ta Phi Çıkarımı](./md/01.Introduction/03/iOS_Inference.md) - - [Android'te Phi Çıkarımı](./md/01.Introduction/03/Android_Inference.md) - - [Jetson'da Phi Çıkarımı](./md/01.Introduction/03/Jetson_Inference.md) - - [AI PC'de Phi Çıkarımı](./md/01.Introduction/03/AIPC_Inference.md) - - [Apple MLX Framework ile Phi Çıkarımı](./md/01.Introduction/03/MLX_Inference.md) - - [Yerel Sunucuda Phi Çıkarımı](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI Toolkit kullanarak Uzaktan Sunucuda Phi Çıkarımı](./md/01.Introduction/03/Remote_Interence.md) - - [Rust ile Phi Çıkarımı](./md/01.Introduction/03/Rust_Inference.md) - - [Yerelde Phi--Vision Çıkarımı](./md/01.Introduction/03/Vision_Inference.md) - - [Kaito AKS, Azure Containers ile Phi Çıkarımı (resmi destek)](./md/01.Introduction/03/Kaito_Inference.md) -- [Phi Ailesinin Kantifikasyonu](./md/01.Introduction/04/QuantifyingPhi.md) - - [llama.cpp kullanarak Phi-3.5 / 4'ü Kantize Etme](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime için Generative AI uzantıları kullanarak Phi-3.5 / 4'ü Kantize Etme](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Intel OpenVINO kullanarak Phi-3.5 / 4'ü Kantize Etme](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Apple MLX Framework kullanarak Phi-3.5 / 4'ü Kantize Etme](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Phi Değerlendirmesi + - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) + +- Phi Ailesinde çıkarım + - [iOS'ta Phi çıkarımı](./md/01.Introduction/03/iOS_Inference.md) + - [Android'te Phi çıkarımı](./md/01.Introduction/03/Android_Inference.md) + - [Jetson'da Phi çıkarımı](./md/01.Introduction/03/Jetson_Inference.md) + - [AI PC'de Phi çıkarımı](./md/01.Introduction/03/AIPC_Inference.md) + - [Apple MLX Framework ile Phi çıkarımı](./md/01.Introduction/03/MLX_Inference.md) + - [Yerel Sunucuda Phi çıkarımı](./md/01.Introduction/03/Local_Server_Inference.md) + - [AI Toolkit kullanarak Uzak Sunucuda Phi çıkarımı](./md/01.Introduction/03/Remote_Interence.md) + - [Rust ile Phi çıkarımı](./md/01.Introduction/03/Rust_Inference.md) + - [Yerelde Phi--Vision çıkarımı](./md/01.Introduction/03/Vision_Inference.md) + - [Kaito AKS, Azure Containers (resmi destek) ile Phi çıkarımı](./md/01.Introduction/03/Kaito_Inference.md) +- [Phi Ailesini Nicelendir](./md/01.Introduction/04/QuantifyingPhi.md) + - [llama.cpp kullanarak Phi-3.5 / 4 nicelendirme](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntime için Üretken AI uzantıları kullanarak Phi-3.5 / 4 nicelendirme](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Intel OpenVINO kullanarak Phi-3.5 / 4 nicelendirme](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Apple MLX Framework kullanarak Phi-3.5 / 4 nicelendirme](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- Phi Değerlendirme - [Sorumlu AI](./md/01.Introduction/05/ResponsibleAI.md) - - [Değerlendirme için Microsoft Foundry](./md/01.Introduction/05/AIFoundry.md) - - [Değerlendirme için Promptflow Kullanımı](./md/01.Introduction/05/Promptflow.md) + - [Microsoft Foundry değerlendirme için](./md/01.Introduction/05/AIFoundry.md) + - [Değerlendirme için Promptflow kullanımı](./md/01.Introduction/05/Promptflow.md) -- Azure AI Search ile RAG - - [Phi-4-mini ve Phi-4-multimodal (RAG) ile Azure AI Search Kullanımı](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- Azure AI Ara ile RAG + - [Phi-4-mini ve Phi-4-multimodal (RAG) Azure AI Ara ile nasıl kullanılır](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Yerel Hibrit RAG SQLite FTS5 ve phi-4-mini ile Zero-Cloud](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Phi uygulama geliştirme örnekleri - Metin & Sohbet Uygulamaları - Phi-4 Örnekleri - [📓] [Phi-4-mini ONNX Modeli ile Sohbet](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 yerel ONNX Modeli .NET ile Sohbet](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Sementic Kernel kullanarak Phi-4 ONNX ile .NET Konsol Uygulaması Sohbet](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - [Phi-4 yerel ONNX Modeli ile Sohbet .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Semantic Kernel kullanarak Phi-4 ONNX ile .NET Konsol Uygulaması Sohbeti](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 Örnekleri - - [Phi3, ONNX Runtime Web ve WebGPU kullanarak tarayıcıda Yerel Sohbetbot](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Sohbet](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Phi3, ONNX Runtime Web ve WebGPU kullanarak tarayıcıda Yerel Sohbet Botu](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [Çoklu Model - Etkileşimli Phi-3-mini ve OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Bir sarıcı oluşturma ve Phi-3'ü MLFlow ile kullanma](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Model Optimizasyonu - Phi-3-min modelini ONNX Runtime Web için Olive ile nasıl optimize edilir](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 Uygulaması Phi-3 mini-4k-instruct-onnx ile](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - - [WinUI3 Çoklu Model Yapay Zeka Destekli Notlar Uygulaması Örneği](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [MLFlow - Bir sarmalayıcı oluşturma ve Phi-3'ü MLFlow ile kullanma](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Model Optimizasyonu - Olive ile ONNX Runtime Web için Phi-3-min modeli nasıl optimize edilir](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Phi-3 mini-4k-instruct-onnx ile WinUI3 Uygulaması](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 Çoklu Model Yapay Zeka Destekli Notlar Uygulaması Örneği](https://github.com/microsoft/ai-powered-notes-winui3-sample) - [Özel Phi-3 modellerini ince ayar yapma ve Prompt flow ile entegre etme](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - [Microsoft Foundry'de Prompt flow ile özel Phi-3 modellerini ince ayar yapma ve entegre etme](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Microsoft'un Sorumlu Yapay Zeka İlkelerine odaklanarak Microsoft Foundry'de ince ayar yapılmış Phi-3 / Phi-3.5 Modelini değerlendirme](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct dil tahmini örneği (Çince/İngilizce)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU kullanarak Phi-3.5-Instruct ONNX ile Prompt flow çözümü oluşturma](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [Microsoft'un Sorumlu AI İlkelerine odaklanarak Microsoft Foundry'de İnce Ayarlı Phi-3 / Phi-3.5 Modelini Değerlendirme](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct dil tahmin örneği (Çince/İngilizce)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG Sohbet Botu](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Phi-3.5-Instruct ONNX ile Windows GPU kullanarak Prompt flow çözümü oluşturma](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Microsoft Phi-3.5 tflite kullanarak Android uygulaması oluşturma](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Microsoft.ML.OnnxRuntime kullanarak yerel ONNX Phi-3 modeli ile Soru-Cevap .NET Örneği](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Microsoft.ML.OnnxRuntime kullanarak yerel ONNX Phi-3 modeliyle .NET Soru&Cevap Örneği](../../md/04.HOL/dotnet/src/LabsPhi301) - [Semantic Kernel ve Phi-3 ile Konsol sohbet .NET uygulaması](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Tahmin SDK Kod Tabanlı Örnekler - - Phi-4 Örnekleri + - Azure AI Çıkarım SDK Kod Tabanlı Örnekler + - Phi-4 Örnekleri - [📓] [Phi-4-multimodal kullanarak proje kodu oluşturma](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 Örnekleri - [Microsoft Phi-3 Ailesi ile kendi Visual Studio Code GitHub Copilot Sohbetinizi oluşturun](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub Modelleri ile Phi-3.5 kullanarak kendi Visual Studio Code Sohbet Copilot Ajanınızı oluşturun](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [Phi-3.5 ile GitHub Modelleri kullanarak kendi Visual Studio Code Sohbet Copilot Ajanınızı oluşturun](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - Gelişmiş Akıl Yürütme Örnekleri - - Phi-4 Örnekleri + - Phi-4 Örnekleri - [📓] [Phi-4-mini-akıl yürütme veya Phi-4-akıl yürütme Örnekleri](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Microsoft Olive ile Phi-4-mini-akıl yürütme ince ayarı](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Apple MLX ile Phi-4-mini-akıl yürütme ince ayarı](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Microsoft Olive ile Phi-4-mini-akıl yürütmeyi ince ayar yapma](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLX ile Phi-4-mini-akıl yürütmeyi ince ayar yapma](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [GitHub Modelleri ile Phi-4-mini-akıl yürütme](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Microsoft Foundry Modelleri ile Phi-4-mini-akıl yürütme](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Demo'lar - - [Hugging Face Spaces'da barındırılan Phi-4-mini demoları](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Hugging Face Spaces'da barındırılan Phi-4-multimodal demoları](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Demolar + - [Hugging Face Spaces üzerinde barındırılan Phi-4-mini demoları](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Hugging Face Spaces üzerinde barındırılan Phi-4-multimodal demoları](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - Görüntü Örnekleri - - Phi-4 Örnekleri - - [📓] [Phi-4-multimodal kullanarak resimleri okuma ve kod üretme](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Phi-4 Örnekleri + - [📓] [Phi-4-multimodal kullanarak resimleri okumak ve kod üretmek](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 Örnekleri - - [📓][Phi-3-görüntü-Resim metinden metne](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [Phi-3-görüntü-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-görüntü CLIP Gömme](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-Resimden metne metne](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) + - [📓][Phi-3-vision CLIP Gömme](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [DEMO: Phi-3 Geri Dönüşüm](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-görüntü - Görsel dil asistanı - Phi3-Görüntü ve OpenVINO ile](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 Görüntü Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 Görüntü OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Görüntü çok kareli veya çoklu resim örneği](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Microsoft.ML.OnnxRuntime .NET kullanarak yerel Phi-3 Görüntü ONNX Modeli](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Menü tabanlı Microsoft.ML.OnnxRuntime .NET kullanarak yerel Phi-3 Görüntü ONNX Modeli](../../md/04.HOL/dotnet/src/LabsPhi304) + - [Phi-3-vision - Görsel dil asistanı - Phi3-Vision ve OpenVINO ile](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision çoklu kare veya çoklu resim örneği](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision Microsoft.ML.OnnxRuntime .NET kullanarak Yerel ONNX Modeli](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Menü tabanlı Phi-3 Vision Microsoft.ML.OnnxRuntime .NET kullanarak Yerel ONNX Modeli](../../md/04.HOL/dotnet/src/LabsPhi304) - Akıl Yürütme-Görüntü Örnekleri - - Phi-4-AkılYürütme-Görüntü-15B - - [📓] [Phi-4-AkılYürütme-Görüntü-15B kullanarak karşıdan karşıya geçme algılama](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Phi-4-AkılYürütme-Görüntü-15B kullanarak matematik](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-AkılYürütme-Görüntü-15B kullanarak UI algılama](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - Phi-4-Akıl Yürütme-Görüntü-15B + - [📓] [Phi-4-Akıl Yürütme-Görüntü-15B kullanarak yaya önceliğini ihlal etme tespiti](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Phi-4-Akıl Yürütme-Görüntü-15B kullanarak matematik](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Phi-4-Akıl Yürütme-Görüntü-15B kullanarak UI tespiti](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - Matematik Örnekleri - - Phi-4-Mini-Flash-Akıl-Yürütme-Yönergeleri Örnekleri [Phi-4-Mini-Flash-Akıl-Yürütme-Yönergeleri ile Matematik Demo](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-Mini-Flash-Akıl Yürütme-Yönlendirme Örnekleri [Phi-4-Mini-Flash-Akıl Yürütme-Yönlendirme ile Matematik Demosu](./md/02.Application/09.Math/MathDemo.ipynb) - Ses Örnekleri - - Phi-4 Örnekleri + - Phi-4 Örnekleri - [📓] [Phi-4-multimodal kullanarak ses transkriptlerini çıkarma](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Phi-4-multimodal Ses Örneği](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Phi-4-multimodal Konuşma Çevirisi Örneği](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET konsol uygulaması Phi-4-multimodal Ses kullanarak bir ses dosyasını analiz etme ve transkript oluşturma](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [.NET konsol uygulaması, Phi-4-multimodal kullanarak ses dosyasını analiz edip transkript oluşturma](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE Örnekleri - Phi-3 / 3.5 Örnekleri - - [📓] [Phi-3.5 Uzman Karışımı Modeller (MoEs) Sosyal Medya Örneği](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search ve LlamaIndex ile Bir Retrieval-Augmented Generation (RAG) Pipeline Oluşturma](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 Uzman Karışımı Modelleri (MoEs) Sosyal Medya Örneği](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE, Azure AI Search ve LlamaIndex kullanarak Retrieval-Augmented Generation (RAG) Boru Hattı oluşturma](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - Fonksiyon Çağırma Örnekleri - Phi-4 Örnekleri 🆕 - - [📓] [Phi-4-mini ile Fonksiyon Çağırma kullanımı](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Phi-4-mini ile çoklu ajanlar oluşturmak için Fonksiyon Çağırma kullanımı](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Ollama ile Fonksiyon Çağırma kullanımı](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [ONNX ile Fonksiyon Çağırma kullanımı](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - [📓] [Phi-4-mini ile Fonksiyon Çağırmayı Kullanma](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Phi-4-mini ile çoklu ajan oluşturmak için Fonksiyon Çağırmayı Kullanma](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Ollama ile Fonksiyon Çağırmayı Kullanma](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [ONNX ile Fonksiyon Çağırmayı Kullanma](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - Çok Modlu Karışım Örnekleri - Phi-4 Örnekleri 🆕 - - [📓] [Phi-4-multimodal'ı Teknoloji gazetecisi olarak kullanma](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET konsol uygulaması Phi-4-multimodal kullanarak resimleri analiz etme](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [📓] [Phi-4-multimodal kullanarak Teknoloji gazetecisi olma](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET konsol uygulaması, Phi-4-multimodal kullanarak resimleri analiz etme](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - Phi İnce Ayar Örnekleri - [İnce Ayar Senaryoları](./md/03.FineTuning/FineTuning_Scenarios.md) - - [İnce Ayar ile RAG Karşılaştırması](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Phi-3'ü bir endüstri uzmanı yapma ince ayarı](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [VS Code için AI Araç Kiti ile Phi-3 ince ayarı](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Azure Machine Learning Hizmeti ile Phi-3 ince ayarı](./md/03.FineTuning/Introduce_AzureML.md) - - [Lora ile Phi-3 ince ayarı](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora ile Phi-3 ince ayarı](./md/03.FineTuning/FineTuning_Qlora.md) - - [Microsoft Foundry ile Phi-3 ince ayarı](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Azure ML CLI/SDK ile Phi-3 ince ayarı](./md/03.FineTuning/FineTuning_MLSDK.md) + - [İnce Ayar ve RAG karşılaştırması](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Phi-3'ün sektör uzmanı olmasını sağlamak için ince ayar](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [AI Araç Seti ile VS Code için Phi-3'ü ince ayar yapma](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure Machine Learning Servisi ile Phi-3'ü ince ayar yapma](./md/03.FineTuning/Introduce_AzureML.md) + - [Lora ile Phi-3'ü ince ayar yapma](./md/03.FineTuning/FineTuning_Lora.md) + - [QLora ile Phi-3'ü ince ayar yapma](./md/03.FineTuning/FineTuning_Qlora.md) + - [Microsoft Foundry ile Phi-3'ü ince ayar yapma](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Azure ML CLI/SDK ile Phi-3'ü ince ayar yapma](./md/03.FineTuning/FineTuning_MLSDK.md) - [Microsoft Olive ile ince ayar](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Microsoft Olive Uygulamalı Laboratuvar ile ince ayar](./md/03.FineTuning/olive-lab/readme.md) - - [Weights and Bias ile Phi-3-görüntü ince ayarı](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Apple MLX Framework ile Phi-3 ince ayarı](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-görüntü ince ayarı (resmi destek)](./md/03.FineTuning/FineTuning_Vision.md) - - [Kaito AKS ile Phi-3 İnce Ayarı, Azure Containers (Resmi Destek)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Weights and Bias ile Phi-3-görüş ince ayar](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Apple MLX Framework ile Phi-3 İnce Ayarı](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision İnce Ayarı (resmi destek)](./md/03.FineTuning/FineTuning_Vision.md) + - [Kaito AKS, Azure Konteynerleri ile Phi-3 İnce Ayarı (resmi Destek)](./md/03.FineTuning/FineTuning_Kaito.md) - [Phi-3 ve 3.5 Vision İnce Ayarı](https://github.com/2U1/Phi3-Vision-Finetune) - Uygulamalı Laboratuvar - - [Öncü modellerin keşfi: LLM'ler, SLM'ler, yerel geliştirme ve daha fazlası](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Keskin modelleri keşfetmek: LLM'ler, SLM'ler, yerel geliştirme ve daha fazlası](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - [NLP Potansiyelini Açığa Çıkarmak: Microsoft Olive ile İnce Ayar](https://github.com/azure/Ignite_FineTuning_workshop) - Akademik Araştırma Makaleleri ve Yayınlar - [Sadece Ders Kitapları II: phi-1.5 teknik raporu](https://arxiv.org/abs/2309.05463) - - [Phi-3 Teknik Raporu: Telefonunuzda Yüksek Kapasiteli Dil Modeli](https://arxiv.org/abs/2404.14219) - - [Phi-4 Teknik Raporu](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini Teknik Raporu: Mixture-of-LoRAs ile Kompakt ama Güçlü Multimodal Dil Modelleri](https://arxiv.org/abs/2503.01743) + - [Phi-3 Teknik Raporu: Telefonunuzda Yerel Olarak Yüksek Kapasiteli Dil Modeli](https://arxiv.org/abs/2404.14219) + - [Phi-4 Teknik Rapor](https://arxiv.org/abs/2412.08905) + - [Phi-4-Mini Teknik Rapor: Karışım-LoRA'lar ile Kompakt ancak Güçlü Çok Modlu Dil Modelleri](https://arxiv.org/abs/2503.01743) - [Araç İçi Fonksiyon Çağrısı için Küçük Dil Modellerinin Optimize Edilmesi](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Çoktan Seçmeli Soru Cevaplama için PHI-3 İnce Ayarı: Yöntem, Sonuçlar ve Zorluklar](https://arxiv.org/abs/2501.01588) - - [Phi-4-mantık Teknik Raporu](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini-mantık Teknik Raporu](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [(WhyPHI) Çoktan Seçmeli Soru Yanıtlama için PHI-3 İnce Ayar: Yöntemoloji, Sonuçlar ve Zorluklar](https://arxiv.org/abs/2501.01588) + - [Phi-4-Çözümleme Teknik Raporu](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini-çözümleme Teknik Raporu](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Phi Modellerini Kullanma -### Microsoft Foundry'de Phi +### Microsoft Foundry üzerinde Phi -Microsoft Phi'yi nasıl kullanacağınızı ve farklı donanım cihazlarınızda uçtan uca çözümler nasıl oluşturacağınızı öğrenebilirsiniz. Phi'yi kendiniz deneyimlemek için modellerle oynamaya başlayın ve Phi'yi senaryolarınıza göre özelleştirin. Daha fazla bilgi için [Microsoft Foundry Azure AI Model Envanteri](https://aka.ms/phi3-azure-ai) ve [Microsoft Foundry ile Başlangıç](./md/02.QuickStart/AzureAIFoundry_QuickStart.md) rehberine göz atabilirsiniz. +Microsoft Phi'yi nasıl kullanacağınızı ve farklı donanım cihazlarınızda uçtan uca çözümler nasıl oluşturacağınızı öğrenebilirsiniz. Phi'yi kendiniz deneyimlemek için, modellerle oynamaya başlayın ve senaryolarınıza göre Phi'yi özelleştirin. Bunun için [Microsoft Foundry Azure AI Model Kataloğu](https://aka.ms/phi3-azure-ai) kullanabilirsiniz. Daha fazlasını [Microsoft Foundry ile Başlarken](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) bölümünde öğrenebilirsiniz. -**Oyun Alanı** -Her modelin test edilmesi için ayrılmış bir oyun alanı vardır: [Azure AI Playground](https://aka.ms/try-phi3). +**Oyun Alanı** +Her modelin test etmek için ayrılmış bir oyun alanı vardır [Azure AI Playground](https://aka.ms/try-phi3). ### GitHub Modellerinde Phi -Microsoft Phi'yi nasıl kullanacağınızı ve farklı donanım cihazlarınızda uçtan uca çözümler oluşturmayı öğrenebilirsiniz. Phi'yi kendiniz deneyimlemek için modelle oynamaya başlayın ve Phi'yi senaryolarınıza göre özelleştirin. Daha fazla bilgi için [GitHub Model Envanteri](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ve [GitHub Model Envanteri ile Başlangıç](./md/02.QuickStart/GitHubModel_QuickStart.md) rehberine göz atabilirsiniz. +Microsoft Phi'yi nasıl kullanacağınızı ve farklı donanım cihazlarınızda uçtan uca çözümler nasıl oluşturacağınızı öğrenebilirsiniz. Phi'yi kendiniz deneyimlemek için, modelle oynamaya başlayın ve senaryolarınıza göre Phi'yi özelleştirin. Bunun için [GitHub Model Kataloğu](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) kullanabilirsiniz. Daha fazlasını [GitHub Model Kataloğu ile Başlarken](/md/02.QuickStart/GitHubModel_QuickStart.md) bölümünde öğrenebilirsiniz. -**Oyun Alanı** -Her modelin test edilmesi için ayrılmış bir [oyun alanı](/md/02.QuickStart/GitHubModel_QuickStart.md) vardır. +**Oyun Alanı** +Her modelin test etmek için ayrılmış bir [oyun alanı vardır](/md/02.QuickStart/GitHubModel_QuickStart.md). -### Hugging Face'te Phi +### Hugging Face üzerinde Phi Modeli ayrıca [Hugging Face](https://huggingface.co/microsoft) üzerinde de bulabilirsiniz. -**Oyun Alanı** -[Hugging Chat oyun alanı](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Oyun Alanı** + [Hugging Chat oyun alanı](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Diğer Kurslar + ## 🎒 Diğer Kurslar Ekibimiz başka kurslar da üretiyor! Göz atın: @@ -260,10 +263,10 @@ Ekibimiz başka kurslar da üretiyor! Göz atın: [![Yeni Başlayanlar için AZD](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Yeni Başlayanlar için Edge AI](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Yeni Başlayanlar için MCP](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Yeni Başlayanlar için AI Ajanlar](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Yeni Başlayanlar için AI Ajanları](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Üretken AI Serisi [![Yeni Başlayanlar için Üretken AI](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![Üretken AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) @@ -271,7 +274,7 @@ Ekibimiz başka kurslar da üretiyor! Göz atın: [![Üretken AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - + ### Temel Öğrenme [![Yeni Başlayanlar için ML](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![Yeni Başlayanlar için Veri Bilimi](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) @@ -282,42 +285,45 @@ Ekibimiz başka kurslar da üretiyor! Göz atın: [![Yeni Başlayanlar için XR Geliştirme](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Copilot Serisi -[![AI Eşli Programlama için Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Yapay Zeka ile Eşli Programlama için Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![C#/.NET için Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot Macerası](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Sorumlu Yapay Zeka +## Sorumlu AI + +Microsoft, müşterilerimizin AI ürünlerimizi sorumlu bir şekilde kullanmalarına yardımcı olmaya; öğrendiklerimizi paylaşmaya ve Şeffaflık Notları ve Etki Değerlendirmeleri gibi araçlarla güvene dayalı ortaklıklar kurmaya kararlıdır. Bu kaynakların çoğu [https://aka.ms/RAI](https://aka.ms/RAI) adresinde bulunabilir. +Microsoft’un sorumlu AI yaklaşımı, adalet, güvenilirlik ve güvenlik, gizlilik ve güvenlik, kapsayıcılık, şeffaflık ve hesap verebilirlik gibi AI ilkelerine dayanır. + +Bu örnekte kullanılanlar gibi büyük ölçekli doğal dil, görüntü ve konuşma modelleri adaletsiz, güvenilmez veya saldırgan davranışlar sergileyebilir ve bu da zararlara yol açabilir. Riskler ve sınırlamalar hakkında bilgi edinmek için lütfen [Azure OpenAI servisi Şeffaflık notu](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) sayfasına bakınız. -Microsoft, müşterilerimizin yapay zeka ürünlerimizi sorumlu bir şekilde kullanmalarına yardımcı olmaya, öğrendiklerimizi paylaşmaya ve Transparan Notlar ve Etki Değerlendirmeleri gibi araçlarla güvene dayalı ortaklıklar kurmaya kararlıdır. Bu kaynakların birçoğuna [https://aka.ms/RAI](https://aka.ms/RAI) adresinden ulaşabilirsiniz. -Microsoft'un sorumlu yapay zeka yaklaşımı, adalet, güvenilirlik ve güvenlik, gizlilik ve güvenlik, kapsayıcılık, şeffaflık ve hesap verebilirlik yapay zeka ilkelerimize dayanır. -Bu örnekte kullanılan gibi büyük ölçekli doğal dil, görüntü ve konuşma modelleri potansiyel olarak adaletsiz, güvenilmez veya saldırgan davranışlar sergileyerek zarar verebilir. Riskler ve sınırlamalar hakkında bilgi edinmek için [Azure OpenAI hizmeti Şeffaflık notuna](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) başvurunuz. -Bu riskleri azaltmak için önerilen yaklaşım, zararlı davranışları tespit edip önleyebilen bir güvenlik sistemini mimarinize dahil etmektir. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview), uygulamalarda ve servislerde zararlı kullanıcı tarafından oluşturulan ve yapay zeka tarafından oluşturulan içeriği tespit edebilen bağımsız bir koruma katmanı sağlar. Azure AI Content Safety, zararlı materyali tespit etmenize olanak tanıyan metin ve resim API'lerini içerir. Microsoft Foundry içinde, Content Safety servisi, farklı modaliteler arasında zararlı içeriği tespit etmek için örnek kodu görüntülemenize, keşfetmenize ve denemenize olanak tanır. Aşağıdaki [hızlı başlangıç dökümü](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) hizmete istek yapma konusunda sizi yönlendirmektedir. +Bu riskleri azaltmak için önerilen yaklaşım, zararlı davranışları algılayabilen ve önleyebilen bir güvenlik sistemini mimarinize dahil etmektir. [Azure AI İçerik Güvenliği](https://learn.microsoft.com/azure/ai-services/content-safety/overview), uygulamalarda ve hizmetlerde zararlı kullanıcı tarafından oluşturulan ve yapay zeka tarafından oluşturulan içeriği algılayabilen bağımsız bir koruma katmanı sağlar. Azure AI İçerik Güvenliği, zararlı materyalleri algılamanıza olanak tanıyan metin ve görüntü API'lerini içerir. Microsoft Foundry içinde, İçerik Güvenliği hizmeti, farklı modalitelerde zararlı içeriği algılamak için örnek kodları görüntülemenize, keşfetmenize ve denemenize olanak tanır. Aşağıdaki [hızlı başlangıç belgeleri](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) size hizmete istek yapmanızda rehberlik eder. -Dikkate alınması gereken bir diğer husus ise genel uygulama performansıdır. Çok modlu ve çok modellli uygulamalarda, performansın, sizin ve kullanıcılarınızın beklentilerine uygun şekilde sistemin çalışması, zararlı çıktılar üretmemesi anlamına geldiğini düşünüyoruz. Genel uygulamanızın performansını değerlendirmek için [Performans ve Kalite ile Risk ve Güvenlik değerlendiricilerini](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) kullanmanız önemlidir. Ayrıca, [özel değerlendiriciler](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) oluşturma ve değerlendirme yeteneğiniz de bulunmaktadır. +Dikkate alınması gereken bir diğer husus ise genel uygulama performansıdır. Çoklu modlu ve çoklu model uygulamalarda performans, sistemin sizin ve kullanıcılarınızın beklentilerine uygun olarak çalışması, zararlı çıktılar üretmemesi anlamına gelir. [Performans ve Kalite ile Risk ve Güvenlik değerlendiricileri](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) kullanarak genel uygulama performansınızı değerlendirmeniz önemlidir. Ayrıca, [özelleştirilmiş değerlendiriciler](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) oluşturma ve bunlarla değerlendirme yapma imkanınız da vardır. -AI uygulamanızı geliştirme ortamınızda [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) kullanarak değerlendirebilirsiniz. Bir test veri seti veya hedef verildiğinde, üretken AI uygulamanızın çıktıları, yerleşik değerlendiriciler veya seçtiğiniz özel değerlendiricilerle niceliksel olarak ölçülür. Sisteminizi değerlendirmek için azure ai evaluation sdk ile başlamak isterseniz, [hızlı başlangıç kılavuzunu](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) takip edebilirsiniz. Bir değerlendirme çalıştırdıktan sonra, sonucu [Microsoft Foundry’de görselleştirebilirsiniz](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Geliştirme ortamınızda [Azure AI Değerlendirme SDK'sı](https://microsoft.github.io/promptflow/index.html) kullanarak AI uygulamanızı değerlendirebilirsiniz. Bir test veri seti veya hedef verildiğinde, üretilen yapay zeka çıktılarınız yerleşik değerlendiriciler veya tercih ettiğiniz özelleştirilmiş değerlendiricilerle niceliksel olarak ölçülür. Sisteminizin değerlendirilmesine başlamak için [hızlı başlangıç kılavuzunu](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) takip edebilirsiniz. Bir değerlendirme çalışması başlattıktan sonra, [sonuçları Microsoft Foundry'de görselleştirebilirsiniz](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). -## Ticarî Markalar +## Ticari Markalar -Bu proje, projeler, ürünler veya servisler için ticarî markalar veya logolar içerebilir. Microsoft ticarî markalarının veya logolarının yetkili kullanımı, [Microsoft'un Ticarî Marka ve Marka Yönergeleri](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) kurallarına tabidir ve bu kurallara uymalıdır. Microsoft ticarî markalarının veya logolarının bu projenin değiştirilmiş sürümlerinde kullanımı karışıklığa yol açmamalı veya Microsoft sponsorluğunu ima etmemelidir. Üçüncü taraf ticarî marka veya logoların kullanımı ise ilgili üçüncü taraf politikalarına tabidir. +Bu proje, projeler, ürünler veya hizmetler için ticari markalar veya logolar içerebilir. Microsoft ticari markalarının veya logolarının yetkili kullanımı, [Microsoft'un Ticari Marka ve Marka Yönergeleri](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) koşullarına uygun olmalıdır. +Bu projenin değiştirilmiş sürümlerinde Microsoft ticari markalarının veya logolarının kullanımı karışıklığa yol açmamalı veya Microsoft sponsorluğunu ima etmemelidir. Üçüncü taraf ticari markalarının veya logolarının kullanımı, ilgili üçüncü taraf politikalarına tabidir. -## Yardım Alma +## Yardım Almak -Eğer takılırsanız veya AI uygulamalar geliştirme hakkında sorularınız olursa, katılın: +Takıldığınızda veya AI uygulamaları geliştirme ile ilgili sorularınız olduğunda şu kanallara katılabilirsiniz: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Eğer ürün geri bildirimi vermek veya yapım sırasında hatalarla karşılaşırsanız ziyaret edin: +Ürün geri bildirimi veya hata bildirimi için: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Feragatname**: -Bu belge, AI çeviri servisi [Co-op Translator](https://github.com/Azure/co-op-translator) kullanılarak çevrilmiştir. Doğruluk için çaba göstermemize rağmen, otomatik çevirilerin hata veya yanlışlık içerebileceğini lütfen unutmayın. Orijinal belge, kendi ana dilindeki haliyle yetkili kaynak olarak kabul edilmelidir. Kritik bilgiler için profesyonel insan çevirisi önerilir. Bu çevirinin kullanımı sonucunda ortaya çıkabilecek yanlış anlaşılma veya yanlış yorumlamalardan sorumlu değiliz. +**Feragatname**: +Bu belge, AI çeviri hizmeti [Co-op Translator](https://github.com/Azure/co-op-translator) kullanılarak çevrilmiştir. Doğruluk için çaba sarf etsek de, otomatik çevirilerin hata veya yanlışlık içerebileceğini lütfen unutmayınız. Orijinal belge, kendi dilinde yetkili kaynak olarak kabul edilmelidir. Kritik bilgiler için profesyonel insan çevirisi önerilir. Bu çevirinin kullanımı sonucu ortaya çıkabilecek yanlış anlamalardan veya yanlış yorumlamalardan sorumlu değiliz. \ No newline at end of file diff --git a/translations/tr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/tr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..553ff1a8d --- /dev/null +++ b/translations/tr/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Microsoft phi-4-mini & SQLite FTS5 ile Hibrit RAG (Sıfır-Bulut SLM)\n", + "\n", + "> **Yazar:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Odak:** Küçük Dil Modelleri (SLM'ler), SQLite FTS5 BM25, Yoğun Gömmeler, Karşılıklı Sıralama Birleşimi (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Motivasyon: Yerel SLM'lerde Anahtar Kelime Hatırlama İkilemi\n", + "Yoğun vektör gömmelerine dayanan standart RAG mimarileri, genellikle tam sayısal tokenleri (örneğin, `2,340,000 TL`, sözleşme kodları, hesap numaraları) doğru şekilde elde edemez. \n", + "Öte yandan, seyrek sözcüksel arama (BM25) anlamsal eşanlamlılar ve dolaylı ifadeleri kaçırır.\n", + "\n", + "Bu tarif kitabı, şunu nasıl uygulayacağınızı gösterir: **yüksek hızlı, bellekte çalışan Hibrit Erişim motoru** şu birleşimi içeren:\n", + "1. **Yoğun Vektörler** (Kosine Benzerliği)\n", + "2. **Seyrek Sözcüksel Arama** (SQLite FTS5 BM25)\n", + "3. **Karşılıklı Sıralama Birleşimi (RRF, $k=60$)**\n", + "4. Microsoft `phi-4-mini` ile **Temellendirilmiş Atıf Üretimi (`[1]`, `[2]`)**\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Çift SQLite Şeması (Yoğun Vektörler + Sanal FTS5 BM25 Tablosu)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Örnek Alma ve Çalıştırma Karşılaştırması\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Microsoft phi-4-mini için Temellendirilmiş İstem Formülasyonu\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Feragatname**:\nBu belge, AI çeviri hizmeti [Co-op Translator](https://github.com/Azure/co-op-translator) kullanılarak çevrilmiştir. Doğruluk için çaba sarf etsek de, otomatik çevirilerin hata veya yanlışlık içerebileceğini lütfen unutmayınız. Orijinal belge, kendi dilinde yetkili kaynak olarak kabul edilmelidir. Kritik bilgiler için profesyonel insan çevirisi önerilir. Bu çevirinin kullanımı sonucu ortaya çıkabilecek yanlış anlamalardan veya yanlış yorumlamalardan sorumlu değiliz.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/uk/.co-op-translator.json b/translations/uk/.co-op-translator.json index 5f7d7f883..89cb8716d 100644 --- a/translations/uk/.co-op-translator.json +++ b/translations/uk/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "uk" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:23:28+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T16:45:14+00:00", "source_file": "README.md", "language_code": "uk" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "uk" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T16:50:26+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "uk" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:32:33+00:00", diff --git a/translations/uk/README.md b/translations/uk/README.md index ceedae4e0..52d2bfe29 100644 --- a/translations/uk/README.md +++ b/translations/uk/README.md @@ -1,43 +1,42 @@ -# Phi Cookbook: Практичні приклади з моделями Phi від Microsoft +# Кулінарна книга Phi: Практичні приклади з моделями Phi від Microsoft [![Відкрити та використовувати приклади в GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Відкрити в Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![Учасники GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![Проблеми GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Запити на злиття GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![Ласкаво просимо PR](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Запити на Pull GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![PRs Ласкаво просимо](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![Спостерігачі GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![Форуки GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![Форки GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![Зірки GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) -[![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) +[![Discord Microsoft Foundry](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi — це серія моделей штучного інтелекту з відкритим кодом, розроблена Microsoft. +Phi — це серія відкритих AI-моделей, розроблених Microsoft. -Phi наразі є найпотужнішою та найефективнішою за вартістю малою мовною моделлю (SLM), з дуже добрими показниками у багатомовності, логічному мисленні, генерації тексту/чату, програмуванні, роботі з зображеннями, аудіо та інших сценаріях. +На сьогодні Phi є найпотужнішою та найефективнішою за вартістю маленькою мовною моделлю (SLM), з дуже хорошими результатами у багатомовних завданнях, логічному мисленні, генерації тексту/чатів, кодуванні, роботі з зображеннями, аудіо та інших сценаріях. -Ви можете розгортати Phi у хмарі або на граничних пристроях, а також легко створювати генеративні AI-додатки з обмеженою обчислювальною потужністю. +Ви можете розгорнути Phi в хмарі або на пристроях на межі мережі, а також легко створювати генеративні AI-застосунки з обмеженими обчислювальними ресурсами. -Виконайте ці кроки, щоб почати використовувати ці ресурси: - -1. **Форкніть репозиторій**: Натисніть [![Форкнути](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +Виконайте ці кроки, щоб почати працювати з цими ресурсами: +1. **Форкніть репозиторій**: Натисніть [![Форки GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. **Клонуйте репозиторій**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**Приєднуйтесь до спільноти Microsoft AI Discord і зустрічайте експертів та інших розробників**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**Приєднуйтесь до спільноти Microsoft AI Discord та спілкуйтесь з експертами й іншими розробниками**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/uk/cover.eb18d1b9605d754b.webp) -### 🌐 Підтримка багатомовності +### 🌐 Підтримка багатьох мов -#### Підтримується через GitHub Action (автоматично та завжди актуально) +#### Підтримується через GitHub Action (Автоматично та завжди актуально) -[Арабська](../ar/README.md) | [Бенгальська](../bn/README.md) | [Болгарська](../bg/README.md) | [Бірманська (М’янма)](../my/README.md) | [Китайська (спрощена)](../zh-CN/README.md) | [Китайська (традиційна, Гонконг)](../zh-HK/README.md) | [Китайська (традиційна, Макао)](../zh-MO/README.md) | [Китайська (традиційна, Тайвань)](../zh-TW/README.md) | [Хорватська](../hr/README.md) | [Чеська](../cs/README.md) | [Данська](../da/README.md) | [Голландська](../nl/README.md) | [Естонська](../et/README.md) | [Фінська](../fi/README.md) | [Французька](../fr/README.md) | [Німецька](../de/README.md) | [Грецька](../el/README.md) | [Іврит](../he/README.md) | [Гінді](../hi/README.md) | [Угорська](../hu/README.md) | [Індонезійська](../id/README.md) | [Італійська](../it/README.md) | [Японська](../ja/README.md) | [Каннада](../kn/README.md) | [Кхмерська](../km/README.md) | [Корейська](../ko/README.md) | [Литовська](../lt/README.md) | [Малайська](../ms/README.md) | [Малаялам](../ml/README.md) | [Маратхі](../mr/README.md) | [Непальська](../ne/README.md) | [Нігерійський Піджин](../pcm/README.md) | [Норвезька](../no/README.md) | [Перська (Фарсі)](../fa/README.md) | [Польська](../pl/README.md) | [Португальська (Бразилія)](../pt-BR/README.md) | [Португальська (Португалія)](../pt-PT/README.md) | [Панджабі (Гурмухі)](../pa/README.md) | [Румунська](../ro/README.md) | [Російська](../ru/README.md) | [Сербська (кирилиця)](../sr/README.md) | [Словацька](../sk/README.md) | [Словенська](../sl/README.md) | [Іспанська](../es/README.md) | [Суахілі](../sw/README.md) | [Шведська](../sv/README.md) | [Тагалог (Філіппінська)](../tl/README.md) | [Тамільська](../ta/README.md) | [Телугу](../te/README.md) | [Тайська](../th/README.md) | [Турецька](../tr/README.md) | [Українська](./README.md) | [Урду](../ur/README.md) | [В’єтнамська](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](../zh-MO/README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](./README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **Віддаєте перевагу клонувати локально?** +> **Віддаєте перевагу локальному клонуванню?** > -> Цей репозиторій включає понад 50 перекладів мовами, що значно збільшує розмір завантаження. Щоб клонувати без перекладів, використовуйте локальне вибіркове завантаження: +> У цьому репозиторії є понад 50 мовних перекладів, що значно збільшує розмір завантаження. Щоб клонувати без перекладів, використовуйте sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -53,23 +52,23 @@ Phi наразі є найпотужнішою та найефективнішо > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Це дає все, що потрібно для проходження курсу, із значно швидшим завантаженням. +> Це дає вам усе необхідне для проходження курсу з набагато швидшим завантаженням. ## Зміст - Вступ - - [Ласкаво просимо до родини Phi](./md/01.Introduction/01/01.PhiFamily.md) + - [Ласкаво просимо до сімейства Phi](./md/01.Introduction/01/01.PhiFamily.md) - [Налаштування вашого середовища](./md/01.Introduction/01/01.EnvironmentSetup.md) - [Розуміння ключових технологій](./md/01.Introduction/01/01.Understandingtech.md) - - [Безпека AI для моделей Phi](./md/01.Introduction/01/01.AISafety.md) - - [Підтримка обладнання Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Моделі Phi та доступність на різних платформах](./md/01.Introduction/01/01.Edgeandcloud.md) - - [Використання Guidance-ai і Phi](./md/01.Introduction/01/01.Guidance.md) - - [Моделі в GitHub Marketplace](https://github.com/marketplace/models) + - [Безпека ШІ для моделей Phi](./md/01.Introduction/01/01.AISafety.md) + - [Підтримка апаратного забезпечення Phi](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Моделі Phi та їх доступність на різних платформах](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Використання Guidance-ai та Phi](./md/01.Introduction/01/01.Guidance.md) + - [Моделі на GitHub Marketplace](https://github.com/marketplace/models) - [Каталог моделей Azure AI](https://ai.azure.com) -- Інференс Phi у різних середовищах +- Запуск Phi в різних середовищах - [Hugging face](./md/01.Introduction/02/01.HF.md) - [Моделі GitHub](./md/01.Introduction/02/02.GitHubModel.md) - [Каталог моделей Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) @@ -78,177 +77,180 @@ Phi наразі є найпотужнішою та найефективнішо - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Інференс родини Phi - - [Інференс Phi на iOS](./md/01.Introduction/03/iOS_Inference.md) - - [Інференс Phi на Android](./md/01.Introduction/03/Android_Inference.md) - - [Інференс Phi на Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Інференс Phi на AI ПК](./md/01.Introduction/03/AIPC_Inference.md) - - [Інференс Phi з Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) - - [Інференс Phi на локальному сервері](./md/01.Introduction/03/Local_Server_Inference.md) - - [Інференс Phi на віддаленому сервері за допомогою AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) - - [Інференс Phi з Rust](./md/01.Introduction/03/Rust_Inference.md) - - [Інференс Phi--Vision локально](./md/01.Introduction/03/Vision_Inference.md) - - [Інференс Phi з Kaito AKS, Azure Containers (офіційна підтримка)](./md/01.Introduction/03/Kaito_Inference.md) -- [Квантифікація родини Phi](./md/01.Introduction/04/QuantifyingPhi.md) - - [Квантизація Phi-3.5 / 4 за допомогою llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Квантизація Phi-3.5 / 4 за допомогою розширень Generative AI для onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Квантизація Phi-3.5 / 4 за допомогою Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Квантизація Phi-3.5 / 4 за допомогою Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Оцінювання Phi - - [Відповідальний AI](./md/01.Introduction/05/ResponsibleAI.md) - - [Microsoft Foundry для оцінювання](./md/01.Introduction/05/AIFoundry.md) - - [Використання Promptflow для оцінювання](./md/01.Introduction/05/Promptflow.md) +- Запуск сімейства Phi + - [Запуск Phi на iOS](./md/01.Introduction/03/iOS_Inference.md) + - [Запуск Phi на Android](./md/01.Introduction/03/Android_Inference.md) + - [Запуск Phi на Jetson](./md/01.Introduction/03/Jetson_Inference.md) + - [Запуск Phi на AI PC](./md/01.Introduction/03/AIPC_Inference.md) + - [Запуск Phi з Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Запуск Phi на локальному сервері](./md/01.Introduction/03/Local_Server_Inference.md) + - [Запуск Phi на віддаленому сервері з AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Запуск Phi із Rust](./md/01.Introduction/03/Rust_Inference.md) + - [Запуск Phi--Vision локально](./md/01.Introduction/03/Vision_Inference.md) + - [Запуск Phi з Kaito AKS, Azure Containers (офіційна підтримка)](./md/01.Introduction/03/Kaito_Inference.md) +- [Квантифікація сімейства Phi](./md/01.Introduction/04/QuantifyingPhi.md) + - [Квантування Phi-3.5 / 4 за допомогою llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Квантування Phi-3.5 / 4 з розширеннями Generative AI для onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Квантування Phi-3.5 / 4 з Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Квантування Phi-3.5 / 4 з Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- Оцінка Phi + - [Відповідальний ШІ](./md/01.Introduction/05/ResponsibleAI.md) + - [Microsoft Foundry для оцінки](./md/01.Introduction/05/AIFoundry.md) + - [Використання Promptflow для оцінки](./md/01.Introduction/05/Promptflow.md) - RAG з Azure AI Search - - [Як використовувати Phi-4-mini та Phi-4-multimodal (RAG) з Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Як використовувати Phi-4-mini та Phi-4-мультимодальний (RAG) з Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Zero-Cloud локальний гібридний RAG з SQLite FTS5 та phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Приклади розробки додатків Phi - - Текстові та чат-додатки - - Phi-4 Приклади - - [📓] [Чат з Phi-4-mini ONNX моделлю](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) +- Зразки розробки застосунків Phi + - Текстові та чат-застосунки + - Зразки Phi-4 + - [📓] [Чат з моделлю Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - [Чат з локальною моделлю Phi-4 ONNX .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Чат-консольний додаток .NET з Phi-4 ONNX за допомогою Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 Приклади - - [Локальний чатбот у браузері з використанням Phi3, ONNX Runtime Web та WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [Консольний чат-додаток .NET з Phi-4 ONNX за допомогою Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Зразки + - [Локальний чат-бот у браузері з використанням Phi3, ONNX Runtime Web та WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Multi Model - Інтерактивний Phi-3-mini та OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Створення обгортки та використання Phi-3 з MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Мульти модель - інтерактивний Phi-3-mini та OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Створення оболонки та використання Phi-3 з MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - [Оптимізація моделі - Як оптимізувати модель Phi-3-min для ONNX Runtime Web за допомогою Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [Додаток WinUI3 з Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Зразок додатку WinUI3 для нотаток з ШІ на основі кількох моделей](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Детюнінг та інтеграція власних моделей Phi-3 з Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Детюнінг та інтеграція власних моделей Phi-3 з Prompt flow у Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Оцінка детюнінгу моделі Phi-3 / Phi-3.5 у Microsoft Foundry з акцентом на принципи відповідального ШІ Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [WinUI3 додаток з Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 Мульти модель AI Powered Notes App Sample](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Тонке налаштування та інтеграція кастомних моделей Phi-3 з Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Тонке налаштування та інтеграція кастомних моделей Phi-3 з Prompt flow у Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Оцінка тонко налаштованих моделей Phi-3 / Phi-3.5 у Microsoft Foundry з акцентом на принципи Відповідального AI від Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Приклад прогнозування мови Phi-3.5-mini-instruct (китайська/англійська)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG чатбот](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [Phi-3.5-Instruct WebGPU RAG Чат-бот](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Використання GPU Windows для створення рішення Prompt flow з Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [Використання Microsoft Phi-3.5 tflite для створення Android додатку](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Приклад Q&A .NET з локальною ONNX моделлю Phi-3 за допомогою Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Використання Microsoft Phi-3.5 tflite для створення Android-додатку](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Приклад Q&A .NET з використанням локальної ONNX моделі Phi-3 через Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - [Консольний чат .NET додаток з Semantic Kernel та Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI Inference SDK Код прикладів - - Phi-4 Приклади - - [📓] [Генерація коду проекту з Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 Приклади - - [Створіть власний чат GitHub Copilot для Visual Studio Code з Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Створіть власного агента для Visual Studio Code Chat Copilot з Phi-3.5 за моделями GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - Розширені приклади логіки - - Phi-4 Приклади - - [📓] [Phi-4-mini-reasoning або Phi-4-reasoning приклади](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [Детюнінг Phi-4-mini-reasoning з Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Детюнінг Phi-4-mini-reasoning з Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning з моделями GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning з моделями Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - Приклади коду Azure AI Inference SDK + - Зразки Phi-4 + - [📓] [Генерація коду проекту з використанням Phi-4-мультимодального](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - Зразки Phi-3 / 3.5 + - [Створіть власний чат Copilot GitHub у Visual Studio Code з Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Створіть власного агента Chat Copilot для Visual Studio Code з Phi-3.5 за моделями GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - Приклади розширеного мислення + - Зразки Phi-4 + - [📓] [Phi-4-mini-мислення або Phi-4-мислення зразки](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [Тонке налаштування Phi-4-mini-мислення з Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Тонке налаштування Phi-4-mini-мислення з Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Phi-4-mini-мислення з моделями GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [Phi-4-mini-мислення з моделями Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - Демонстрації - - [Phi-4-mini демонстрації на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-мультимодальні демонстрації на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Vision Приклади - - Phi-4 Приклади - - [📓] [Використання Phi-4-multimodal для читання зображень та генерації коду](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 Приклади - - [📓][Phi-3-vision-Зображення текст у текст](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [Phi-4-mini демо, розміщені на Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-мультимодальні демо, розміщені на Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Приклади зору + - Зразки Phi-4 + - [📓] [Використання Phi-4-мультимодального для читання зображень і генерації коду](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - Зразки Phi-3 / 3.5 + - [📓][Phi-3-vision-Перетворення тексту на текст із зображенням](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-vision CLIP embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision CLIP Embedding](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [DEMO: Phi-3 Переробка](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - Візуальний мовний асистент - з Phi3-Vision та OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3-vision - візуальний мовний асистент - з Phi3-Vision та OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Vision мультифреймовий або мультизображень зразок](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 Vision локальна ONNX модель з використанням Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Меню для Phi-3 Vision Локальна ONNX модель з використанням Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - - Приклади Reasoning-Vision - - Phi-4-Reasoning-Vision-15B - - [📓] [Використання Phi-4-Reasoning-Vision-15B для виявлення пішоходів поза переходом](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Використання Phi-4-Reasoning-Vision-15B для математики](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Використання Phi-4-Reasoning-Vision-15B для виявлення UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - - Математичні приклади - - Phi-4-Mini-Flash-Reasoning-Instruct приклади [Демо математики з Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) - - - Аудіо приклади - - Phi-4 Приклади - - [📓] [Витяг аудіотранскриптів за допомогою Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal аудіозразок](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal зразок перекладу мовлення](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET консольний додаток за допомогою Phi-4-multimodal аудіо для аналізу аудіофайлу та генерації транскрипту](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [📓][Phi-3.5 Vision мультифрейм або мультизображення приклад](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 Vision Локальна ONNX модель з використанням Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) + - [Меню на основі Phi-3 Vision Локальна ONNX модель з використанням Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + + - Приклади мислення-зором + - Phi-4-Мислення-Зор-15B + - [📓] [Використання Phi-4-Мислення-Зор-15B для виявлення порушень правил переходу в недозволеному місці](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Використання Phi-4-Мислення-Зор-15B для математики](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Використання Phi-4-Мислення-Зор-15B для виявлення UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + + - Приклади математики + - Зразки Phi-4-Mini-Flash-Reasoning-Instruct [Математична демонстрація з Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + + - Приклади аудіо + - Зразки Phi-4 + - [📓] [Витяг аудіотранскриптів з Phi-4-мультимодального](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-мультимодальний аудіо приклад](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-мультимодальний приклад перекладу мови](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET консольний додаток з використанням Phi-4-мультимодального аудіо для аналізу аудіофайлу та генерації транскрипту](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE Приклади - - Phi-3 / 3.5 Приклади - - [📓] [Phi-3.5 Модель суміші експертів (MoEs) для соціальних мереж](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Побудова pipeline генерації з доповненням пошуком (RAG) з NVIDIA NIM Phi-3 MOE, Azure AI Search та LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - Зразки Phi-3 / 3.5 + - [📓] [Phi-3.5 Моделі змішання експертів (MoEs) приклад з соціальних мереж](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Створення конвеєра із генерацією з використанням RAG, NVIDIA NIM Phi-3 MOE, Azure AI Search та LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Приклади Function Calling - - Phi-4 Приклади 🆕 - - [📓] [Використання Function Calling з Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Використання Function Calling для створення мультиагентів з Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [Використання Function Calling з Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [Використання Function Calling з ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Приклади Multimodal Mixing - - Phi-4 Приклади 🆕 - - [📓] [Використання Phi-4-multimodal як технологічного журналіста](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET консольний додаток за допомогою Phi-4-multimodal для аналізу зображень](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- Детюнінг Phi Прикладів - - [Сценарії детюнінгу](./md/03.FineTuning/FineTuning_Scenarios.md) - - [Детюнінг vs RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [Детюнінг: нехай Phi-3 стане експертом у галузі](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Детюнінг Phi-3 за допомогою AI Toolkit для VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Детюнінг Phi-3 з Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - - [Детюнінг Phi-3 з Lora](./md/03.FineTuning/FineTuning_Lora.md) - - [Детюнінг Phi-3 з QLora](./md/03.FineTuning/FineTuning_Qlora.md) - - [Детюнінг Phi-3 з Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [Детюнінг Phi-3 з Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - - [Детюнінг з Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Практична лабораторія детюнінгу з Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - - [Детюнінг Phi-3-vision з Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Детюнінг Phi-3 з Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) - - [Офіційна підтримка детюнінгу Phi-3-vision](./md/03.FineTuning/FineTuning_Vision.md) + - Функціональні зразки виклику + - Зразки Phi-4 🆕 + - [📓] [Використання виклику функції з Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Використання виклику функції для створення мультиагентів з Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Використання виклику функції з Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Використання виклику функції з ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - Приклади змішування мультимодальних моделей + - Зразки Phi-4 🆕 + - [📓] [Використання Phi-4-мультимодального як технологічного журналіста](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET консольний додаток з використанням Phi-4-мультимодального для аналізу зображень](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- Тонке налаштування Phi зразків + - [Сценарії тонкого налаштування](./md/03.FineTuning/FineTuning_Scenarios.md) + - [Тонке налаштування проти RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Тонке налаштування: нехай Phi-3 стане галузевим експертом](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [Тонке налаштування Phi-3 з AI Toolkit для VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Тонке налаштування Phi-3 з Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) + - [Тонке налаштування Phi-3 з Lora](./md/03.FineTuning/FineTuning_Lora.md) + - [Тонке налаштування Phi-3 з QLora](./md/03.FineTuning/FineTuning_Qlora.md) + - [Тонке налаштування Phi-3 з Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Тонке налаштування Phi-3 з Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Тонке налаштування з Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Тонке налаштування з Microsoft Olive Hands-On Lab](./md/03.FineTuning/olive-lab/readme.md) + - [Тонке налаштування Phi-3-vision з Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [Тонке налаштування Phi-3 за допомогою Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Тонке налаштування Phi-3-vision (офіційна підтримка)](./md/03.FineTuning/FineTuning_Vision.md) - [Тонке налаштування Phi-3 з Kaito AKS, Azure Containers (офіційна підтримка)](./md/03.FineTuning/FineTuning_Kaito.md) - [Тонке налаштування Phi-3 та 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - Практична лабораторія - - [Вивчення найсучасніших моделей: LLM, SLM, локальна розробка та інше](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [Вивчення новітніх моделей: LLMs, SLMs, локальна розробка та інше](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - [Розкриття потенціалу NLP: тонке налаштування з Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Академічні наукові статті та публікації - - [Textbooks Are All You Need II: технічний звіт phi-1.5](https://arxiv.org/abs/2309.05463) - - [Phi-3 Technical Report: високо здібна мовна модель локально на вашому телефоні](https://arxiv.org/abs/2404.14219) - - [Phi-4 Technical Report](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini Technical Report: компактні, але потужні багатомодальні мовні моделі за допомогою Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Оптимізація маленьких мовних моделей для виклику функцій у транспортних засобах](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Тонке налаштування PHI-3 для множинного вибору питань: методологія, результати та виклики](https://arxiv.org/abs/2501.01588) - - [Phi-4-reasoning Technical Report](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini-reasoning Technical Report](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) +- Наукові дослідні роботи та публікації + - [Підручники — все, що потрібно II: технічний звіт phi-1.5](https://arxiv.org/abs/2309.05463) + - [Технічний звіт Phi-3: високо здібна мовна модель локально на вашому телефоні](https://arxiv.org/abs/2404.14219) + - [Технічний звіт Phi-4](https://arxiv.org/abs/2412.08905) + - [Технічний звіт Phi-4-Mini: компактні, але потужні мультимодальні мовні моделі за допомогою Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Оптимізація малих мовних моделей для виклику функцій у транспортному засобі](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Тонке налаштування PHI-3 для відповідей на запитання з вибором відповіді: методологія, результати та виклики](https://arxiv.org/abs/2501.01588) + - [Технічний звіт Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Технічний звіт Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## Використання моделей Phi ### Phi на Microsoft Foundry -Ви можете дізнатися, як використовувати Microsoft Phi та як створювати E2E рішення на різних апаратних пристроях. Щоб самостійно спробувати Phi, почніть з роботи з моделями та налаштуйте Phi для своїх сценаріїв, використовуючи [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Більше інформації доступно в розділі Початок роботи з [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Ви можете навчитися користуватися Microsoft Phi та створювати комплексні рішення для різних апаратних пристроїв. Щоб самостійно випробувати Phi, почніть гратися з моделями та налаштовувати Phi для своїх сценаріїв за допомогою [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai), докладніше можна дізнатися в посібнику Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) **Плейграунд** -Кожна модель має власний плейграунд для тестування моделі [Azure AI Playground](https://aka.ms/try-phi3). +Кожна модель має власний майданчик для тестування - [Azure AI Playground](https://aka.ms/try-phi3). ### Phi на GitHub Models -Ви можете дізнатися, як використовувати Microsoft Phi та як створювати E2E рішення на різних апаратних пристроях. Щоб самостійно спробувати Phi, почніть з роботи з моделлю та налаштуйте Phi для своїх сценаріїв, використовуючи [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Більше інформації доступно в розділі Початок роботи з [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Ви можете навчитися користуватися Microsoft Phi та створювати комплексні рішення для різних апаратних пристроїв. Щоб самостійно випробувати Phi, почніть гратися з моделлю та налаштовувати Phi для своїх сценаріїв за допомогою [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo), докладніше можна дізнатися в посібнику Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) **Плейграунд** -Кожна модель має власний [плейграунд для тестування моделі](/md/02.QuickStart/GitHubModel_QuickStart.md). +Кожна модель має присвячений [плейграунд для тестування моделі](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi на Hugging Face -Ви також можете знайти модель на [Hugging Face](https://huggingface.co/microsoft) +Також модель можна знайти на [Hugging Face](https://huggingface.co/microsoft) **Плейграунд** -[Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) + [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Інші курси + ## 🎒 Інші курси -Наша команда створює інші курси! Ознайомтеся: +Наша команда також створює інші курси! Ознайомтеся: ### LangChain @@ -257,26 +259,26 @@ Phi наразі є найпотужнішою та найефективнішо [![LangChain для початківців](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agents +### Azure / Edge / MCP / Агенти [![AZD для початківців](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![Edge AI для початківців](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![MCP для початківців](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents для початківців](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Агенти для початківців](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### Серія генеративного ШІ -[![Генеративний ШІ для початківців](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Генеративний ШІ (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Генеративний ШІ (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Генеративний ШІ (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### Серія Generative AI +[![Generative AI для початківців](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- -### Основне навчання -[![Машинне навчання для початківців](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Наука про дані для початківців](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![ШІ для початківців](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +### Основи навчання +[![ML для початківців](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science для початківців](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI для початківців](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![Кібербезпека для початківців](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) [![Веб-розробка для початківців](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![IoT для початківців](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) @@ -285,41 +287,43 @@ Phi наразі є найпотужнішою та найефективнішо --- ### Серія Copilot -[![Copilot для парного програмування з AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot для AI парного програмування](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) [![Copilot для C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot Пригоди](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## Відповідальний ШІ +## Відповідальний ШІ + +Microsoft прагне допомогти своїм клієнтам відповідально використовувати продукти ШІ, ділитися набутим досвідом і будувати партнерства на основі довіри за допомогою інструментів, таких як Transparency Notes та Impact Assessments. Багато з цих ресурсів можна знайти за адресою [https://aka.ms/RAI](https://aka.ms/RAI). +Підхід Microsoft до відповідального ШІ ґрунтується на наших принципах ШІ: справедливість, надійність і безпека, конфіденційність і безпека, інклюзивність, прозорість і підзвітність. + +Великі моделі природної мови, зображень і мовлення — як ті, що використовуються у цьому прикладі — можуть потенційно поводитися несправедливо, ненадійно або образливо, що може спричинити шкоду. Будь ласка, ознайомтеся з [Приміткою прозорості сервісу Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), щоб бути поінформованими про ризики та обмеження. -Microsoft прагне допомагати нашим клієнтам відповідально використовувати наші продукти ШІ, ділитися своїм досвідом і будувати партнерські відносини на основі довіри за допомогою інструментів, таких як Нотатки прозорості та Оцінки впливу. Багато з цих ресурсів можна знайти за адресою [https://aka.ms/RAI](https://aka.ms/RAI). -Підхід Microsoft до відповідального ШІ базується на наших принципах ШІ: справедливість, надійність і безпека, конфіденційність і безпека, інклюзивність, прозорість та підзвітність. -Великі масштаби моделей природної мови, зображень і мовлення — як ті, що використовуються в цьому прикладі — потенційно можуть поводитися несправедливо, ненадійно або образливо, що може призвести до шкоди. Будь ласка, ознайомтеся з [нотаткою прозорості служби Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text), щоб бути поінформованими про ризики та обмеження. -Рекомендуваним підходом пом'якшення цих ризиків є включення системи безпеки у вашу архітектуру, яка може виявляти та запобігати шкідливій поведінці. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) забезпечує незалежний рівень захисту, здатний виявляти шкідливий контент, створений користувачами та ШІ, у додатках і службах. Azure AI Content Safety включає API для тексту та зображень, які дозволяють виявляти шкідливий матеріал. У Microsoft Foundry служба Content Safety дозволяє переглядати, досліджувати та випробовувати прикладний код для виявлення шкідливого контенту у різних модальностях. Наступна [документація швидкого запуску](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) проведе вас через процес виконання запитів до служби. +Рекомендований підхід до пом'якшення цих ризиків полягає у включенні системи безпеки у вашу архітектуру, яка може виявляти та запобігати шкідливій поведінці. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) забезпечує незалежний рівень захисту, здатний виявляти шкідливий контент, створений користувачами та ШІ у додатках і сервісах. Azure AI Content Safety включає API для тексту та зображень, які дозволяють виявляти шкідливий матеріал. У межах Microsoft Foundry служба Content Safety дозволяє переглядати, досліджувати та випробовувати прикладний код для виявлення шкідливого контенту в різних модальностях. Наступна [документація швидкого старту](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) проводить вас через процес надсилання запитів до служби. -Ще одним аспектом, який слід враховувати, є загальна продуктивність додатку. У мульти-модальних і мульти-модельних додатках під продуктивністю розуміють те, що система працює так, як очікуєте ви і ваші користувачі, включно з тим, що вона не генерує шкідливі результати. Важливо оцінити продуктивність вашого загального додатку за допомогою [Оцінювачів продуктивності, якості, ризику та безпеки](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Ви також маєте можливість створювати та оцінювати за допомогою [кастомних оцінювачів](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Ще одним аспектом, який слід враховувати, є загальна продуктивність застосунку. У випадку мультимодальних та мультимодельних застосунків під продуктивністю ми розуміємо, що система працює так, як ви та ваші користувачі очікують, включно з тим, що вона не генерує шкідливий вихід. Важливо оцінити продуктивність вашого загального застосунку за допомогою [оцінювачів продуктивності, якості, ризику та безпеки](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in). Ви також маєте можливість створювати та оцінювати за допомогою [власних оцінювачів](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Ви можете оцінити свій ШІ-додаток у середовищі розробки за допомогою [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Враховуючи тестовий набір даних або ціль, генерації вашого генеративного ШІ-додатку кількісно вимірюються вбудованими або визначеними вами кастомними оцінювачами. Щоб почати працювати з Azure AI Evaluation SDK для оцінки вашої системи, ви можете скористатися [інструкцією швидкого запуску](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Після виконання запуску оцінювання ви можете [візуалізувати результати в Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Ви можете оцінити ваш AI-застосунок у вашому середовищі розробки, використовуючи [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). За наявності тестового набору даних або цілі, генерації вашого генеративного AI-застосунку кількісно вимірюються за допомогою вбудованих або власних оцінювачів на ваш вибір. Щоб розпочати роботу з azure ai evaluation sdk для оцінки вашої системи, ви можете слідувати за [інструкцією швидкого старту](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Після запуску оцінювання ви можете [візуалізувати результати в Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Торгові марки -Цей проєкт може містити торгові марки або логотипи проєктів, продуктів чи послуг. Авторизоване використання торгових марок або логотипів Microsoft підлягає дотриманню і повинно відповідати [Правилам використання торгових марок і брендів Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Використання торгових марок або логотипів Microsoft у змінених версіях цього проєкту не повинно викликати плутанину або створювати враження спонсорства Microsoft. Будь-яке використання торгових марок або логотипів третіх сторін підпорядковується політикам цих третіх сторін. +Цей проект може містити торгові марки або логотипи проектів, продуктів або сервісів. Дозволене використання торгових марок або логотипів Microsoft підпорядковується та має відповідати [Керівництву Microsoft щодо торгових марок і брендів](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Використання торгових марок або логотипів Microsoft у змінених версіях цього проекту не повинно викликати плутанину або натякати на спонсорство Microsoft. Будь-яке використання торгових марок або логотипів сторонніх організацій підпорядковується політикам цих сторонніх організацій. ## Отримання допомоги -Якщо у вас виникли труднощі або є питання щодо створення ШІ-додатків, приєднуйтеся: +Якщо ви застрягли або маєте запитання щодо створення AI-застосунків, приєднуйтесь: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Якщо у вас є відгуки про продукт або помилки під час створення, відвідайте: +Якщо у вас є відгуки про продукт або помилки під час розробки, відвідайте: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Відмова від відповідальності**: -Цей документ було перекладено за допомогою сервісу штучного інтелекту [Co-op Translator](https://github.com/Azure/co-op-translator). Хоч ми й докладаємо зусиль для забезпечення точності, просимо враховувати, що автоматичні переклади можуть містити помилки чи неточності. Оригінальний документ рідною мовою слід вважати авторитетним джерелом. Для критично важливої інформації рекомендується професійний людський переклад. Ми не несемо відповідальності за будь-які непорозуміння чи неправильні тлумачення, що виникли внаслідок використання цього перекладу. +**Відмова від відповідальності**: +Цей документ було перекладено за допомогою сервісу штучного інтелекту для перекладу [Co-op Translator](https://github.com/Azure/co-op-translator). Хоча ми прагнемо до точності, будь ласка, майте на увазі, що автоматичні переклади можуть містити помилки або неточності. Оригінальний документ рідною мовою слід вважати авторитетним джерелом. Для критично важливої інформації рекомендується професійний людський переклад. Ми не несемо відповідальності за будь-які непорозуміння або неправильні тлумачення, що виникли внаслідок використання цього перекладу. \ No newline at end of file diff --git a/translations/uk/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/uk/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..1dd2ca13a --- /dev/null +++ b/translations/uk/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 Гібридний RAG з Microsoft phi-4-mini та SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Автор:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Фокус:** Малі мовні моделі (SLMs), SQLite FTS5 BM25, Щільні вбудовування, Взаємна ранжирувальна фузія (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Мотивація: Дилема пошуку ключових слів у локальних SLM\n", + "Стандартні архітектури RAG, що базуються виключно на щільних векторних вбудовуваннях, часто не здатні витягнути точні числові токени (наприклад, `2,340,000 TL`, коди контрактів, номери рахунків). \n", + "Навпаки, розріджений лексичний пошук (BM25) не знайде семантичні синоніми та перефразовані запитання.\n", + "\n", + "У цьому посібнику показано, як реалізувати **швидкий гібридний движок пошуку в пам’яті**, який поєднує:\n", + "1. **Щільні вектори** (Косинусна схожість)\n", + "2. **Розріджений лексичний пошук** (SQLite FTS5 BM25)\n", + "3. **Взаємна ранжирувальна фузія (RRF, $k=60$)**\n", + "4. **Обґрунтоване генерація цитат (`[1]`, `[2]`)** з Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Подвійна схема SQLite (щільні вектори + віртуальна таблиця FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Приклад завантаження та оцінювання виконання\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Формулювання обґрунтованих промптів для Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Відмова від відповідальності**:\nЦей документ було перекладено за допомогою сервісу штучного інтелекту для перекладу [Co-op Translator](https://github.com/Azure/co-op-translator). Хоча ми прагнемо до точності, будь ласка, майте на увазі, що автоматичні переклади можуть містити помилки або неточності. Оригінальний документ рідною мовою слід вважати авторитетним джерелом. Для критично важливої інформації рекомендується професійний людський переклад. Ми не несемо відповідальності за будь-які непорозуміння або неправильні тлумачення, що виникли внаслідок використання цього перекладу.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/ur/.co-op-translator.json b/translations/ur/.co-op-translator.json index 7e36cf06a..b214966c7 100644 --- a/translations/ur/.co-op-translator.json +++ b/translations/ur/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "ur" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T07:43:47+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:50:42+00:00", "source_file": "README.md", "language_code": "ur" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "ur" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:53:35+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "ur" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:29:44+00:00", diff --git a/translations/ur/README.md b/translations/ur/README.md index 538e85a67..fa3b3351c 100644 --- a/translations/ur/README.md +++ b/translations/ur/README.md @@ -1,42 +1,42 @@ -# فی کک بک: مائیکروسافٹ کے فی ماڈلز کے ساتھ ہاتھوں پر مبنی مثالیں +# فی کلیاتی کتاب: مائیکروسافٹ کے فی ماڈلز کے ساتھ عملی مثالیں -[![گیٹ ہب کوڈ اسپیسز میں نمونے کھولیں اور استعمال کریں](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![گیٹ ہب کوڈسپیسز میں نمونے کھولیں اور استعمال کریں](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![ڈیولپمنٹ کنٹینرز میں کھولیں](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![گیٹ ہب شراکت دار](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![گیٹ ہب ایشوز](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![گیٹ ہب پول ریکویسٹ](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![گیٹ ہب میں تعاون کرنے والے](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![گیٹ ہب مسائل](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![گیٹ ہب پل ریکویسٹ](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![PRs خوش آمدید](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![گیٹ ہب دیکھنے والے](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![گیٹ ہب نگران](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![گیٹ ہب فورکس](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![گیٹ ہب اسٹارز](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![مائیکروسافٹ فاؤنڈری ڈسکارڈ](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -فی مائیکروسافٹ کی طرف سے تیار کردہ ایک سلسلہ ہے جو اوپن سورس AI ماڈلز پر مشتمل ہے۔ +فی ایک سلسلہ ہے اوپن سورس AI ماڈلز کا جو مائیکروسافٹ نے تیار کیے ہیں۔ -فی فی الوقت سب سے طاقتور اور کم لاگت والا چھوٹا زبان ماڈل (SLM) ہے، جس نے کئی زبانوں، منطق، متن/چیٹ تخلیق، کوڈنگ، تصاویر، آڈیو اور دیگر منظرناموں میں بہترین بیچ مارکس حاصل کیے ہیں۔ +فی اس وقت سب سے طاقتور اور کم لاگت والا چھوٹا زبان ماڈل (SLM) ہے، جس کے بہت اچھے معیار ہیں مختلف زبانوں میں، استدلال، متن/چیٹ تخلیق، کوڈنگ، تصاویر، آڈیو اور دیگر حالات میں۔ -آپ فی کو کلاؤڈ یا ایج ڈیوائسز پر تعینات کر سکتے ہیں، اور محدود کمپیوٹنگ طاقت کے ساتھ آسانی سے جنریٹو AI ایپلیکیشنز بنا سکتے ہیں۔ +آپ فی کو کلاؤڈ یا ایج ڈیوائسز پر تعینات کر سکتے ہیں، اور آپ آسانی سے محدود کمپیوٹنگ طاقت کے ساتھ جنریشن AI ایپلی کیشنز بنا سکتے ہیں۔ -ان وسائل کو استعمال شروع کرنے کے لیے درج ذیل مراحل پر عمل کریں: -1. **ریپوزٹری کو فورک کریں**: کلک کریں [![گیٹ ہب فورکس](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **ریپوزٹری کو کلون کریں**: `git clone https://github.com/microsoft/PhiCookBook.git` -3. [**مائیکروسافٹ AI ڈسکارڈ کمیونٹی میں شامل ہوں اور ماہرین و دیگر ڈویلپرز سے ملیں**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +ان وسائل کو استعمال شروع کرنے کے لئے یہ اقدامات کریں: +1. **ریپوزیٹری کو فورک کریں**: کلک کریں [![گیٹ ہب فورکس](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **ریپوزیٹری کو کلون کریں**: `git clone https://github.com/microsoft/PhiCookBook.git` +3. [**مائیکروسافٹ AI ڈسکارڈ کمیونٹی میں شامل ہوں اور ماہرین اور دیگر ڈویلپرز سے ملاقات کریں**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/ur/cover.eb18d1b9605d754b.webp) -### 🌐 متعدد زبانوں کی حمایت +### 🌐 کثیر الزبان معاونت #### GitHub ایکشن کے ذریعے معاونت (خودکار اور ہمیشہ تازہ ترین) -[عربی](../ar/README.md) | [بنگالی](../bn/README.md) | [بلغاریائی](../bg/README.md) | [برمی (میانمار)](../my/README.md) | [چینی (سادہ)](../zh-CN/README.md) | [چینی (روایتی، ہانگ کانگ)](../zh-HK/README.md) | [چینی (روایتی، میکاؤ)](../zh-MO/README.md) | [چینی (روایتی، تائیوان)](../zh-TW/README.md) | [کروشیائی](../hr/README.md) | [چیک](../cs/README.md) | [ڈینش](../da/README.md) | [ڈچ](../nl/README.md) | [ایسٹونین](../et/README.md) | [فنلش](../fi/README.md) | [فرانسیسی](../fr/README.md) | [جرمن](../de/README.md) | [یونانی](../el/README.md) | [عبرانی](../he/README.md) | [ہندی](../hi/README.md) | [ہنگیرین](../hu/README.md) | [انڈونیشین](../id/README.md) | [اطالوی](../it/README.md) | [جاپانی](../ja/README.md) | [کنڑ](../kn/README.md) | [خمیر](../km/README.md) | [کوریائی](../ko/README.md) | [لیتھوینین](../lt/README.md) | [ملائی](../ms/README.md) | [مالیالم](../ml/README.md) | [مراٹھی](../mr/README.md) | [نیپالی](../ne/README.md) | [نائجیریائی پیجین](../pcm/README.md) | [ناروے (ناروے زبان)](../no/README.md) | [فارسی (فارسی)](../fa/README.md) | [پولش](../pl/README.md) | [پرتگالی (برازیل)](../pt-BR/README.md) | [پرتگالی (پرتگال)](../pt-PT/README.md) | [پنجابی (گورمکھی)](../pa/README.md) | [رومانیائی](../ro/README.md) | [روسی](../ru/README.md) | [سربیائی (سریلیک)](../sr/README.md) | [سلوواک](../sk/README.md) | [سلووینیائی](../sl/README.md) | [ہسپانوی](../es/README.md) | [سواحلی](../sw/README.md) | [سویڈش](../sv/README.md) | [تاگالوگ (فلپائنی)](../tl/README.md) | [تمل](../ta/README.md) | [تیلگو](../te/README.md) | [تھائی](../th/README.md) | [ترکی](../tr/README.md) | [یوکرائنی](../uk/README.md) | [اردو](./README.md) | [ویتنامی](../vi/README.md) +[عربی](../ar/README.md) | [بنگالی](../bn/README.md) | [بلغاریائی](../bg/README.md) | [برمی (میانمار)](../my/README.md) | [چینی (سادہ)](../zh-CN/README.md) | [چینی (روایتی، ہانگ کانگ)](../zh-HK/README.md) | [چینی (روایتی، میکاؤ)](../zh-MO/README.md) | [چینی (روایتی، تائیوان)](../zh-TW/README.md) | [کروشیاوی](../hr/README.md) | [چیک](../cs/README.md) | [ڈینش](../da/README.md) | [ڈچ](../nl/README.md) | [ایسٹونین](../et/README.md) | [فنش](../fi/README.md) | [فرانسیسی](../fr/README.md) | [جرمن](../de/README.md) | [یونانی](../el/README.md) | [عبرانی](../he/README.md) | [ہندی](../hi/README.md) | [ہنگریائی](../hu/README.md) | [انڈونیشین](../id/README.md) | [اطالوی](../it/README.md) | [جاپانی](../ja/README.md) | [کنڑا](../kn/README.md) | [خمیر](../km/README.md) | [کوریائی](../ko/README.md) | [لتھونین](../lt/README.md) | [ملائی](../ms/README.md) | [ملایالم](../ml/README.md) | [مراٹھی](../mr/README.md) | [نیپالی](../ne/README.md) | [نائجیرین پیجین](../pcm/README.md) | [ناروے](../no/README.md) | [فارسی (فارسی)](../fa/README.md) | [پولش](../pl/README.md) | [پرتگالی (برازیل)](../pt-BR/README.md) | [پرتگالی (پرتگال)](../pt-PT/README.md) | [پنجابی (گرمکھی)](../pa/README.md) | [رومانیائی](../ro/README.md) | [روسی](../ru/README.md) | [سربیائی (سریلیک)](../sr/README.md) | [سلوواک](../sk/README.md) | [سلووینیائی](../sl/README.md) | [ہسپانوی](../es/README.md) | [سواحلی](../sw/README.md) | [سویڈش](../sv/README.md) | [تاگالوگ (فلپائنی)](../tl/README.md) | [تمل](../ta/README.md) | [تیلگو](../te/README.md) | [تھائی](../th/README.md) | [ترک](../tr/README.md) | [یوکرائنی](../uk/README.md) | [اردو](./README.md) | [ویتنامی](../vi/README.md) -> **لوکل کلوننگ کو ترجیح دیں؟** +> **کیا آپ لوکل کلون کرنا پسند کریں گے؟** > -> یہ ریپوزٹری 50+ زبانوں میں تراجم شامل کرتی ہے جو ڈاؤن لوڈ سائز کو نمایاں طور پر بڑھاتی ہے۔ بغیر تراجم کے کلون کرنے کے لیے، sparse checkout کا استعمال کریں: +> یہ ریپوزیٹری 50+ زبانوں کی ترجمے شامل ہے جو ڈاؤن لوڈ کے حجم میں نمایاں اضافہ کرتے ہیں۔ ترجمے کے بغیر کلون کرنے کے لیے، سپارس چیک آؤٹ استعمال کریں: > > **باش / میک او ایس / لینکس:** > ```bash @@ -45,225 +45,228 @@ > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` > -> **کمانڈ پرامپٹ (ونڈوز):** +> **CMD (ونڈوز):** > ```cmd > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> اس سے آپ کو تیزی سے ڈاؤن لوڈ کے ساتھ کورس مکمل کرنے کے لیے تمام ضروریات مل جاتی ہیں۔ +> اس سے آپ کو کورس مکمل کرنے کے لیے سب کچھ مل جاتا ہے اور ڈاؤن لوڈ بہت تیز ہو جاتا ہے۔ ## فہرست مضامین - تعارف - - [فی فیملی میں خوش آمدید](./md/01.Introduction/01/01.PhiFamily.md) + - [فی خاندان میں خوش آمدید](./md/01.Introduction/01/01.PhiFamily.md) - [اپنا ماحول ترتیب دینا](./md/01.Introduction/01/01.EnvironmentSetup.md) - [اہم ٹیکنالوجیز کو سمجھنا](./md/01.Introduction/01/01.Understandingtech.md) - - [فی ماڈلز کے لیے AI حفاظت](./md/01.Introduction/01/01.AISafety.md) + - [فی ماڈلز کے لیے AI سیفٹی](./md/01.Introduction/01/01.AISafety.md) - [فی ہارڈویئر سپورٹ](./md/01.Introduction/01/01.Hardwaresupport.md) - - [فی ماڈلز اور پلیٹ فارمز پر دستیابی](./md/01.Introduction/01/01.Edgeandcloud.md) - - [گائیڈنس-ai اور فی کا استعمال](./md/01.Introduction/01/01.Guidance.md) + - [پی اور کلاؤڈ کے درمیان فی ماڈلز اور دستیابی](./md/01.Introduction/01/01.Edgeandcloud.md) + - [گائیڈنس- ای آئی اور فی کا استعمال](./md/01.Introduction/01/01.Guidance.md) - [گیٹ ہب مارکیٹ پلیس ماڈلز](https://github.com/marketplace/models) - [ایزور AI ماڈل کیٹلاگ](https://ai.azure.com) -- مختلف ماحول میں فی کی انفرنس - - [ہیگنگ فیس](./md/01.Introduction/02/01.HF.md) +- مختلف ماحول میں فی کی پیشگوئی + - [ہگیگ فیس](./md/01.Introduction/02/01.HF.md) - [گیٹ ہب ماڈلز](./md/01.Introduction/02/02.GitHubModel.md) - [مائیکروسافٹ فاؤنڈری ماڈل کیٹلاگ](./md/01.Introduction/02/03.AzureAIFoundry.md) - [اولاما](./md/01.Introduction/02/04.Ollama.md) - [AI ٹول کٹ VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [فاؤنڈری لوکل](./md/01.Introduction/02/07.FoundryLocal.md) - -- فی فیملی کی انفرنس - - [iOS میں فی کی انفرنس](./md/01.Introduction/03/iOS_Inference.md) - - [Android میں فی کی انفرنس](./md/01.Introduction/03/Android_Inference.md) - - [Jetson میں فی کی انفرنس](./md/01.Introduction/03/Jetson_Inference.md) - - [AI پی سی میں فی کی انفرنس](./md/01.Introduction/03/AIPC_Inference.md) - - [ایپل MLX فریم ورک کے ساتھ فی کی انفرنس](./md/01.Introduction/03/MLX_Inference.md) - - [لوکل سرور میں فی کی انفرنس](./md/01.Introduction/03/Local_Server_Inference.md) - - [AI ٹول کٹ کا استعمال کرتے ہوئے ریموٹ سرور پر فی کی انفرنس](./md/01.Introduction/03/Remote_Interence.md) - - [رسٹ کے ساتھ فی کی انفرنس](./md/01.Introduction/03/Rust_Inference.md) - - [لوکل میں فی وژن کی انفرنس](./md/01.Introduction/03/Vision_Inference.md) - - [کائٹو AKS، ایزور کنٹینرز (سرکاری سپورٹ) کے ساتھ فی کی انفرنس](./md/01.Introduction/03/Kaito_Inference.md) -- [فی فیملی کو مقداری بنانا](./md/01.Introduction/04/QuantifyingPhi.md) - - [llama.cpp کے ذریعے Phi-3.5 / 4 کو مقداری بنانا](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [onnxruntime کے لیے جنریٹو AI ایکسٹینشنز کے ذریعے Phi-3.5 / 4 کو مقداری بنانا](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [انٹل اوپن وی نو کے ذریعے Phi-3.5 / 4 کو مقداری بنانا](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [ایپل MLX فریم ورک کے ذریعے Phi-3.5 / 4 کو مقداری بنانا](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- فی کی جانچ - - [ردعمل AI](./md/01.Introduction/05/ResponsibleAI.md) - - [تجزیے کے لیے مائیکروسافٹ فاؤنڈری](./md/01.Introduction/05/AIFoundry.md) - - [جانچ کے لیے Promptflow کا استعمال](./md/01.Introduction/05/Promptflow.md) + - [لوکل فاؤنڈری](./md/01.Introduction/02/07.FoundryLocal.md) + +- فی خاندان کی پیشگوئی + - [iOS میں فی کی پیشگوئی](./md/01.Introduction/03/iOS_Inference.md) + - [اینڈرائیڈ میں فی کی پیشگوئی](./md/01.Introduction/03/Android_Inference.md) + - [جیٹسن میں فی کی پیشگوئی](./md/01.Introduction/03/Jetson_Inference.md) + - [AI پی سی میں فی کی پیشگوئی](./md/01.Introduction/03/AIPC_Inference.md) + - [ایپل MLX فریم ورک کے ساتھ فی کی پیشگوئی](./md/01.Introduction/03/MLX_Inference.md) + - [لوکل سرور میں فی کی پیشگوئی](./md/01.Introduction/03/Local_Server_Inference.md) + - [AI ٹول کٹ استعمال کرتے ہوئے ریموٹ سرور پر فی کی پیشگوئی](./md/01.Introduction/03/Remote_Interence.md) + - [رسٹ کے ساتھ فی کی پیشگوئی](./md/01.Introduction/03/Rust_Inference.md) + - [لوکل میں فی وژن کی پیشگوئی](./md/01.Introduction/03/Vision_Inference.md) + - [کایٹو AKS، ایزور کنٹینرز (سرکاری حمایت) کے ساتھ فی کی پیشگوئی](./md/01.Introduction/03/Kaito_Inference.md) +- [فی خاندان کی مقداری جانچ](./md/01.Introduction/04/QuantifyingPhi.md) + - [لاما.cpp کے ذریعے فی-3.5 / 4 کی مقداری جانچ](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [onnxruntime کے لیے جنریٹو AI توسیعات کے ذریعے فی-3.5 / 4 کی مقداری جانچ](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [انٹیل اوپن وی نو کے ذریعے فی-3.5 / 4 کی مقداری جانچ](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [ایپل MLX فریم ورک کے ذریعے فی-3.5 / 4 کی مقداری جانچ](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- فی کی تشخیص + - [ذمہ دار AI](./md/01.Introduction/05/ResponsibleAI.md) + - [مائیکروسافٹ فاؤنڈری برائے تشخیص](./md/01.Introduction/05/AIFoundry.md) + - [پرومپٹ فلو کے ذریعے تشخیص](./md/01.Introduction/05/Promptflow.md) - ایزور AI سرچ کے ساتھ RAG - - [Phi-4-mini اور Phi-4-multimodal(RAG) کو ایزور AI سرچ کے ساتھ کیسے استعمال کریں](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [فی-4-منی اور فی-4-ملٹی موڈل (RAG) کو ایزور AI سرچ کے ساتھ استعمال کرنا](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [زیرو کلاؤڈ لوکل ہائبرڈ RAG SQLite FTS5 اور فی-4-منی کے ساتھ](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi ایپلیکیشن ڈویلپمنٹ کے نمونے +- فی ایپلیکیشن ڈیولپمنٹ نمونے - متن اور چیٹ ایپلیکیشنز - - Phi-4 کے نمونے - - [📓] [Phi-4-mini ONNX ماڈل کے ساتھ چیٹ](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Phi-4 لوکل ONNX ماڈل .NET کے ساتھ چیٹ](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Phi-4 ONNX کے ساتھ سمنٹک کرنل استعمال کرتے ہوئے .NET کنسول ایپ میں چیٹ](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5 کے نمونے - - [Phi3، ONNX Runtime Web اور WebGPU استعمال کرتے ہوئے براؤزر میں لوکل چیٹ بوٹ](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - فی-4 کے نمونے + - [📓] [فی-4-منی ONNX ماڈل کے ساتھ چیٹ](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [فی-4 لوکل ONNX ماڈل کے ساتھ چیٹ .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Sementic Kernel کا استعمال کرتے ہوئے فی-4 ONNX کے ساتھ چیٹ .NET کونسول ایپ](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - فائی-3 / 3.5 نمونے + - [براؤزر میں مقامی چیٹ بوٹ جو فی 3، ONNX Runtime Web اور WebGPU استعمال کرتا ہے](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [اوپن وینو چیٹ](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [ملٹی ماڈل - انٹرایکٹو فی-3-منی اور اوپن اے آئی وسپر](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [ایم ایل فلو - ایک ریپر بنانے اور ایم ایل فلو کے ساتھ فی-3 کا استعمال](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [ماڈل کی اصلاح - فی-3-منی ماڈل کو ONNX رن ٹائم ویب کے لیے اولیو کے ساتھ کیسے بہتر بنائیں](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 ایپ کے ساتھ Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[ WinUI3 ملٹی ماڈل AI پاورڈ نوٹس ایپ سیمپل](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [حسب ضرورت فی-3 ماڈلز کو پرامپٹ فلو کے ساتھ فائن ٹیون اور انٹیگریٹ کریں](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [مائیکروسافٹ فاؤنڈری میں پرامپٹ فلو کے ساتھ حسب ضرورت فی-3 ماڈلز کو فائن ٹیون اور انٹیگریٹ کریں](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [مائیکروسافٹ فاؤنڈری میں مائیکروسافٹ کے ذمہ دار AI اصولوں پر توجہ دیتے ہوئے فائن ٹیون شدہ فی-3 / فی-3.5 ماڈل کا جائزہ لیں](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct زبان کی پیش گوئی کا سیمپل (چینی/انگریزی)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [کثیر ماڈل - انٹرایکٹو Phi-3-mini اور OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - ریپر بنانے اور Phi-3 کو MLFlow کے ساتھ استعمال کرنے کا طریقہ](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [ماڈل کی اصلاح - ONNX Runtime Web کے لیے Phi-3-min ماڈل کو Olive کے ساتھ کیسے بہتر بنایا جائے](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 ایپ Phi-3 mini-4k-instruct-onnx کے ساتھ](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 ملٹی ماڈل AI پاورڈ نوٹس ایپ کا نمونہ](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [کسٹم فائی-3 ماڈلز کو فائن ٹون کریں اور پرامپٹ فلو کے ساتھ انضمام کریں](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Microsoft Foundry میں پرامپٹ فلو کے ساتھ کسٹم فائی-3 ماڈلز کو فائن ٹون کریں اور انضمام کریں](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Microsoft Foundry میں فائن ٹون کیے گئے Phi-3 / Phi-3.5 ماڈل کا جائزہ لیں، Microsoft کی ذمہ دار AI اصولوں پر توجہ کے ساتھ](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct زبان کی پیش گوئی کا نمونہ (چینی/انگریزی)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG چیٹ بوٹ](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [Windows GPU کا استعمال کرتے ہوئے پرامپٹ فلو حل تخلیق کرنا Phi-3.5-Instruct ONNX کے ساتھ](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [مائیکروسافٹ Phi-3.5 tflite کا استعمال کرتے ہوئے اینڈرائیڈ ایپ بنائیں](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET مثال مقامی ONNX Phi-3 ماڈل استعمال کرتے ہوئے Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [کنسول چیٹ .NET ایپ Semantic Kernel اور Phi-3 کے ساتھ](../../md/04.HOL/dotnet/src/LabsPhi302) - - - ایژور AI انفرنس SDK کوڈ کی بنیاد پر نمونے - - Phi-4 نمونے - - [📓] [Phi-4-multimodal کا استعمال کرتے ہوئے پروجیکٹ کوڈ تیار کریں](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Phi-3 / 3.5 نمونے - - [اپنا Visual Studio Code GitHub Copilot Chat مائیکروسافٹ Phi-3 فیملی کے ساتھ بنائیں](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [GitHub ماڈلز کے ساتھ Phi-3.5 کے ذریعے اپنا Visual Studio Code Chat Copilot Agent بنائیں](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - - ایڈوانسڈ ریزننگ نمونے - - Phi-4 نمونے + - [Phi-3.5-Instruct ONNX کے ساتھ پرامپٹ فلو حل بنانے کے لیے ونڈوز GPU کا استعمال](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [مائیکروسافٹ Phi-3.5 tflite استعمال کرتے ہوئے اینڈرائیڈ ایپ بنائیں](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [Q&A .NET مثال مائیکروسافٹ.ML.OnnxRuntime استعمال کرتے ہوئے مقامی ONNX Phi-3 ماڈل کے ساتھ](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Semantic Kernel اور Phi-3 کے ساتھ کنسول چیٹ .NET ایپ](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI Inference SDK کوڈ کی بنیاد پر نمونے + - فائی-4 نمونے + - [📓] [Phi-4-multimodal استعمال کرتے ہوئے پروجیکٹ کوڈ جنریٹ کریں](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - فائی-3 / 3.5 نمونے + - [مائیکروسافٹ Phi-3 فیملی کے ساتھ اپنا Visual Studio Code GitHub Copilot چیٹ بنائیں](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [GitHub ماڈلز کے ذریعے Phi-3.5 کے ساتھ اپنا Visual Studio Code Chat Copilot ایجنٹ بنائیں](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + + - جدید استدلال کے نمونے + - فائی-4 نمونے - [📓] [Phi-4-mini-reasoning یا Phi-4-reasoning نمونے](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [مائیکروسافٹ اولیو کے ساتھ Phi-4-mini-reasoning کی فائن ٹیوننگ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [ایپل MLX کے ساتھ Phi-4-mini-reasoning کی فائن ٹیوننگ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Microsoft Olive کے ساتھ Phi-4-mini-reasoning کی فائن ٹوننگ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [Apple MLX کے ساتھ Phi-4-mini-reasoning کی فائن ٹوننگ](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [GitHub ماڈلز کے ساتھ Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [مائیکروسافٹ فاؤنڈری ماڈلز کے ساتھ Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [📓] [Microsoft Foundry ماڈلز کے ساتھ Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - ڈیموز - - [Phi-4-mini ڈیموز جو Hugging Face Spaces پر میزبان ہیں](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal ڈیموز جو Hugginge Face Spaces پر میزبان ہیں](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-mini ڈیموز جو Hugging Face Spaces پر میزبانی کی گئی ہیں](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal ڈیموز جو Hugginge Face Spaces پر میزبانی کی گئی ہیں](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - وژن نمونے - - Phi-4 نمونے - - [📓] [تصاویر پڑھنے اور کوڈ تیار کرنے کے لیے Phi-4-multimodal کا استعمال کریں](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Phi-3 / 3.5 نمونے - - [📓][Phi-3-vision-تصویر متن سے متن تک](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - فائی-4 نمونے + - [📓] [Phi-4-multimodal استعمال کرتے ہوئے تصاویر پڑھیں اور کوڈ بنائیں](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - فائی-3 / 3.5 نمونے + - [📓][Phi-3-vision-تصویر سے متن تک متن](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP ایمبیڈنگ](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [ڈیمو: Phi-3 ری سائیکلنگ](https://github.com/jennifermarsman/PhiRecycling/) + - [ڈیمو: Phi-3 ریسائیکلنگ](https://github.com/jennifermarsman/PhiRecycling/) - [Phi-3-vision - بصری زبان کا معاون - Phi3-Vision اور OpenVINO کے ساتھ](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 وژن Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 وژن اوپن وینو](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 وژن ملٹی فریم یا ملٹی تصویر کا سیمپل](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 وژن لوکل ONNX ماڈل Microsoft.ML.OnnxRuntime .NET استعمال کرتے ہوئے](../../md/04.HOL/dotnet/src/LabsPhi303) - - [مینیو پر مبنی Phi-3 وژن لوکل ONNX ماڈل Microsoft.ML.OnnxRuntime .NET استعمال کرتے ہوئے](../../md/04.HOL/dotnet/src/LabsPhi304) - - - ریزننگ-وژن نمونے - - Phi-4-Reasoning-Vision-15B - - [📓] [Phi-4-Reasoning-Vision-15B کا استعمال کرتے ہوئے جے واکنگ کا پتہ لگانا](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [Phi-3 وژن OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 وژن کثیر فریم یا کثیر تصویر کا نمونہ](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [Phi-3 وژن مقامی ONNX ماڈل مائیکروسافٹ.ML.OnnxRuntime .NET استعمال کرتے ہوئے](../../md/04.HOL/dotnet/src/LabsPhi303) + - [مینیو پر مبنی Phi-3 وژن مقامی ONNX ماڈل مائیکروسافٹ.ML.OnnxRuntime .NET استعمال کرتے ہوئے](../../md/04.HOL/dotnet/src/LabsPhi304) + + - استدلال-وژن نمونے + - Phi-4-Reasoning-Vision-15B + - [📓] [Phi-4-Reasoning-Vision-15B کا استعمال کرتے ہوئے جے واکنگ کی شناخت](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - [📓] [Phi-4-Reasoning-Vision-15B کا استعمال کرتے ہوئے ریاضی](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Phi-4-Reasoning-Vision-15B کا استعمال کرتے ہوئے UI کا پتہ لگانا](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [Phi-4-Reasoning-Vision-15B کا استعمال کرتے ہوئے UI کی شناخت](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - ریاضی کے نمونے - - Phi-4-Mini-Flash-Reasoning-Instruct نمونے [ریاضی ڈیمو Phi-4-Mini-Flash-Reasoning-Instruct کے ساتھ](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-Mini-Flash-Reasoning-Instruct نمونے [Phi-4-Mini-Flash-Reasoning-Instruct کے ساتھ ریاضی کا ڈیمو](./md/02.Application/09.Math/MathDemo.ipynb) - آڈیو نمونے - - Phi-4 نمونے - - [📓] [Phi-4-multimodal کا استعمال کرتے ہوئے آڈیو ٹرانسکرپٹس نکالنا](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal آڈیو سیمپل](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal تقریر ترجمہ سیمپل](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET کنسول ایپلیکیشن Phi-4-multimodal آڈیو کا استعمال کرتے ہوئے آڈیو فائل کا تجزیہ اور ٹرانسکرپٹ تیار کرنا](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - ماہرین کے مرکب کے نمونے (MOE) - - Phi-3 / 3.5 نمونے - - [📓] [Phi-3.5 ماہرین کے مرکب (MoEs) سوشل میڈیا سیمپل](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [NVIDIA NIM Phi-3 MOE، ایژور AI سرچ، اور لاما انڈیکس کے ساتھ ریٹریول آگمینٹڈ جنریشن (RAG) پائپ لائن بنانا](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - فائی-4 نمونے + - [📓] [Phi-4-multimodal کا استعمال کرتے ہوئے آڈیو نقلیں نکالنا](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal آڈیو نمونہ](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal تقریر ترجمہ کا نمونہ](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET کنسول ایپلی کیشن کا استعمال کرتے ہوئے Phi-4-multimodal آڈیو فائل کا تجزیہ اور نقل تیار کرنا](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - MOE نمونے + - فائی-3 / 3.5 نمونے + - [📓] [Phi-3.5 ماہرین کے امتزاج (MoEs) سے متعلق سوشل میڈیا نمونہ](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [NVIDIA NIM Phi-3 MOE، Azure AI Search، اور LlamaIndex کے ساتھ ریٹریول-آگمینٹڈ جنریشن (RAG) پائپ لائن بنانا](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - فنکشن کالنگ نمونے - - Phi-4 نمونے 🆕 + - فائی-4 نمونے 🆕 - [📓] [Phi-4-mini کے ساتھ فنکشن کالنگ کا استعمال](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [فنکشن کالنگ کا استعمال کرتے ہوئے Phi-4-mini کے ساتھ ملٹی ایجنٹس بنانا](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [اولاما کے ساتھ فنکشن کالنگ کا استعمال](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [Phi-4-mini کے ساتھ ملٹی ایجنٹس بنانے کے لیے فنکشن کالنگ کا استعمال](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [Ollama کے ساتھ فنکشن کالنگ کا استعمال](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [ONNX کے ساتھ فنکشن کالنگ کا استعمال](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - کثیرالماڈل مکسنگ نمونے - - Phi-4 نمونے 🆕 - - [📓] [ٹیکنالوجی جرنلسٹ کے طور پر Phi-4-multimodal کا استعمال](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET کنسول ایپلیکیشن جس میں Phi-4-multimodal کا استعمال تصاویر کا تجزیہ کرنے کے لیے](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - -- فی نمونوں کی فائن ٹیوننگ - - [فائن ٹیوننگ کے مناظر](./md/03.FineTuning/FineTuning_Scenarios.md) - - [فائن ٹیوننگ بمقابلہ RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [فی-3 کو صنعت کا ماہر بننے دیں](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [VS کوڈ کے لیے AI ٹول کٹ کے ساتھ فی-3 کی فائن ٹیوننگ](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [ایژور مشین لرننگ سروس کے ساتھ فی-3 کی فائن ٹیوننگ](./md/03.FineTuning/Introduce_AzureML.md) - - [لوڑا کے ساتھ فی-3 کی فائن ٹیوننگ](./md/03.FineTuning/FineTuning_Lora.md) - - [QLora کے ساتھ فی-3 کی فائن ٹیوننگ](./md/03.FineTuning/FineTuning_Qlora.md) - - [مائیکروسافٹ فاؤنڈری کے ساتھ فی-3 کی فائن ٹیوننگ](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [ایژور ML CLI/SDK کے ساتھ فی-3 کی فائن ٹیوننگ](./md/03.FineTuning/FineTuning_MLSDK.md) - - [مائیکروسافٹ اولیو کے ساتھ فائن ٹیوننگ](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [مائیکروسافٹ اولیو ہینڈز آن لیب کے ساتھ فائن ٹیوننگ](./md/03.FineTuning/olive-lab/readme.md) - - [ویز اور بایس کے ساتھ فی-3-ویژن کی فائن ٹیوننگ](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [ایپل MLX فریم ورک کے ساتھ فی-3 کی فائن ٹیوننگ](./md/03.FineTuning/FineTuning_MLX.md) - - [فی-3-ویژن کی فائن ٹیوننگ (سرکاری حمایت)](./md/03.FineTuning/FineTuning_Vision.md) - - [کائتو AKS کے ساتھ Phi-3 کی فائن ٹوننگ، Azure کنٹینرز (سرکاری سپورٹ)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Phi-3 اور 3.5 وژن کی فائن ٹوننگ](https://github.com/2U1/Phi3-Vision-Finetune) - -- ہینڈز آن لیب - - [جدید ماڈلز کی دریافت: LLMs، SLMs، مقامی ترقی اور مزید](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [NLP کی صلاحیتوں کا انلاک کرنا: Microsoft Olive کے ساتھ فائن ٹوننگ](https://github.com/azure/Ignite_FineTuning_workshop) - -- علمی تحقیقی مقالات اور اشاعتیں - - [کتابیں سب کچھ ہیں II: phi-1.5 تکنیکی رپورٹ](https://arxiv.org/abs/2309.05463) - - [Phi-3 تکنیکی رپورٹ: ایک اعلیٰ صلاحیت والا زبان ماڈل آپ کے فون پر مقامی طور پر](https://arxiv.org/abs/2404.14219) + - ملٹی موڈل مکسنگ نمونے + - فائی-4 نمونے 🆕 + - [📓] [ٹیکنالوجی صحافی کے طور پر Phi-4-multimodal کا استعمال](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET کنسول ایپلی کیشن کا استعمال کرتے ہوئے Phi-4-multimodal تصاویر کا تجزیہ](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + +- فائن ٹوننگ فی نمونے + - [فائن ٹوننگ کے منظرنامے](./md/03.FineTuning/FineTuning_Scenarios.md) + - [فائن ٹوننگ بمقابلہ RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Phi-3 کو صنعتی ماہر بنائیں](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [VS Code کے لیے AI Toolkit کے ساتھ Phi-3 کی فائن ٹوننگ](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Azure Machine Learning سروس کے ساتھ Phi-3 کی فائن ٹوننگ](./md/03.FineTuning/Introduce_AzureML.md) + - [Lora کے ساتھ Phi-3 کی فائن ٹوننگ](./md/03.FineTuning/FineTuning_Lora.md) + - [QLora کے ساتھ Phi-3 کی فائن ٹوننگ](./md/03.FineTuning/FineTuning_Qlora.md) + - [Microsoft Foundry کے ساتھ Phi-3 کی فائن ٹوننگ](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [Azure ML CLI/SDK کے ساتھ Phi-3 کی فائن ٹوننگ](./md/03.FineTuning/FineTuning_MLSDK.md) + - [Microsoft Olive کے ساتھ فائن ٹوننگ](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Microsoft Olive Hands-On Lab سے فائن ٹوننگ](./md/03.FineTuning/olive-lab/readme.md) + - [Weights اور Bias کے ساتھ Phi-3-vision کی فائن ٹوننگ](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + + - [ایپل MLX فریم ورک کے ساتھ فی-3 کو فائن ٹون کرنا](./md/03.FineTuning/FineTuning_MLX.md) + - [Phi-3-vision کو فائن ٹون کرنا (سرکاری حمایت)](./md/03.FineTuning/FineTuning_Vision.md) + - [کائٹو AKS، ایزور کنٹینرز کے ساتھ Phi-3 کو فائن ٹون کرنا (سرکاری حمایت)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Phi-3 اور 3.5 وژن کو فائن ٹون کرنا](https://github.com/2U1/Phi3-Vision-Finetune) + +- عملی لیب + - [جدید ترین ماڈلز کی تلاش: LLMs، SLMs، مقامی ترقی اور مزید](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [NLP کی صلاحیت کھولنا: مائیکروسافٹ اولیو کے ساتھ فائن ٹوننگ](https://github.com/azure/Ignite_FineTuning_workshop) + +- تعلیمی تحقیقی مقالے اور اشاعتیں + - [صرف درسی کتابیں کافی ہیں II: phi-1.5 تکنیکی رپورٹ](https://arxiv.org/abs/2309.05463) + - [Phi-3 تکنیکی رپورٹ: آپ کے فون پر ایک انتہائی قابل زبان ماڈل](https://arxiv.org/abs/2404.14219) - [Phi-4 تکنیکی رپورٹ](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini تکنیکی رپورٹ: کمپیکٹ لیکن طاقتور ملٹی موڈل زبان ماڈلز LoRAs کے مِکسچر کے ذریعے](https://arxiv.org/abs/2503.01743) - - [گاڑی میں فنکشن کالنگ کے لیے چھوٹے زبان ماڈلز کو بہتر بنانا](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) PHI-3 کی فائن ٹوننگ برائے ملٹی پل چوائس سوالات کے جواب: طریقہ کار، نتائج، اور چیلنجز](https://arxiv.org/abs/2501.01588) + - [Phi-4-Mini تکنیکی رپورٹ: مختصر مگر طاقتور ملٹی ماڈل زبان کے ماڈلز LoRA کے امتزاج کے ذریعے](https://arxiv.org/abs/2503.01743) + - [گاڑی میں فنکشن کالنگ کے لیے چھوٹے زبان کے ماڈلز کی اصلاح](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) فائن ٹوننگ PHI-3 برائے کثیر انتخاب سوالات کا جواب دینا: طریقہ کار، نتائج، اور چیلنجز](https://arxiv.org/abs/2501.01588) - [Phi-4-ریزننگ تکنیکی رپورٹ](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Phi-4-منی-ریزننگ تکنیکی رپورٹ](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Phi ماڈلز کا استعمال +## فی ماڈلز کا استعمال -### Microsoft Foundry پر Phi +### مائیکروسافٹ فاؤنڈری پر Phi -آپ سیکھ سکتے ہیں کہ Microsoft Phi کو کیسے استعمال کیا جائے اور اپنے مختلف ہارڈویئر ڈیوائسز میں E2E حل کیسے بنائیں۔ خود Phi کا تجربہ کرنے کے لیے، ماڈلز کے ساتھ کھیلنا شروع کریں اور [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) استعمال کرتے ہوئے اپنے منظرناموں کے لیے Phi کو ذاتی بنائیں۔ آپ مزید جان سکتے ہیں Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) سے۔ +آپ مائیکروسافٹ فی کا استعمال کرنا اور اپنے مختلف ہارڈویئر آلات میں ای2ای حل تیار کرنا سیکھ سکتے ہیں۔ خود فی کا تجربہ کرنے کے لیے، ماڈلز کے ساتھ کھیلنا شروع کریں اور اپنے منظرناموں کے لیے فی کو حسب ضرورت بنائیں [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) کا استعمال کرتے ہوئے آپ مزید جان سکتے ہیں Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**پلے گراؤنڈ** -ہر ماڈل کا ایک مخصوص پلے گراؤنڈ ہوتا ہے جہاں آپ ماڈل کو آزما سکتے ہیں [Azure AI Playground](https://aka.ms/try-phi3)۔ +**پلے گراؤنڈ** +ہر ماڈل کے لیے ایک مخصوص پلے گراؤنڈ موجود ہے تاکہ ماڈل کو ٹیسٹ کیا جا سکے [Azure AI Playground](https://aka.ms/try-phi3). ### GitHub ماڈلز پر Phi -آپ سیکھ سکتے ہیں کہ Microsoft Phi کو کیسے استعمال کیا جائے اور اپنے مختلف ہارڈویئر ڈیوائسز میں E2E حل کیسے بنائیں۔ خود Phi کا تجربہ کرنے کے لیے، ماڈل کے ساتھ کھیلنا شروع کریں اور اپنے منظرناموں کے لیے Phi کو ذاتی بنانے کے لیے [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) استعمال کریں۔ آپ مزید جان سکتے ہیں Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) سے۔ +آپ مائیکروسافٹ فی کا استعمال کرنا اور اپنے مختلف ہارڈویئر آلات میں ای2ای حل تیار کرنا سیکھ سکتے ہیں۔ خود فی کا تجربہ کرنے کے لیے، ماڈل کے ساتھ کھیلنا شروع کریں اور اپنے منظرناموں کے لیے فی کو حسب ضرورت بنائیں [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) کا استعمال کرتے ہوئے آپ مزید جان سکتے ہیں Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) -**پلے گراؤنڈ** -ہر ماڈل کا مخصوص [پلے گراؤنڈ جہاں ماڈل کو آزمایا جا سکتا ہے](/md/02.QuickStart/GitHubModel_QuickStart.md) موجود ہے۔ +**پلے گراؤنڈ** +ہر ماڈل کے لیے ایک مخصوص [پلے گراؤنڈ تاکہ ماڈل کو ٹیسٹ کیا جا سکے](/md/02.QuickStart/GitHubModel_QuickStart.md) موجود ہے۔ ### Hugging Face پر Phi آپ ماڈل کو [Hugging Face](https://huggingface.co/microsoft) پر بھی تلاش کر سکتے ہیں۔ -**پلے گراؤنڈ** -[Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**پلے گراؤنڈ** + [Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 دیگر کورسز + ## 🎒 دیگر کورسز ہماری ٹیم دیگر کورسز بھی تیار کرتی ہے! دیکھیں: -### LangChain +### لینگ چین [![نو آموزوں کے لیے LangChain4j](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) [![نو آموزوں کے لیے LangChain.js](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) [![نو آموزوں کے لیے LangChain](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / ایجینٹس +### ایزور / ایج / MCP / ایجنٹس [![نو آموزوں کے لیے AZD](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) [![نو آموزوں کے لیے Edge AI](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) [![نو آموزوں کے لیے MCP](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![نو آموزوں کے لیے AI ایجینٹس](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![نو آموزوں کے لیے AI ایجنٹس](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### جنریٹیو AI سیریز [![نو آموزوں کے لیے جنریٹیو AI](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) [![جنریٹیو AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) @@ -271,53 +274,56 @@ [![جنریٹیو AI (جاوا اسکرپٹ)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### بنیادی سیکھنا + +### کور لرننگ [![نو آموزوں کے لیے ML](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) [![نو آموزوں کے لیے ڈیٹا سائنس](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![نو آموزوں کے لیے AI](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![نو آموزوں کے لیے سائبر سیکیورٹی](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![نو آموزوں کے لیے ویب ڈیولپمنٹ](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![نو آموزوں کے لیے ویب ڈویلپمنٹ](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![نو آموزوں کے لیے IoT](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![نو آموزوں کے لیے XR ڈیولپمنٹ](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![نو آموزوں کے لیے XR ڈیویلپمنٹ](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### کوپایلٹ سیریز -[![AI کے لیے جوڑے ہوئے پروگرامنگ کے کوپایلٹ](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![C#/.NET کے لیے کوپایلٹ](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![کوپایلٹ ایڈونچر](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + +### کوپائلٹ سیریز +[![AI جوڑ توڑ پروگرامنگ کے لیے کوپائلٹ](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![C#/.NET کے لیے کوپائلٹ](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![کوپائلٹ ایڈونچر](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## ذمہ دار AI -Microsoft اپنے صارفین کی مدد کرنے کے لیے پرعزم ہے کہ وہ ہمارے AI مصنوعات کو ذمہ داری کے ساتھ استعمال کریں، اپنی سیکھ کو شیئر کریں، اور ٹرانسپیرنسی نوٹس اور امپیکٹ اسیسمنٹس جیسے ٹولز کے ذریعے بھروسے پر مبنی شراکت داریاں قائم کریں۔ ان میں سے بہت سے وسائل [https://aka.ms/RAI](https://aka.ms/RAI) پر دستیاب ہیں۔ -Microsoft کا ذمہ دار AI کا نظریہ ہمارے AI کے اصولوں پر مبنی ہے: انصاف، قابل اعتماد ہونا اور حفاظت، رازداری اور سیکورٹی، شمولیت، شفافیت، اور جوابدہی۔ +مائیکروسافٹ ہمارے صارفین کی مدد کرنے کے لیے پرعزم ہے کہ وہ ہمارے AI مصنوعات کو ذمے داری کے ساتھ استعمال کریں، ہماری سیکھوں کو شیئر کریں، اور ٹرانسپیرنسی نوٹس اور امپیکٹ اسسمنٹس جیسے اوزار کے ذریعے اعتماد پر مبنی شراکت داری قائم کریں۔ ان میں سے بہت سے وسائل [https://aka.ms/RAI](https://aka.ms/RAI) پر دستیاب ہیں۔ +مائیکروسافٹ کا ذمہ دار AI کا طریقہ کار ہمارے AI اصولوں پر مبنی ہے جس میں منصفانہ، قابل اعتماد اور محفوظ، پرائیویسی اور سیکیورٹی، شمولیت، شفافیت، اور جوابدہی شامل ہیں۔ + +بڑے پیمانے پر قدرتی زبان، تصویر، اور تقریر کے ماڈلز - جیسا کہ اس نمونے میں استعمال کیے گئے ہیں - ممکنہ طور پر ایسے طریقوں سے برتاو کر سکتے ہیں جو غیر منصفانہ، غیر قابل اعتماد، یا ناگوار ہوں، جو نقصان کا باعث بن سکتے ہیں۔ براہ کرم رجوع کریں [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) تا کہ خطرات اور حدود کے بارے میں معلومات حاصل کر سکیں۔ + -بڑے پیمانے کے قدرتی زبان، تصویر، اور تقریر کے ماڈلز — جیسے کہ اس نمونے میں استعمال ہونے والے — ممکن ہے کہ کچھ معاملات میں ناجائز، ناقابل اعتماد، یا توہین آمیز رویہ اپنائیں، جو نقصان کا سبب بن سکتی ہے۔ براہ کرم خطرات اور حد بندیوں سے آگاہ ہونے کے لیے [Azure OpenAI service Transparency note](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) کو دیکھیں۔ -ان خطرات کو کم کرنے کے لیے تجویز کردہ طریقہ یہ ہے کہ آپ اپنی آرکیٹیکچر میں ایک حفاظتی نظام شامل کریں جو نقصان دہ رویے کا پتہ لگا سکے اور اسے روک سکے۔ [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ایک آزاد حفاظتی پرت فراہم کرتا ہے، جو ایپلیکیشنز اور خدمات میں نقصان دہ صارف-بنائی گئی اور AI-بنائی گئی مواد کا پتہ لگا سکتا ہے۔ Azure AI Content Safety میں ٹیکسٹ اور امیج APIs شامل ہیں جو آپ کو نقصان دہ مواد کا پتہ لگانے کی اجازت دیتے ہیں۔ Microsoft Foundry کے اندر، Content Safety سروس آپ کو مختلف طریقوں سے نقصان دہ مواد کا پتہ لگانے کے لیے نمونہ کوڈ دیکھنے، دریافت کرنے اور آزمانے کی سہولت فراہم کرتی ہے۔ درج ذیل [quickstart دستاویزات](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) آپ کو سروس کو درخواستیں دینے کے عمل سے رہنمائی کرتی ہیں۔ +ان خطرات کو کم کرنے کے لیے تجویز کردہ طریقہ کار یہ ہے کہ آپ اپنے آرکیٹیکچر میں ایک حفاظتی نظام شامل کریں جو نقصان دہ رویے کا پتہ لگا سکے اور اسے روک سکے۔ [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) ایک آزاد حفاظتی پرت فراہم کرتا ہے، جو ایپلیکیشنز اور خدمات میں نقصان دہ صارف کی طرف سے تیار کردہ اور AI کی طرف سے تیار کردہ مواد کا پتہ لگانے کے قابل ہے۔ Azure AI Content Safety میں متن اور تصویر کے APIs شامل ہیں جو آپ کو نقصان دہ مواد کا پتہ لگانے کی اجازت دیتے ہیں۔ Microsoft Foundry کے اندر، Content Safety سروس آپ کو مختلف طریقوں سے نقصان دہ مواد کا پتہ لگانے کے لیے نمونہ کوڈ دیکھنے، دریافت کرنے اور آزمانے کی اجازت دیتی ہے۔ درج ذیل [quickstart ڈاکیومنٹیشن](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) آپ کو سروس کو درخواستیں کرنے کے طریقہ کار سے رہنمائی فراہم کرتی ہے۔ -ایک اور پہلو جس پر غور کرنا ضروری ہے وہ مجموعی ایپلیکیشن کی کارکردگی ہے۔ کثیر المناحل اور کثیر ماڈلز ایپلیکیشنز کے ساتھ، ہم کارکردگی سے مراد لیتے ہیں کہ نظام آپ اور آپ کے صارفین کی توقعات کے مطابق کام کرے، جس میں نقصان دہ نتائج پیدا نہ کرنا بھی شامل ہے۔ آپ کے لیے ضروری ہے کہ آپ اپنی مجموعی ایپلیکیشن کی کارکردگی کا جائزہ لیں [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) کا استعمال کرتے ہوئے۔ آپ کے پاس [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) بنانے اور جانچنے کی بھی صلاحیت موجود ہے۔ +ایک اور پہلو جس کا خیال رکھنا ضروری ہے وہ مجموعی ایپلیکیشن کی کارکردگی ہے۔ کثیر وضع اور کثیر ماڈل ایپلیکیشنز کے ساتھ، ہم کارکردگی کو اس طرح سمجھتے ہیں کہ سسٹم ویسا ہی کام کرے جیسا کہ آپ اور آپ کے صارفین توقع کرتے ہیں، جس میں نقصان دہ نتائج پیدا نہ ہوں۔ اپنی مجموعی ایپلیکیشن کی کارکردگی کا اندازہ لگانا ضروری ہے جس کے لیے [Performance and Quality and Risk and Safety evaluators](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) کا استعمال کریں۔ آپ کے پاس [custom evaluators](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) تخلیق کرنے اور ان سے جانچ کرنے کی بھی صلاحیت موجود ہے۔ -آپ اپنے AI ایپلیکیشن کو اپنے ترقیاتی ماحول میں [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) استعمال کرکے جانچ سکتے ہیں۔ چاہے آپ کے پاس ٹیسٹ ڈیٹا سیٹ ہو یا کوئی ہدف، آپ کی جینیریٹو AI ایپلیکیشن کی تخلیقات کو بلٹ ان یا آپ کے منتخب کردہ کسٹم ایویلیویٹرز کے ذریعے مقداری طور پر ماپا جاتا ہے۔ اپنے نظام کا جائزہ لینے کے لیے azure ai evaluation sdk کے ساتھ شروع کرنے کے لیے، آپ [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) کی پیروی کر سکتے ہیں۔ ایک بار جب آپ جائزہ چلائیں گے، تو آپ [Microsoft Foundry میں نتائج کو بصری صورت میں دیکھ سکتے ہیں](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)۔ +آپ اپنے AI ایپلیکیشن کا اپنے ترقیاتی ماحول میں [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html) کے ذریعے جائزہ لے سکتے ہیں۔ چاہے ایک ٹیسٹ ڈیٹاسیٹ ہو یا ہدف، آپ کی جنریٹیو AI ایپلیکیشن کی جنریشنز کو بلٹ ان ایوالویٹرز یا اپنی پسند کے کسٹم ایوالویٹرز کے ذریعے مقداری طور پر ماپا جاتا ہے۔ اپنے سسٹم کا جائزہ لینے کے لیے Azure AI Evaluation SDK کے ساتھ شروع ہونے کے لیے، آپ [quickstart guide](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk) کی پیروی کر سکتے ہیں۔ ایک بار جب آپ جائزے کا عمل انجام دیتے ہیں، تو آپ [Microsoft Foundry میں نتائج کا تجزیہ کر سکتے ہیں](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)۔ -## ٹریڈ مارکس +## تجارتی نشان -یہ پروجیکٹ پروجیکٹس، مصنوعات، یا خدمات کے لیے ٹریڈ مارکس یا لوگوز پر مشتمل ہو سکتا ہے۔ Microsoft کے ٹریڈ مارکس یا لوگوز کے مجاز استعمال پر [Microsoft کے ٹریڈ مارک اور برانڈ گائیڈ لائنز](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) لاگو ہوتے ہیں اور انہیں لازمی طور پر ان قواعد کی پیروی کرنا ہوتی ہے۔ اس پروجیکٹ کے ترمیم شدہ ورژنز میں Microsoft کے ٹریڈ مارکس یا لوگوز کا استعمال الجھن پیدا نہیں کرے گا اور نہ ہی Microsoft کی اسپانسرشپ کا اشارہ دے گا۔ تیسرے فریق کے ٹریڈ مارکس یا لوگوز کا کوئی بھی استعمال ان تیسرے فریق کی پالیسیوں کے تابع ہوگا۔ +یہ پروجیکٹ پروجیکٹس، مصنوعات، یا خدمات کے لیے تجارتی نشان یا لوگوز رکھ سکتا ہے۔ Microsoft کے تجارتی نشانوں یا لوگوز کے مجاز استعمال کے لیے [Microsoft کے تجارتی نشان اور برانڈ کی ہدایات](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general) کی پیروی کرنا ضروری ہے۔ +Microsoft کے تجارتی نشانوں یا لوگوز کے ترمیم شدہ ورژنز میں استعمال سے الجھن پیدا نہیں ہونی چاہیے اور نہ ہی وہ Microsoft کی اسپانسرشپ ظاہر کریں۔ کسی بھی تیسرے فریق کے تجارتی نشانوں یا لوگوز کے استعمال پر ان تیسرے فریق کی پالیسیاں لاگو ہوتی ہیں۔ -## مدد حاصل کرنا +## مدد حاصل کریں -اگر آپ پھنس جائیں یا AI ایپس بنانے کے بارے میں کوئی سوال ہو، تو شامل ہوں: +اگر آپ کو پھنسنے کا سامنا ہو یا AI ایپس بنانے کے بارے میں کوئی سوال ہو تو شامل ہوں: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -اگر آپ کو مصنوعات کے بارے میں تاثرات یا تعمیر کے دوران خرابیوں کا سامنا ہو، تو وزٹ کریں: +اگر آپ کے پاس مصنوع کے بارے میں آراء ہوں یا تعمیر کے دوران کوئی خرابی ہو تو دیکھیں: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**ڈسکلیمر**: -یہ دستاویز AI ترجمہ سروس [Co-op Translator](https://github.com/Azure/co-op-translator) کے ذریعے ترجمہ کی گئی ہے۔ اگرچہ ہم درستگی کی کوشش کرتے ہیں، براہ کرم آگاہ رہیں کہ خودکار ترجمے میں غلطیاں یا غیر یقینی باتیں ہو سکتی ہیں۔ اصل دستاویز اپنی مادری زبان میں معتبر ذریعہ سمجھی جانی چاہیے۔ اہم معلومات کے لیے پیشہ ور انسانی ترجمہ کی سفارش کی جاتی ہے۔ اس ترجمے کے استعمال سے پیدا ہونے والی کسی بھی غلط فہمی یا غلط تشریح کی ذمہ داری ہم پر عائد نہیں ہوتی۔ +**ڈس کلیمر**: +یہ دستاویز AI ترجمہ سروس [Co-op Translator](https://github.com/Azure/co-op-translator) کے ذریعے ترجمہ کی گئی ہے۔ جبکہ ہم درستگی کے لیے کوشاں ہیں، براہ کرم اس بات سے آگاہ رہیں کہ خودکار ترجمے میں غلطیاں یا عدم درستیاں ہو سکتی ہیں۔ اصل دستاویز اپنے مادری زبان میں مستند ماخذ سمجھی جائے گی۔ حساس معلومات کے لیے پیشہ ور انسانی ترجمہ کی سفارش کی جاتی ہے۔ اس ترجمے کے استعمال سے پیدا ہونے والی کسی بھی غلط فہمی یا غلط تشریح کی ذمہ داری ہم قبول نہیں کرتے۔ \ No newline at end of file diff --git a/translations/ur/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/ur/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..0319474ab --- /dev/null +++ b/translations/ur/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 مائیکروسافٹ phi-4-mini اور SQLite FTS5 کے ساتھ ہائبرڈ RAG (زیرو-کلاؤڈ SLM)\n", + "\n", + "> **مصنف:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **توجہ:** چھوٹے زبان ماڈلز (SLMs)، SQLite FTS5 BM25، گھنے ایمبیڈنگز، ریسپروکل رینک فیوژن (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. محرک: مقامی SLMs میں کی ورڈ ریکال کا مسئلہ\n", + "معیاری RAG ساختیں جو صرف گھنے ویکٹر ایمبیڈنگز پر انحصار کرتی ہیں، اکثر بالکل درست عددی ٹوکنز (مثلاً `2,340,000 TL`، معاہدہ کوڈز، اکاؤنٹ نمبر) بازیافت کرنے میں ناکام رہتی ہیں۔ \n", + "اس کے برعکس، سپارس لِیکسی کل سرچ (BM25) معنوی مترادفات اور پیرافرایز کیے گئے سوالات کو نظر انداز کر دیتا ہے۔\n", + "\n", + "یہ کوک بک دکھاتی ہے کہ **تیز رفتار، اندرونی میموری ہائبرڈ ریٹریول انجن** کیسے نافذ کیا جائے جس میں شامل ہیں:\n", + "1. **گھنے ویکٹرز** (کوسائن مماثلت)\n", + "2. **سپارس لِیکسی کل سرچ** (SQLite FTS5 BM25)\n", + "3. **ریس پروکل رینک فیوژن (RRF, $k=60$)**\n", + "4. **مایکروسافٹ `phi-4-mini` کے ساتھ زمینی حوالہ جات کی پیداوار (`[1]`, `[2]`)۔**\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. ڈوئل SQLite اسکیمہ (گنجان ویکٹرز + ورچوئل FTS5 BM25 ٹیبل)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. نمونہ جذب اور عملدرآمد بینچ مارک\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. مائیکروسافٹ phi-4-mini کے لیے Grounded پرامپٹ کی تشکیل\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**ڈس کلیمر**:\nیہ دستاویز AI ترجمہ سروس [Co-op Translator](https://github.com/Azure/co-op-translator) کے ذریعے ترجمہ کی گئی ہے۔ جبکہ ہم درستگی کے لیے کوشاں ہیں، براہ کرم اس بات سے آگاہ رہیں کہ خودکار ترجمے میں غلطیاں یا عدم درستیاں ہو سکتی ہیں۔ اصل دستاویز اپنے مادری زبان میں مستند ماخذ سمجھی جائے گی۔ حساس معلومات کے لیے پیشہ ور انسانی ترجمہ کی سفارش کی جاتی ہے۔ اس ترجمے کے استعمال سے پیدا ہونے والی کسی بھی غلط فہمی یا غلط تشریح کی ذمہ داری ہم قبول نہیں کرتے۔\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/vi/.co-op-translator.json b/translations/vi/.co-op-translator.json index cfef19150..757b70d73 100644 --- a/translations/vi/.co-op-translator.json +++ b/translations/vi/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "vi" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T01:08:35+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:07:52+00:00", "source_file": "README.md", "language_code": "vi" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "vi" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:08:43+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "vi" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:31:32+00:00", diff --git a/translations/vi/README.md b/translations/vi/README.md index 4c0ac3354..5c034a77e 100644 --- a/translations/vi/README.md +++ b/translations/vi/README.md @@ -1,42 +1,42 @@ -# Phi Cookbook: Ví dụ Thực hành với các Mô hình Phi của Microsoft +# Sổ Tay Phi: Ví Dụ Thực Hành với Các Mô Hình Phi của Microsoft [![Mở và sử dụng các mẫu trong GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) [![Mở trong Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![Người đóng góp trên GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![Vấn đề trên GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![Yêu cầu kéo trên GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![Chào đón PR](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![Người đóng góp GitHub](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![Vấn đề GitHub](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![Yêu cầu kéo GitHub](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![Chào mừng Pull Requests](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![Người theo dõi trên GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![Fork trên GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![Sao trên GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![Người theo dõi GitHub](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![Sao chép GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![Lượt sao GitHub](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi là một chuỗi các mô hình AI mã nguồn mở được phát triển bởi Microsoft. +Phi là một loạt các mô hình AI mã nguồn mở được phát triển bởi Microsoft. -Phi hiện là mô hình ngôn ngữ nhỏ (SLM) mạnh mẽ và tiết kiệm chi phí nhất, với các đánh giá rất tốt trong đa ngôn ngữ, suy luận, tạo văn bản/trò chuyện, lập trình, hình ảnh, âm thanh và các kịch bản khác. +Phi hiện là mô hình ngôn ngữ nhỏ (SLM) mạnh mẽ và hiệu quả chi phí nhất, với các chỉ số đánh giá rất tốt trong đa ngôn ngữ, suy luận, tạo văn bản/trò chuyện, mã hóa, hình ảnh, âm thanh và các kịch bản khác. -Bạn có thể triển khai Phi lên đám mây hoặc thiết bị biên, và bạn có thể dễ dàng xây dựng các ứng dụng AI sinh tạo với năng lực tính toán giới hạn. +Bạn có thể triển khai Phi lên đám mây hoặc thiết bị biên, và bạn có thể dễ dàng xây dựng các ứng dụng AI tạo sinh với công suất tính toán hạn chế. -Thực hiện các bước sau để bắt đầu sử dụng tài nguyên này: -1. **Fork Repository**: Nhấp [![Fork trên GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Clone Repository**: `git clone https://github.com/microsoft/PhiCookBook.git` +Thực hiện theo các bước sau để bắt đầu sử dụng các tài nguyên này: +1. **Tạo bản sao kho lưu trữ**: Nhấn [![Sao chép GitHub](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Tải kho lưu trữ về máy**: `git clone https://github.com/microsoft/PhiCookBook.git` 3. [**Tham gia cộng đồng Microsoft AI Discord và gặp gỡ các chuyên gia cùng nhà phát triển**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/vi/cover.eb18d1b9605d754b.webp) +![bìa](../../translated_images/vi/cover.eb18d1b9605d754b.webp) -### 🌐 Hỗ trợ Đa Ngôn Ngữ +### 🌐 Hỗ trợ đa ngôn ngữ -#### Hỗ trợ qua GitHub Action (Tự động & luôn Cập nhật) +#### Hỗ trợ qua GitHub Action (Tự động & Luôn được cập nhật) -[Tiếng Ả Rập](../ar/README.md) | [Tiếng Bengal](../bn/README.md) | [Tiếng Bungari](../bg/README.md) | [Tiếng Miến Điện (Myanmar)](../my/README.md) | [Tiếng Trung (Giản thể)](../zh-CN/README.md) | [Tiếng Trung (Phồn thể, Hồng Kông)](../zh-HK/README.md) | [Tiếng Trung (Phồn thể, Macau)](../zh-MO/README.md) | [Tiếng Trung (Phồn thể, Đài Loan)](../zh-TW/README.md) | [Tiếng Croatia](../hr/README.md) | [Tiếng Séc](../cs/README.md) | [Tiếng Đan Mạch](../da/README.md) | [Tiếng Hà Lan](../nl/README.md) | [Tiếng Estonia](../et/README.md) | [Tiếng Phần Lan](../fi/README.md) | [Tiếng Pháp](../fr/README.md) | [Tiếng Đức](../de/README.md) | [Tiếng Hy Lạp](../el/README.md) | [Tiếng Do Thái](../he/README.md) | [Tiếng Hindi](../hi/README.md) | [Tiếng Hungary](../hu/README.md) | [Tiếng Indonesia](../id/README.md) | [Tiếng Ý](../it/README.md) | [Tiếng Nhật](../ja/README.md) | [Tiếng Kannada](../kn/README.md) | [Tiếng Khmer](../km/README.md) | [Tiếng Hàn](../ko/README.md) | [Tiếng Litva](../lt/README.md) | [Tiếng Mã Lai](../ms/README.md) | [Tiếng Malayalam](../ml/README.md) | [Tiếng Marathi](../mr/README.md) | [Tiếng Nepal](../ne/README.md) | [Tiếng Pidgin Nigeria](../pcm/README.md) | [Tiếng Na Uy](../no/README.md) | [Tiếng Ba Tư (Farsi)](../fa/README.md) | [Tiếng Ba Lan](../pl/README.md) | [Tiếng Bồ Đào Nha (Brazil)](../pt-BR/README.md) | [Tiếng Bồ Đào Nha (Bồ Đào Nha)](../pt-PT/README.md) | [Tiếng Punjabi (Gurmukhi)](../pa/README.md) | [Tiếng Rumani](../ro/README.md) | [Tiếng Nga](../ru/README.md) | [Tiếng Serbia (Chữ Kirin)](../sr/README.md) | [Tiếng Slovakia](../sk/README.md) | [Tiếng Slovenia](../sl/README.md) | [Tiếng Tây Ban Nha](../es/README.md) | [Tiếng Swahili](../sw/README.md) | [Tiếng Thụy Điển](../sv/README.md) | [Tiếng Tagalog (Philippines)](../tl/README.md) | [Tiếng Tamil](../ta/README.md) | [Tiếng Telugu](../te/README.md) | [Tiếng Thái](../th/README.md) | [Tiếng Thổ Nhĩ Kỳ](../tr/README.md) | [Tiếng Ukraina](../uk/README.md) | [Tiếng Urdu](../ur/README.md) | [Tiếng Việt](./README.md) +[Tiếng Ả Rập](../ar/README.md) | [Tiếng Bengal](../bn/README.md) | [Tiếng Bulgaria](../bg/README.md) | [Tiếng Miến Điện (Myanmar)](../my/README.md) | [Tiếng Trung (Giản thể)](../zh-CN/README.md) | [Tiếng Trung (Phồn thể, Hồng Kông)](../zh-HK/README.md) | [Tiếng Trung (Phồn thể, Macau)](../zh-MO/README.md) | [Tiếng Trung (Phồn thể, Đài Loan)](../zh-TW/README.md) | [Tiếng Croatia](../hr/README.md) | [Tiếng Séc](../cs/README.md) | [Tiếng Đan Mạch](../da/README.md) | [Tiếng Hà Lan](../nl/README.md) | [Tiếng Estonia](../et/README.md) | [Tiếng Phần Lan](../fi/README.md) | [Tiếng Pháp](../fr/README.md) | [Tiếng Đức](../de/README.md) | [Tiếng Hy Lạp](../el/README.md) | [Tiếng Do Thái](../he/README.md) | [Tiếng Hindi](../hi/README.md) | [Tiếng Hungary](../hu/README.md) | [Tiếng Indonesia](../id/README.md) | [Tiếng Ý](../it/README.md) | [Tiếng Nhật](../ja/README.md) | [Tiếng Kannada](../kn/README.md) | [Tiếng Khmer](../km/README.md) | [Tiếng Hàn](../ko/README.md) | [Tiếng Lithuania](../lt/README.md) | [Tiếng Mã Lai](../ms/README.md) | [Tiếng Malayalam](../ml/README.md) | [Tiếng Marathi](../mr/README.md) | [Tiếng Nepal](../ne/README.md) | [Tiếng Pidgin Nigeria](../pcm/README.md) | [Tiếng Na Uy](../no/README.md) | [Tiếng Ba Tư (Farsi)](../fa/README.md) | [Tiếng Ba Lan](../pl/README.md) | [Tiếng Bồ Đào Nha (Brazil)](../pt-BR/README.md) | [Tiếng Bồ Đào Nha (Bồ Đào Nha)](../pt-PT/README.md) | [Tiếng Punjabi (Gurmukhi)](../pa/README.md) | [Tiếng Romania](../ro/README.md) | [Tiếng Nga](../ru/README.md) | [Tiếng Serbia (Chữ Kirin)](../sr/README.md) | [Tiếng Slovak](../sk/README.md) | [Tiếng Slovenia](../sl/README.md) | [Tiếng Tây Ban Nha](../es/README.md) | [Tiếng Swahili](../sw/README.md) | [Tiếng Thụy Điển](../sv/README.md) | [Tiếng Tagalog (Philippines)](../tl/README.md) | [Tiếng Tamil](../ta/README.md) | [Tiếng Telugu](../te/README.md) | [Tiếng Thái](../th/README.md) | [Tiếng Thổ Nhĩ Kỳ](../tr/README.md) | [Tiếng Ukraina](../uk/README.md) | [Tiếng Urdu](../ur/README.md) | [Tiếng Việt](./README.md) -> **Ưa thích Clone cục bộ?** +> **Ưu tiên Clone cục bộ?** > -> Kho này bao gồm hơn 50 bản dịch ngôn ngữ, điều này làm tăng đáng kể kích thước tải về. Để clone mà không có bản dịch, sử dụng sparse checkout: +> Kho lưu trữ này bao gồm hơn 50 bản dịch ngôn ngữ, điều này làm tăng đáng kể kích thước tải xuống. Để clone mà không có bản dịch, sử dụng sparse checkout: > > **Bash / macOS / Linux:** > ```bash @@ -52,273 +52,278 @@ Thực hiện các bước sau để bắt đầu sử dụng tài nguyên này: > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> Điều này cung cấp cho bạn tất cả những gì cần thiết để hoàn thành khóa học với tốc độ tải nhanh hơn rất nhiều. +> Điều này cung cấp cho bạn tất cả những gì cần thiết để hoàn thành khóa học với tốc độ tải xuống nhanh hơn nhiều. -## Mục lục +## Mục Lục - Giới thiệu - [Chào mừng đến với Gia đình Phi](./md/01.Introduction/01/01.PhiFamily.md) - [Cài đặt môi trường của bạn](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [Hiểu biết về Các Công nghệ Chính](./md/01.Introduction/01/01.Understandingtech.md) - - [An toàn AI cho các Mô hình Phi](./md/01.Introduction/01/01.AISafety.md) - - [Hỗ trợ Phần cứng Phi](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Mô hình Phi & Khả năng sẵn có trên các nền tảng](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Hiểu biết về các công nghệ chính](./md/01.Introduction/01/01.Understandingtech.md) + - [An toàn AI cho các mô hình Phi](./md/01.Introduction/01/01.AISafety.md) + - [Hỗ trợ phần cứng Phi](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Các mô hình Phi & Khả năng sử dụng trên các nền tảng](./md/01.Introduction/01/01.Edgeandcloud.md) - [Sử dụng Guidance-ai và Phi](./md/01.Introduction/01/01.Guidance.md) - [Mô hình trên GitHub Marketplace](https://github.com/marketplace/models) - - [Danh mục Mô hình AI Azure](https://ai.azure.com) + - [Danh mục mô hình AI Azure](https://ai.azure.com) - Suy luận Phi trong các môi trường khác nhau - [Hugging face](./md/01.Introduction/02/01.HF.md) - [Mô hình GitHub](./md/01.Introduction/02/02.GitHubModel.md) - - [Danh mục Mô hình Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [Danh mục mô hình Microsoft Foundry](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [Bộ công cụ AI VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- Suy luận Gia đình Phi +- Suy luận về Gia đình Phi - [Suy luận Phi trên iOS](./md/01.Introduction/03/iOS_Inference.md) - [Suy luận Phi trên Android](./md/01.Introduction/03/Android_Inference.md) - [Suy luận Phi trên Jetson](./md/01.Introduction/03/Jetson_Inference.md) - - [Suy luận Phi trên AI PC](./md/01.Introduction/03/AIPC_Inference.md) - - [Suy luận Phi với Apple MLX Framework](./md/01.Introduction/03/MLX_Inference.md) + - [Suy luận Phi trên PC AI](./md/01.Introduction/03/AIPC_Inference.md) + - [Suy luận Phi với Framework MLX của Apple](./md/01.Introduction/03/MLX_Inference.md) - [Suy luận Phi trên Máy chủ cục bộ](./md/01.Introduction/03/Local_Server_Inference.md) - - [Suy luận Phi trên Máy chủ từ xa sử dụng AI Toolkit](./md/01.Introduction/03/Remote_Interence.md) + - [Suy luận Phi trên Máy chủ từ xa sử dụng Bộ công cụ AI](./md/01.Introduction/03/Remote_Interence.md) - [Suy luận Phi với Rust](./md/01.Introduction/03/Rust_Inference.md) - [Suy luận Phi--Thị giác tại địa phương](./md/01.Introduction/03/Vision_Inference.md) - [Suy luận Phi với Kaito AKS, Azure Containers (hỗ trợ chính thức)](./md/01.Introduction/03/Kaito_Inference.md) - [Lượng tử hóa Gia đình Phi](./md/01.Introduction/04/QuantifyingPhi.md) - - [Lượng tử hóa Phi-3.5 / 4 bằng llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [Lượng tử hóa Phi-3.5 / 4 bằng tiện ích mở rộng Generative AI cho onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [Lượng tử hóa Phi-3.5 / 4 bằng Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [Lượng tử hóa Phi-3.5 / 4 bằng Apple MLX Framework](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [Lượng tử hóa Phi-3.5 / 4 sử dụng llama.cpp](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [Lượng tử hóa Phi-3.5 / 4 sử dụng mở rộng AI tạo sinh cho onnxruntime](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [Lượng tử hóa Phi-3.5 / 4 sử dụng Intel OpenVINO](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [Lượng tử hóa Phi-3.5 / 4 sử dụng Framework MLX của Apple](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- Đánh giá Phi - - [AI có trách nhiệm](./md/01.Introduction/05/ResponsibleAI.md) +- Đánh giá Phi + - [Responsible AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry cho Đánh giá](./md/01.Introduction/05/AIFoundry.md) - - [Sử dụng Promptflow để Đánh giá](./md/01.Introduction/05/Promptflow.md) + - [Sử dụng Promptflow cho Đánh giá](./md/01.Introduction/05/Promptflow.md) - RAG với Azure AI Search - - [Cách sử dụng Phi-4-mini và Phi-4-multimodal(RAG) với Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [Cách sử dụng Phi-4-mini và Phi-4-multimodal (RAG) với Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [RAG huyền thoại cục bộ không đám mây với SQLite FTS5 và phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Ví dụ phát triển ứng dụng Phi +- Mẫu phát triển ứng dụng Phi - Ứng dụng Văn bản & Trò chuyện - - Mẫu Phi-4 - - [📓] [Trò chuyện với Mô hình Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [Trò chuyện với Mô hình ONNX Phi-4 cục bộ .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [Ứng dụng Console trò chuyện .NET với Phi-4 ONNX sử dụng Sementic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Mẫu Phi-3 / 3.5 + - Mẫu Phi-4 + - [📓] [Trò chuyện với mô hình Phi-4-mini ONNX](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [Trò chuyện với mô hình ONNX Phi-4 cục bộ .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [Ứng dụng Console .NET trò chuyện với Phi-4 ONNX sử dụng Semantic Kernel](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 Ví dụ mẫu - [Chatbot cục bộ trong trình duyệt sử dụng Phi3, ONNX Runtime Web và WebGPU](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - - [OpenVino Chat](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - - [Mở nhiều mô hình - Tương tác Phi-3-mini và OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - Xây dựng bộ đóng gói và sử dụng Phi-3 với MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [Tối ưu hóa mô hình - Cách tối ưu hóa mô hình Phi-3-min cho ONNX Runtime Web với Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [Chat OpenVino](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) + - [Mô hình đa dạng - Phi-3-mini tương tác và OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) + - [MLFlow - Xây dựng wrapper và dùng Phi-3 với MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [Tối ưu mô hình - Cách tối ưu mô hình Phi-3-min cho ONNX Runtime Web bằng Olive](https://github.com/microsoft/Olive/tree/main/examples/phi3) - [Ứng dụng WinUI3 với Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[Mẫu ứng dụng Ghi chú sử dụng AI với Nhiều mô hình WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [Tinh chỉnh và tích hợp các mô hình Phi-3 tùy chỉnh với Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [Tinh chỉnh và tích hợp các mô hình Phi-3 tùy chỉnh với Prompt flow trong Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [Đánh giá mô hình Phi-3 / Phi-3.5 đã tinh chỉnh trong Microsoft Foundry với trọng tâm là các Nguyên tắc AI có Trách nhiệm của Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Mẫu dự đoán ngôn ngữ Phi-3.5-mini-instruct (Tiếng Trung/Anh)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG Chatbot](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + -[Ví dụ mẫu Ứng dụng Ghi chú AI đa mô hình WinUI3](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [Tinh chỉnh và tích hợp mô hình Phi-3 tùy chỉnh với Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [Tinh chỉnh và tích hợp mô hình Phi-3 tùy chỉnh với Prompt flow trong Microsoft Foundry](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [Đánh giá mô hình Phi-3 / Phi-3.5 được tinh chỉnh trong Microsoft Foundry tập trung vào Nguyên tắc AI có Trách nhiệm của Microsoft](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Ví dụ dự đoán ngôn ngữ Phi-3.5-mini-instruct (Tiếng Trung/Anh)](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Chatbot Phi-3.5-Instruct WebGPU RAG](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - [Sử dụng GPU Windows để tạo giải pháp Prompt flow với Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - [Sử dụng Microsoft Phi-3.5 tflite để tạo ứng dụng Android](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Ví dụ Hỏi & Đáp .NET sử dụng mô hình Phi-3 ONNX cục bộ với Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Ứng dụng chat .NET console với Semantic Kernel và Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Ví dụ .NET Q&A sử dụng mô hình ONNX Phi-3 cục bộ dùng Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [Ứng dụng chat .NET trên console với Semantic Kernel và Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) - - Mẫu Code SDK Azure AI Inference - - Mẫu Phi-4 + - Ví dụ mẫu Azure AI Inference SDK dựa trên mã nguồn + - Ví dụ Phi-4 - [📓] [Tạo mã dự án sử dụng Phi-4-multimodal](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - - Mẫu Phi-3 / 3.5 - - [Xây dựng Chat Copilot Visual Studio Code của riêng bạn với Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [Tạo Chat Copilot Agent Visual Studio Code của riêng bạn với Phi-3.5 bằng các Mô hình GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - Ví dụ Phi-3 / 3.5 + - [Xây dựng Visual Studio Code GitHub Copilot Chat riêng của bạn với Microsoft Phi-3 Family](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [Tạo đại lý Chat Copilot riêng cho Visual Studio Code với Phi-3.5 bằng Mô hình GitHub](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - - Mẫu Lý luận Nâng cao - - Mẫu Phi-4 - - [📓] [Mẫu Phi-4-mini-lý luận hoặc Phi-4-lý luận](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - Ví dụ mẫu Lý luận nâng cao + - Ví dụ Phi-4 + - [📓] [Ví dụ Phi-4-mini-lý luận hoặc Phi-4-lý luận](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - [📓] [Tinh chỉnh Phi-4-mini-lý luận với Microsoft Olive](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Tinh chỉnh Phi-4-mini-lý luận với Apple MLX](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [Phi-4-mini-lý luận với Mô hình GitHub](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [Phi-4-mini-lý luận với Mô hình Microsoft Foundry](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - Bản Demo - - [Bản demo Phi-4-mini được lưu trữ trên Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Bản demo Phi-4-multimodal được lưu trữ trên Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - - Mẫu Thị giác - - Mẫu Phi-4 + - Demos + - [Phi-4-mini demo được lưu trữ trên Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal demo được lưu trữ trên Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - Ví dụ mẫu Thị giác + - Ví dụ Phi-4 - [📓] [Sử dụng Phi-4-multimodal để đọc hình ảnh và tạo mã](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - - Mẫu Phi-3 / 3.5 - - [📓][Phi-3-thị giác - chuyển văn bản hình ảnh thành văn bản](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - Ví dụ Phi-3 / 3.5 + - [📓][Phi-3-thị giác-Ảnh văn bản thành văn bản](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-thị giác-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-thị giác nhúng CLIP](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 Tái chế](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-thị giác - Trợ lý ngôn ngữ thị giác - với Phi3-Vision và OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [DEMO: Tái chế Phi-3](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-thị giác - Trợ lý ngôn ngữ thị giác - với Phi3-Thị giác và OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 Thị giác Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 Thị giác OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 Thị giác đa khung hoặc đa hình ảnh mẫu](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [📓][Phi-3.5 Thị giác mẫu đa khung hoặc đa ảnh](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - [Phi-3 Thị giác Mô hình ONNX cục bộ sử dụng Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [Mô hình ONNX cục bộ Phi-3 Thị giác dựa trên Menu sử dụng Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [Mô hình ONNX Thị giác Phi-3 cục bộ dựa trên menu sử dụng Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - - Mẫu Lý luận-Thị giác + - Ví dụ mẫu Lý luận-Thị giác - Phi-4-Lý luận-Thị giác-15B - - [📓] [Sử dụng Phi-4-Lý luận-Thị giác-15B để phát hiện hành vi đi bộ không đúng nơi quy định](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [Sử dụng Phi-4-Lý luận-Thị giác-15B cho toán học](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [Sử dụng Phi-4-Lý luận-Thị giác-15B để phát hiện giao diện người dùng](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [Sử dụng Phi-4-Lý luận-Thị giác-15B để phát hiện jaywalking](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [Sử dụng Phi-4-Lý luận-Thị giác-15B để toán học](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [Sử dụng Phi-4-Lý luận-Thị giác-15B để phát hiện giao diện người dùng (UI)](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - - Mẫu Toán học - - Mẫu Phi-4-Mini-Flash-Reasoning-Instruct [Demo Toán học với Phi-4-Mini-Flash-Reasoning-Instruct](./md/02.Application/09.Math/MathDemo.ipynb) + - Ví dụ mẫu Toán học + - Ví dụ Phi-4-Mini-Flash-Lý luận-Hướng dẫn [Demo Toán học với Phi-4-Mini-Flash-Lý luận-Hướng dẫn](./md/02.Application/09.Math/MathDemo.ipynb) - - Mẫu Âm thanh - - Mẫu Phi-4 + - Ví dụ mẫu Âm thanh + - Ví dụ Phi-4 - [📓] [Trích xuất bản ghi âm thanh sử dụng Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Mẫu âm thanh Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Mẫu dịch nói Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [Ứng dụng console .NET sử dụng Phi-4-multimodal Audio phân tích tệp âm thanh và tạo bản ghi](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - - Mẫu MOE - - Mẫu Phi-3 / 3.5 - - [📓] [Mô hình hỗn hợp Chuyên gia Phi-3.5 (MoEs) Mẫu mạng xã hội](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [Xây dựng dây chuyền Tăng cường Truy xuất (RAG) với NVIDIA NIM Phi-3 MOE, Azure AI Search và LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Ví dụ Âm thanh Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Ví dụ Dịch phát âm Phi-4-multimodal](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [Ứng dụng console .NET sử dụng Phi-4-multimodal Âm thanh để phân tích file âm thanh và tạo bản ghi](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + + - Ví dụ mẫu MOE + - Ví dụ Phi-3 / 3.5 + - [📓] [Mô hình Mixture of Experts Phi-3.5 (MoEs) ví dụ mẫu mạng xã hội](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [Xây dựng pipeline Sinh tạo tăng cường truy xuất thông tin (RAG) với NVIDIA NIM Phi-3 MOE, Azure AI Search và LlamaIndex](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - Mẫu Gọi Hàm - - Mẫu Phi-4 🆕 - - [📓] [Sử dụng Gọi Hàm với Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [Sử dụng Gọi Hàm để tạo đa tác nhân Với Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - Ví dụ mẫu Gọi Hàm + - Ví dụ Phi-4 🆕 + - [📓] [Sử dụng Gọi Hàm Với Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [Sử dụng Gọi Hàm để tạo đa đại lý Với Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - [📓] [Sử dụng Gọi Hàm với Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - [📓] [Sử dụng Gọi Hàm với ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - Mẫu Trộn Đa phương thức - - Mẫu Phi-4 🆕 - - [📓] [Sử dụng Phi-4-multimodal như một Nhà báo Công nghệ](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - Ví dụ mẫu Hòa trộn Đa phương thức + - Ví dụ Phi-4 🆕 + - [📓] [Sử dụng Phi-4-multimodal như nhà báo Công nghệ](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [Ứng dụng console .NET sử dụng Phi-4-multimodal để phân tích hình ảnh](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- Tinh chỉnh mẫu Phi +- Tinh chỉnh Phi Ví dụ - [Tình huống tinh chỉnh](./md/03.FineTuning/FineTuning_Scenarios.md) - - [So sánh Tinh chỉnh và RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [Tinh chỉnh và RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - [Tinh chỉnh để Phi-3 trở thành chuyên gia ngành](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [Tinh chỉnh Phi-3 với Bộ Công cụ AI cho VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [Tinh chỉnh Phi-3 với Dịch vụ Machine Learning Azure](./md/03.FineTuning/Introduce_AzureML.md) + - [Tinh chỉnh Phi-3 với AI Toolkit cho VS Code](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [Tinh chỉnh Phi-3 với Azure Machine Learning Service](./md/03.FineTuning/Introduce_AzureML.md) - [Tinh chỉnh Phi-3 với Lora](./md/03.FineTuning/FineTuning_Lora.md) - [Tinh chỉnh Phi-3 với QLora](./md/03.FineTuning/FineTuning_Qlora.md) - [Tinh chỉnh Phi-3 với Microsoft Foundry](./md/03.FineTuning/FineTuning_AIFoundry.md) - [Tinh chỉnh Phi-3 với Azure ML CLI/SDK](./md/03.FineTuning/FineTuning_MLSDK.md) - [Tinh chỉnh với Microsoft Olive](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Thực hành Tinh chỉnh với Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) + - [Thực hành tinh chỉnh với Microsoft Olive](./md/03.FineTuning/olive-lab/readme.md) - [Tinh chỉnh Phi-3-thị giác với Weights and Bias](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [Tinh chỉnh Phi-3 với Khung Apple MLX](./md/03.FineTuning/FineTuning_MLX.md) - - [Tinh chỉnh Phi-3-thị giác (hỗ trợ chính thức)](./md/03.FineTuning/FineTuning_Vision.md) - - [Tinh chỉnh Phi-3 với Kaito AKS, Azure Containers (Hỗ trợ chính thức)](./md/03.FineTuning/FineTuning_Kaito.md) - - [Tinh chỉnh Phi-3 và 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) -- Thực hành + - [Điều chỉnh tinh Phi-3 với Apple MLX Framework](./md/03.FineTuning/FineTuning_MLX.md) + - [Điều chỉnh tinh Phi-3-vision (hỗ trợ chính thức)](./md/03.FineTuning/FineTuning_Vision.md) + - [Điều chỉnh tinh Phi-3 với Kaito AKS , Azure Containers (hỗ trợ chính thức)](./md/03.FineTuning/FineTuning_Kaito.md) + - [Điều chỉnh tinh Phi-3 và 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + +- Phòng Thực Hành - [Khám phá các mô hình tiên tiến: LLMs, SLMs, phát triển cục bộ và hơn thế nữa](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [Khám phá tiềm năng NLP: Tinh chỉnh với Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) + - [Khai phá tiềm năng NLP: Điều chỉnh tinh với Microsoft Olive](https://github.com/azure/Ignite_FineTuning_workshop) -- Các bài báo và ấn phẩm nghiên cứu học thuật +- Bài Báo và Công Bố Nghiên Cứu Học Thuật - [Textbooks Are All You Need II: báo cáo kỹ thuật phi-1.5](https://arxiv.org/abs/2309.05463) - - [Báo cáo kỹ thuật Phi-3: Mô hình ngôn ngữ có khả năng cao cục bộ trên điện thoại của bạn](https://arxiv.org/abs/2404.14219) + - [Báo cáo kỹ thuật Phi-3: Mô hình ngôn ngữ có năng lực cao trên điện thoại của bạn](https://arxiv.org/abs/2404.14219) - [Báo cáo kỹ thuật Phi-4](https://arxiv.org/abs/2412.08905) - - [Báo cáo kỹ thuật Phi-4-Mini: Mô hình ngôn ngữ đa phương tiện nhỏ gọn nhưng mạnh mẽ thông qua Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) - - [Tối ưu hóa các mô hình ngôn ngữ nhỏ cho gọi hàm trong xe](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) Tinh chỉnh PHI-3 cho câu hỏi trắc nghiệm: Phương pháp, kết quả và thách thức](https://arxiv.org/abs/2501.01588) - - [Báo cáo kỹ thuật Phi-4-reasoning](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Báo cáo kỹ thuật Phi-4-mini-reasoning](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Báo cáo kỹ thuật Phi-4-Mini: Mô hình ngôn ngữ đa phương tiện gọn nhẹ nhưng mạnh mẽ qua Mixture-of-LoRAs](https://arxiv.org/abs/2503.01743) + - [Tối ưu mô hình ngôn ngữ nhỏ cho chức năng gọi hàm trên xe](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) Điều chỉnh PHI-3 cho trả lời câu hỏi trắc nghiệm: Phương pháp, Kết quả và Thách thức](https://arxiv.org/abs/2501.01588) + - [Báo cáo kỹ thuật Phi-4-lý luận](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Báo cáo kỹ thuật Phi-4-mini-lý luận](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## Sử dụng các mô hình Phi +## Sử dụng Mô hình Phi ### Phi trên Microsoft Foundry -Bạn có thể học cách sử dụng Microsoft Phi và cách xây dựng các giải pháp E2E trên các thiết bị phần cứng khác nhau của bạn. Để trải nghiệm Phi cho chính bạn, hãy bắt đầu bằng cách thử chơi với các mô hình và tùy chỉnh Phi cho các kịch bản của bạn bằng cách sử dụng [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai). Bạn có thể tìm hiểu thêm tại Bắt đầu với [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +Bạn có thể học cách sử dụng Microsoft Phi và xây dựng các giải pháp E2E trên các thiết bị phần cứng khác nhau của bạn. Để trải nghiệm Phi trực tiếp, hãy bắt đầu chơi với các mô hình và tùy chỉnh Phi cho các kịch bản của bạn bằng cách sử dụng [Danh mục mô hình Azure AI Microsoft Foundry](https://aka.ms/phi3-azure-ai) bạn có thể tìm hiểu thêm tại Bắt đầu với [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -**Khu vui chơi** -Mỗi mô hình đều có một sân chơi riêng để kiểm tra mô hình [Azure AI Playground](https://aka.ms/try-phi3). +**Sân chơi** +Mỗi mô hình có một sân chơi riêng để thử nghiệm mô hình [Azure AI Playground](https://aka.ms/try-phi3). -### Phi trên GitHub Models +### Phi trên các Mô hình GitHub -Bạn có thể học cách sử dụng Microsoft Phi và cách xây dựng các giải pháp E2E trên các thiết bị phần cứng khác nhau của bạn. Để trải nghiệm Phi cho chính bạn, hãy bắt đầu bằng cách thử chơi với mô hình và tùy chỉnh Phi cho các kịch bản của bạn bằng cách sử dụng [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo). Bạn có thể tìm hiểu thêm tại Bắt đầu với [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) +Bạn có thể học cách sử dụng Microsoft Phi và xây dựng các giải pháp E2E trên các thiết bị phần cứng khác nhau của bạn. Để trải nghiệm Phi trực tiếp, hãy bắt đầu chơi với mô hình và tùy chỉnh Phi cho các kịch bản của bạn bằng cách sử dụng [Danh mục mô hình GitHub](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) bạn có thể tìm hiểu thêm tại Bắt đầu với [Danh mục mô hình GitHub](/md/02.QuickStart/GitHubModel_QuickStart.md) -**Khu vui chơi** -Mỗi mô hình đều có một [sân chơi riêng để kiểm tra mô hình](/md/02.QuickStart/GitHubModel_QuickStart.md). +**Sân chơi** +Mỗi mô hình có một [sân chơi để thử nghiệm mô hình](/md/02.QuickStart/GitHubModel_QuickStart.md). ### Phi trên Hugging Face Bạn cũng có thể tìm thấy mô hình trên [Hugging Face](https://huggingface.co/microsoft) -**Khu vui chơi** -[Sân chơi Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +**Sân chơi** + [Sân chơi Hugging Chat](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 Các khóa học khác + ## 🎒 Các Khóa Học Khác -Nhóm của chúng tôi còn có các khóa học khác! Hãy khám phá: +Nhóm chúng tôi sản xuất các khóa học khác! Hãy xem: ### LangChain -[![LangChain4j cho Người mới bắt đầu](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js cho Người mới bắt đầu](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain cho Người mới bắt đầu](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain for Beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- ### Azure / Edge / MCP / Agents -[![AZD cho Người mới bắt đầu](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI cho Người mới bắt đầu](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP cho Người mới bắt đầu](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents cho Người mới bắt đầu](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Chuỗi AI Tạo Sinh -[![AI Tạo Sinh cho Người mới bắt đầu](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Tạo Sinh (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![AI Tạo Sinh (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![AI Tạo Sinh (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### Học tập Cốt lõi -[![ML cho Người mới bắt đầu](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Khoa học Dữ liệu cho Người mới bắt đầu](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI cho Người mới bắt đầu](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![An ninh mạng cho Người mới bắt đầu](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Phát triển Web cho Người mới bắt đầu](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT cho Người mới bắt đầu](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![Phát triển XR cho Người mới bắt đầu](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) + +### Học cốt lõi +[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - + ### Chuỗi Copilot -[![Copilot cho Lập trình ghép đôi AI](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot cho C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Cuộc phiêu lưu Copilot](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) +[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## AI có trách nhiệm +## AI Có Trách Nhiệm + +Microsoft cam kết giúp khách hàng sử dụng sản phẩm AI của chúng tôi một cách có trách nhiệm, chia sẻ những hiểu biết, và xây dựng các quan hệ hợp tác dựa trên sự tin cậy thông qua các công cụ như Ghi chú Minh bạch và Đánh giá Tác động. Nhiều tài nguyên này có thể được tìm thấy tại [https://aka.ms/RAI](https://aka.ms/RAI). +Phương pháp của Microsoft đối với AI có trách nhiệm dựa trên các nguyên tắc AI của chúng tôi về công bằng, độ tin cậy và an toàn, bảo mật và quyền riêng tư, tính bao trùm, minh bạch và trách nhiệm giải trình. + +Các mô hình ngôn ngữ, hình ảnh và giọng nói quy mô lớn - như các mô hình được sử dụng trong mẫu này - có thể hoạt động theo cách không công bằng, không đáng tin cậy hoặc gây xúc phạm, từ đó gây ra những tổn hại. Vui lòng tham khảo [ghi chú minh bạch dịch vụ Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) để được thông tin về các rủi ro và giới hạn. -Microsoft cam kết giúp khách hàng sử dụng các sản phẩm AI của chúng tôi một cách có trách nhiệm, chia sẻ những bài học của chúng tôi, và xây dựng các quan hệ đối tác dựa trên sự tin cậy thông qua các công cụ như Ghi chú Minh bạch và Đánh giá Tác động. Nhiều tài nguyên này có thể được tìm thấy tại [https://aka.ms/RAI](https://aka.ms/RAI). -Cách tiếp cận AI có trách nhiệm của Microsoft dựa trên các nguyên tắc AI của chúng tôi về công bằng, độ tin cậy và an toàn, quyền riêng tư và bảo mật, tính bao trùm, minh bạch và trách nhiệm giải trình. -Các mô hình ngôn ngữ tự nhiên, hình ảnh và giọng nói quy mô lớn - như những mô hình được sử dụng trong ví dụ này - có thể hành xử theo cách không công bằng, không đáng tin cậy, hoặc gây xúc phạm, từ đó gây hại. Vui lòng tham khảo [Ghi chú Minh bạch dịch vụ Azure OpenAI](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) để được thông tin về các rủi ro và hạn chế. -Cách tiếp cận được khuyến nghị để giảm thiểu những rủi ro này là bao gồm một hệ thống an toàn trong kiến trúc của bạn có thể phát hiện và ngăn chặn hành vi gây hại. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) cung cấp một lớp bảo vệ độc lập, có khả năng phát hiện nội dung do người dùng và AI tạo ra gây hại trong các ứng dụng và dịch vụ. Azure AI Content Safety bao gồm các API văn bản và hình ảnh cho phép bạn phát hiện các tài liệu có hại. Trong Microsoft Foundry, dịch vụ Content Safety cho phép bạn xem, khám phá và thử mã mẫu để phát hiện nội dung gây hại trên các phương thức khác nhau. Tài liệu [bắt đầu nhanh sau đây](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) hướng dẫn bạn cách gửi yêu cầu đến dịch vụ. +Cách tiếp cận được khuyến nghị để giảm thiểu các rủi ro này là bao gồm một hệ thống an toàn trong kiến trúc của bạn có thể phát hiện và ngăn chặn hành vi gây hại. [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) cung cấp một lớp bảo vệ độc lập, có khả năng phát hiện nội dung gây hại do người dùng hoặc AI tạo ra trong các ứng dụng và dịch vụ. Azure AI Content Safety bao gồm các API văn bản và hình ảnh cho phép bạn phát hiện các tài liệu gây hại. Trong Microsoft Foundry, dịch vụ Content Safety cho phép bạn xem, khám phá và thử mã mẫu để phát hiện nội dung gây hại trên các phương thức khác nhau. Tài liệu [bắt đầu nhanh](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) sau đây hướng dẫn bạn cách gửi yêu cầu đến dịch vụ. -Một khía cạnh khác cần lưu ý là hiệu suất tổng thể của ứng dụng. Với các ứng dụng đa phương thức và đa mô hình, chúng tôi coi hiệu suất là hệ thống hoạt động như bạn và người dùng mong đợi, bao gồm cả việc không tạo ra các kết quả gây hại. Việc đánh giá hiệu suất tổng thể của ứng dụng của bạn bằng cách sử dụng [đánh giá Hiệu suất và Chất lượng và Rủi ro và An toàn](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) là rất quan trọng. Bạn cũng có khả năng tạo và đánh giá với [bộ đánh giá tùy chỉnh](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). +Một khía cạnh khác cần xem xét là hiệu suất tổng thể của ứng dụng. Với các ứng dụng đa phương thức và đa mô hình, chúng tôi coi hiệu suất là hệ thống hoạt động đúng như bạn và người dùng mong đợi, bao gồm không tạo ra các kết quả đầu ra gây hại. Việc đánh giá hiệu suất tổng thể của ứng dụng của bạn bằng cách sử dụng [trình đánh giá Hiệu suất và Chất lượng cùng Đánh giá Rủi ro và An toàn](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) rất quan trọng. Bạn cũng có khả năng tạo và đánh giá với các [trình đánh giá tùy chỉnh](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators). -Bạn có thể đánh giá ứng dụng AI của mình trong môi trường phát triển bằng cách sử dụng [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Với bộ dữ liệu kiểm thử hoặc mục tiêu đã cho, các kết quả tạo AI tổng hợp của bạn sẽ được đo lường định lượng bằng các bộ đánh giá tích hợp hoặc bộ đánh giá tùy chỉnh mà bạn lựa chọn. Để bắt đầu với azure ai evaluation sdk nhằm đánh giá hệ thống của bạn, bạn có thể theo dõi [hướng dẫn nhanh](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Khi bạn thực hiện một lần chạy đánh giá, bạn có thể [trực quan hóa kết quả trong Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). +Bạn có thể đánh giá ứng dụng AI của mình trong môi trường phát triển bằng cách sử dụng [Azure AI Evaluation SDK](https://microsoft.github.io/promptflow/index.html). Khi có bộ dữ liệu thử nghiệm hoặc mục tiêu, các kết quả tạo ra bởi ứng dụng AI của bạn sẽ được đo lường định lượng với các trình đánh giá sẵn có hoặc trình đánh giá tùy chỉnh mà bạn chọn. Để bắt đầu với SDK đánh giá azure ai để đánh giá hệ thống của bạn, bạn có thể theo dõi [hướng dẫn bắt đầu nhanh](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk). Khi bạn thực hiện một lần chạy đánh giá, bạn có thể [trực quan hóa kết quả trong Microsoft Foundry](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results). ## Nhãn hiệu -Dự án này có thể chứa nhãn hiệu hoặc logo cho các dự án, sản phẩm hoặc dịch vụ. Việc sử dụng được phép các nhãn hiệu hoặc logo của Microsoft phải tuân thủ và theo [Hướng dẫn Nhãn hiệu & Thương hiệu của Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). -Việc sử dụng nhãn hiệu hoặc logo của Microsoft trong các phiên bản chỉnh sửa của dự án này không được gây hiểu nhầm hoặc ngụ ý rằng Microsoft bảo trợ. Bất kỳ việc sử dụng nhãn hiệu hoặc logo của bên thứ ba phải tuân thủ chính sách của bên thứ ba đó. +Dự án này có thể chứa nhãn hiệu hoặc logo cho các dự án, sản phẩm hoặc dịch vụ. Việc sử dụng nhãn hiệu hoặc logo của Microsoft được ủy quyền phải tuân theo và tuân thủ [Hướng dẫn Nhãn hiệu & Thương hiệu của Microsoft](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general). +Việc sử dụng nhãn hiệu hoặc logo của Microsoft trong các phiên bản sửa đổi của dự án này không được gây nhầm lẫn hoặc ngụ ý sự tài trợ của Microsoft. Bất kỳ việc sử dụng nhãn hiệu hoặc logo của bên thứ ba nào cũng phải tuân theo chính sách của bên thứ ba đó. ## Nhận trợ giúp -Nếu bạn gặp khó khăn hoặc có bất kỳ câu hỏi nào về việc xây dựng ứng dụng AI, hãy tham gia: +Nếu bạn gặp khó khăn hoặc có bất kỳ câu hỏi nào về xây dựng ứng dụng AI, hãy tham gia: [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -Nếu bạn có phản hồi về sản phẩm hoặc lỗi trong quá trình xây dựng, hãy truy cập: +Nếu bạn có phản hồi về sản phẩm hoặc lỗi khi xây dựng, hãy truy cập: [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**Từ chối trách nhiệm**: -Tài liệu này đã được dịch bằng dịch vụ dịch thuật AI [Co-op Translator](https://github.com/Azure/co-op-translator). Mặc dù chúng tôi cố gắng đảm bảo độ chính xác, xin lưu ý rằng các bản dịch tự động có thể chứa lỗi hoặc không chính xác. Tài liệu gốc bằng ngôn ngữ gốc của nó được coi là nguồn chính thức. Đối với thông tin quan trọng, nên sử dụng dịch thuật chuyên nghiệp bởi con người. Chúng tôi không chịu trách nhiệm về bất kỳ sự hiểu lầm hoặc diễn giải sai nào phát sinh từ việc sử dụng bản dịch này. +**Tuyên bố miễn trừ trách nhiệm**: +Tài liệu này đã được dịch bằng dịch vụ dịch thuật AI [Co-op Translator](https://github.com/Azure/co-op-translator). Mặc dù chúng tôi cố gắng đảm bảo độ chính xác, xin lưu ý rằng bản dịch tự động có thể chứa lỗi hoặc sai sót. Tài liệu gốc bằng ngôn ngữ gốc nên được coi là nguồn tin chính thức. Đối với thông tin quan trọng, nên sử dụng dịch vụ dịch thuật chuyên nghiệp bởi con người. Chúng tôi không chịu trách nhiệm về bất kỳ hiểu lầm hoặc giải thích sai nào phát sinh từ việc sử dụng bản dịch này. \ No newline at end of file diff --git a/translations/vi/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/vi/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..c2125083b --- /dev/null +++ b/translations/vi/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 RAG Lai ghép với Microsoft phi-4-mini & SQLite FTS5 (Zero-Cloud SLM)\n", + "\n", + "> **Tác giả:** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **Chủ đề:** Mô hình Ngôn ngữ Nhỏ (SLMs), SQLite FTS5 BM25, Biểu diễn Mật độ, Kết hợp Xếp hạng Lẫn nhau (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. Động lực: Thách thức Gợi nhớ Từ khóa trong SLMs cục bộ\n", + "Các kiến trúc RAG tiêu chuẩn dựa hoàn toàn vào biểu diễn vectơ mật độ thường thất bại trong việc truy xuất chính xác các ký tự số (ví dụ: `2,340,000 TL`, mã hợp đồng, số tài khoản). \n", + "Ngược lại, tìm kiếm từ ngữ thưa thớt (BM25) bỏ lỡ các từ đồng nghĩa ngữ nghĩa và câu hỏi được diễn đạt lại.\n", + "\n", + "Tài liệu hướng dẫn này trình bày cách triển khai **cỗ máy Truy hồi Lai ghép tốc độ cao, trong bộ nhớ** kết hợp:\n", + "1. **Vectơ Mật độ** (Độ tương đồng Cosine)\n", + "2. **Tìm kiếm Từ ngữ Thưa thớt** (SQLite FTS5 BM25)\n", + "3. **Kết hợp Xếp hạng Lẫn nhau (RRF, $k=60$)**\n", + "4. **Sinh Trích dẫn Căn cứ (`[1]`, `[2]`)** với Microsoft `phi-4-mini`.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. Lược đồ SQLite kép (Vector Dày đặc + Bảng Ảo FTS5 BM25)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. Đánh giá hiệu năng Thu nhận mẫu & Thực thi\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Công thức Mở Prompt Có Căn Cứ cho Microsoft phi-4-mini\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**Tuyên bố miễn trừ trách nhiệm**:\nTài liệu này đã được dịch bằng dịch vụ dịch thuật AI [Co-op Translator](https://github.com/Azure/co-op-translator). Mặc dù chúng tôi cố gắng đảm bảo độ chính xác, xin lưu ý rằng bản dịch tự động có thể chứa lỗi hoặc sai sót. Tài liệu gốc bằng ngôn ngữ gốc nên được coi là nguồn tin chính thức. Đối với thông tin quan trọng, nên sử dụng dịch vụ dịch thuật chuyên nghiệp bởi con người. Chúng tôi không chịu trách nhiệm về bất kỳ hiểu lầm hoặc giải thích sai nào phát sinh từ việc sử dụng bản dịch này.\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/zh-CN/.co-op-translator.json b/translations/zh-CN/.co-op-translator.json index 344ce2e4e..605f6fd70 100644 --- a/translations/zh-CN/.co-op-translator.json +++ b/translations/zh-CN/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "zh-CN" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-20T07:45:59+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T18:52:56+00:00", "source_file": "README.md", "language_code": "zh-CN" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "zh-CN" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T18:53:49+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "zh-CN" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:29:49+00:00", diff --git a/translations/zh-CN/README.md b/translations/zh-CN/README.md index b92244a4f..7576acecb 100644 --- a/translations/zh-CN/README.md +++ b/translations/zh-CN/README.md @@ -1,42 +1,42 @@ -# Phi Cookbook：微软Phi模型的动手示例 +# Phi 烹饪书：微软 Phi 模型的动手示例 -[![在GitHub Codespaces中打开并使用示例](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![在Dev Containers中打开](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![在 GitHub Codespaces 中打开并使用示例](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![在 Dev Containers 中打开](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub贡献者](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub问题](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub拉取请求](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![欢迎PR](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 贡献者](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 问题](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 拉取请求](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![欢迎提交 PR](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub关注者](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub分叉](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub星标](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 关注者](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 派生](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 星标](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi 是微软开发的一系列开源AI模型。 +Phi 是微软开发的一系列开源 AI 模型。 -Phi 目前是最强大且具有成本效益的小型语言模型（SLM），在多语言、推理、文本/聊天生成、编码、图像、音频及其他场景中表现出非常好的基准测试成绩。 +Phi 当前是最强大且性价比最高的小型语言模型（SLM），在多语言、推理、文本/聊天生成、编码、图像、音频和其他场景中拥有非常优秀的基准表现。 -您可以将Phi部署到云端或边缘设备上，并且可以轻松使用有限的计算能力构建生成式AI应用。 +你可以将 Phi 部署到云端或边缘设备，并且可以轻松构建有限计算能力下的生成式 AI 应用。 -请按照以下步骤开始使用这些资源： -1. **Fork 仓库**：点击 [![GitHub分叉](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +按照以下步骤开始使用这些资源： +1. **Fork 仓库**：点击 [![GitHub 派生](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) 2. 克隆仓库：`git clone https://github.com/microsoft/PhiCookBook.git` -3. [**加入微软AI Discord社区，结识专家和开发者同行**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +3. [**加入微软 AI Discord 社区，结识专家和同行开发者**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) -![cover](../../translated_images/zh-CN/cover.eb18d1b9605d754b.webp) +![封面](../../translated_images/zh-CN/cover.eb18d1b9605d754b.webp) ### 🌐 多语言支持 -#### 通过GitHub Action支持（自动化且随时更新） +#### 通过 GitHub Action 支持（自动化且始终保持最新） -[阿拉伯语](../ar/README.md) | [孟加拉语](../bn/README.md) | [保加利亚语](../bg/README.md) | [缅甸语 (Myanmar)](../my/README.md) | [中文（简体）](./README.md) | [中文（繁体，香港）](../zh-HK/README.md) | [中文（繁体，澳门）](../zh-MO/README.md) | [中文（繁体，台湾）](../zh-TW/README.md) | [克罗地亚语](../hr/README.md) | [捷克语](../cs/README.md) | [丹麦语](../da/README.md) | [荷兰语](../nl/README.md) | [爱沙尼亚语](../et/README.md) | [芬兰语](../fi/README.md) | [法语](../fr/README.md) | [德语](../de/README.md) | [希腊语](../el/README.md) | [希伯来语](../he/README.md) | [印地语](../hi/README.md) | [匈牙利语](../hu/README.md) | [印尼语](../id/README.md) | [意大利语](../it/README.md) | [日语](../ja/README.md) | [卡纳达语](../kn/README.md) | [高棉语](../km/README.md) | [韩语](../ko/README.md) | [立陶宛语](../lt/README.md) | [马来语](../ms/README.md) | [马拉雅拉姆语](../ml/README.md) | [马拉地语](../mr/README.md) | [尼泊尔语](../ne/README.md) | [尼日利亚皮钦语](../pcm/README.md) | [挪威语](../no/README.md) | [波斯语 (法尔西语)](../fa/README.md) | [波兰语](../pl/README.md) | [葡萄牙语 (巴西)](../pt-BR/README.md) | [葡萄牙语 (葡萄牙)](../pt-PT/README.md) | [旁遮普语 (古鲁穆奇)](../pa/README.md) | [罗马尼亚语](../ro/README.md) | [俄语](../ru/README.md) | [塞尔维亚语 (西里尔字母)](../sr/README.md) | [斯洛伐克语](../sk/README.md) | [斯洛文尼亚语](../sl/README.md) | [西班牙语](../es/README.md) | [斯瓦希里语](../sw/README.md) | [瑞典语](../sv/README.md) | [他加禄语 (菲律宾语)](../tl/README.md) | [泰米尔语](../ta/README.md) | [泰卢固语](../te/README.md) | [泰语](../th/README.md) | [土耳其语](../tr/README.md) | [乌克兰语](../uk/README.md) | [乌尔都语](../ur/README.md) | [越南语](../vi/README.md) +[阿拉伯语](../ar/README.md) | [孟加拉语](../bn/README.md) | [保加利亚语](../bg/README.md) | [缅甸语 (Myanmar)](../my/README.md) | [中文（简体）](./README.md) | [中文（繁体，香港）](../zh-HK/README.md) | [中文（繁体，澳门）](../zh-MO/README.md) | [中文（繁体，台湾）](../zh-TW/README.md) | [克罗地亚语](../hr/README.md) | [捷克语](../cs/README.md) | [丹麦语](../da/README.md) | [荷兰语](../nl/README.md) | [爱沙尼亚语](../et/README.md) | [芬兰语](../fi/README.md) | [法语](../fr/README.md) | [德语](../de/README.md) | [希腊语](../el/README.md) | [希伯来语](../he/README.md) | [印地语](../hi/README.md) | [匈牙利语](../hu/README.md) | [印尼语](../id/README.md) | [意大利语](../it/README.md) | [日语](../ja/README.md) | [卡纳达语](../kn/README.md) | [高棉语](../km/README.md) | [韩语](../ko/README.md) | [立陶宛语](../lt/README.md) | [马来语](../ms/README.md) | [马拉雅拉姆语](../ml/README.md) | [马拉地语](../mr/README.md) | [尼泊尔语](../ne/README.md) | [尼日利亚皮钦语](../pcm/README.md) | [挪威语](../no/README.md) | [波斯语 (法尔斯语)](../fa/README.md) | [波兰语](../pl/README.md) | [葡萄牙语（巴西）](../pt-BR/README.md) | [葡萄牙语（葡萄牙）](../pt-PT/README.md) | [旁遮普语（古鲁姆ukhi）](../pa/README.md) | [罗马尼亚语](../ro/README.md) | [俄语](../ru/README.md) | [塞尔维亚语（西里尔字母）](../sr/README.md) | [斯洛伐克语](../sk/README.md) | [斯洛文尼亚语](../sl/README.md) | [西班牙语](../es/README.md) | [斯瓦希里语](../sw/README.md) | [瑞典语](../sv/README.md) | [他加禄语（菲律宾语）](../tl/README.md) | [泰米尔语](../ta/README.md) | [泰卢固语](../te/README.md) | [泰语](../th/README.md) | [土耳其语](../tr/README.md) | [乌克兰语](../uk/README.md) | [乌尔都语](../ur/README.md) | [越南语](../vi/README.md) -> **偏好本地克隆？** +> **更喜欢本地克隆？** > -> 本仓库包含50多种语言的翻译，显著增加了下载大小。若想不含翻译的克隆，可使用稀疏检出： +> 本仓库包含 50 多种语言的翻译，极大增加了下载大小。若想不含翻译内容克隆，请使用 sparse checkout： > > **Bash / macOS / Linux：** > ```bash @@ -45,280 +45,285 @@ Phi 目前是最强大且具有成本效益的小型语言模型（SLM），在 > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` > -> **CMD (Windows)：** +> **CMD（Windows）：** > ```cmd > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> 这样可让您以更快的速度获取完成课程所需的全部内容。 +> 这样你能获得完成课程所需的一切，同时大幅缩短下载时间。 ## 目录 - 介绍 - - [欢迎加入Phi大家庭](./md/01.Introduction/01/01.PhiFamily.md) - - [环境搭建](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [欢迎加入 Phi 家族](./md/01.Introduction/01/01.PhiFamily.md) + - [环境设置](./md/01.Introduction/01/01.EnvironmentSetup.md) - [关键技术理解](./md/01.Introduction/01/01.Understandingtech.md) - - [Phi模型的AI安全](./md/01.Introduction/01/01.AISafety.md) - - [Phi硬件支持](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi模型及跨平台可用性](./md/01.Introduction/01/01.Edgeandcloud.md) - - [使用Guidance-ai和Phi](./md/01.Introduction/01/01.Guidance.md) - - [GitHub市场模型](https://github.com/marketplace/models) - - [Azure AI模型目录](https://ai.azure.com) - -- 不同环境中推理Phi + - [Phi 模型的 AI 安全](./md/01.Introduction/01/01.AISafety.md) + - [Phi 硬件支持](./md/01.Introduction/01/01.Hardwaresupport.md) + - [Phi 模型与跨平台可用性](./md/01.Introduction/01/01.Edgeandcloud.md) + - [使用 Guidance-ai 和 Phi](./md/01.Introduction/01/01.Guidance.md) + - [GitHub 市场模型](https://github.com/marketplace/models) + - [Azure AI 模型目录](https://ai.azure.com) + +- 在不同环境推理 Phi - [Hugging face](./md/01.Introduction/02/01.HF.md) - - [GitHub模型](./md/01.Introduction/02/02.GitHubModel.md) - - [Microsoft Foundry模型目录](./md/01.Introduction/02/03.AzureAIFoundry.md) + - [GitHub 模型](./md/01.Introduction/02/02.GitHubModel.md) + - [微软 Foundry 模型目录](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [AI工具包VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry本地](./md/01.Introduction/02/07.FoundryLocal.md) - -- 推理Phi家族 - - [在iOS中推理Phi](./md/01.Introduction/03/iOS_Inference.md) - - [在Android中推理Phi](./md/01.Introduction/03/Android_Inference.md) - - [在Jetson中推理Phi](./md/01.Introduction/03/Jetson_Inference.md) - - [在AI PC中推理Phi](./md/01.Introduction/03/AIPC_Inference.md) - - [使用Apple MLX框架推理Phi](./md/01.Introduction/03/MLX_Inference.md) - - [在本地服务器推理Phi](./md/01.Introduction/03/Local_Server_Inference.md) - - [使用AI工具包在远程服务器推理Phi](./md/01.Introduction/03/Remote_Interence.md) - - [使用Rust推理Phi](./md/01.Introduction/03/Rust_Inference.md) - - [本地推理Phi视觉模型](./md/01.Introduction/03/Vision_Inference.md) - - [使用Kaito AKS和Azure容器推理Phi（官方支持）](./md/01.Introduction/03/Kaito_Inference.md) -- [Phi家族量化](./md/01.Introduction/04/QuantifyingPhi.md) - - [使用llama.cpp对Phi-3.5 / 4进行量化](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [使用onnxruntime的生成式AI扩展量化Phi-3.5 / 4](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - - [使用Intel OpenVINO量化Phi-3.5 / 4](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [使用Apple MLX框架量化Phi-3.5 / 4](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - -- Phi评估 - - [负责任的AI](./md/01.Introduction/05/ResponsibleAI.md) - - [Microsoft Foundry评估](./md/01.Introduction/05/AIFoundry.md) - - [使用Promptflow进行评估](./md/01.Introduction/05/Promptflow.md) + - [Foundry 本地](./md/01.Introduction/02/07.FoundryLocal.md) + +- 推理 Phi 家族 + - [iOS 上推理 Phi](./md/01.Introduction/03/iOS_Inference.md) + - [Android 上推理 Phi](./md/01.Introduction/03/Android_Inference.md) + - [Jetson 上推理 Phi](./md/01.Introduction/03/Jetson_Inference.md) + - [AI PC 上推理 Phi](./md/01.Introduction/03/AIPC_Inference.md) + - [使用 Apple MLX 框架推理 Phi](./md/01.Introduction/03/MLX_Inference.md) + - [本地服务器推理 Phi](./md/01.Introduction/03/Local_Server_Inference.md) + - [使用 AI Toolkit 远程服务器推理 Phi](./md/01.Introduction/03/Remote_Interence.md) + - [使用 Rust 推理 Phi](./md/01.Introduction/03/Rust_Inference.md) + - [本地推理 Phi--Vision](./md/01.Introduction/03/Vision_Inference.md) + - [使用 Kaito AKS、Azure 容器（官方支持）推理 Phi](./md/01.Introduction/03/Kaito_Inference.md) +- [Phi 家族量化](./md/01.Introduction/04/QuantifyingPhi.md) + - [使用 llama.cpp 量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) + - [使用 onnxruntime 的生成式 AI 扩展量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [使用 Intel OpenVINO 量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) + - [使用 Apple MLX 框架量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + +- 评估 Phi + - [负责任的 AI](./md/01.Introduction/05/ResponsibleAI.md) + - [微软 Foundry 评估](./md/01.Introduction/05/AIFoundry.md) + - [使用 Promptflow 进行评估](./md/01.Introduction/05/Promptflow.md) -- 与Azure AI搜索的RAG - - [如何在Azure AI搜索中使用Phi-4-mini和Phi-4-multimodal (RAG)](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) - -- Phi应用开发示例 - - 文本和聊天应用 - - Phi-4示例 - - [📓] [使用Phi-4-mini ONNX模型聊天](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [使用Phi-4本地ONNX模型的.NET聊天](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [使用语义内核的Phi-4 ONNX .NET控制台聊天应用](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - - Phi-3 / 3.5示例 - - [使用Phi3、ONNX Runtime Web和WebGPU在浏览器中部署本地聊天机器人](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) +- 结合 Azure AI 搜索的 RAG + - [如何使用 Phi-4-mini 和 Phi-4-multimodal (RAG) 与 Azure AI 搜索](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [使用 SQLite FTS5 和 phi-4-mini 的零云本地混合 RAG](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) + +- Phi 应用开发示例 + - 文本与聊天应用 + - Phi-4 示例 + - [📓] [与 Phi-4-mini ONNX 模型聊天](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [使用 Phi-4 本地 ONNX 模型的聊天 .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [使用 Sementic Kernel 的 Phi-4 ONNX 聊天 .NET 控制台应用](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + + - Phi-3 / 3.5 示例 + - [使用 Phi3、ONNX Runtime Web 和 WebGPU 在浏览器中本地聊天机器人](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino 聊天](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [多模型 - 交互式 Phi-3-mini 和 OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - 构建包装器并使用 Phi-3 与 MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [模型优化 - 如何使用 Olive 优化 Phi-3-min 模型以适配 ONNX Runtime Web](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [带有 Phi-3 mini-4k-instruct-onnx 的 WinUI3 应用](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 多模型 AI 驱动的笔记应用示例](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [模型优化 - 如何使用 Olive 优化 Phi-3-min 模型以用于 ONNX Runtime Web](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 应用示例，内置 Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 多模型 AI 助力记事本应用示例](https://github.com/microsoft/ai-powered-notes-winui3-sample) - [微调并集成自定义 Phi-3 模型与 Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [在 Microsoft Foundry 中微调并集成自定义 Phi-3 模型与 Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [在 Microsoft Foundry 中评估微调后的 Phi-3 / Phi-3.5 模型，重点关注微软的负责任 AI 原则](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct 语言预测示例（中英双语）](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [在 Microsoft Foundry 中使用 Prompt flow 微调并集成自定义 Phi-3 模型](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [在 Microsoft Foundry 中评估微调后的 Phi-3 / Phi-3.5 模型，重点关注微软的责任人工智能原则](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct 语言预测示例（中英文）](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG 聊天机器人](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [使用 Windows GPU 创建基于 Phi-3.5-Instruct ONNX 的 Prompt flow 解决方案](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [使用 Microsoft Phi-3.5 tflite 创建 Android 应用](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [使用本地 ONNX Phi-3 模型的 .NET Q&A 示例，使用 Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [带有 Semantic Kernel 和 Phi-3 的控制台聊天 .NET 应用](../../md/04.HOL/dotnet/src/LabsPhi302) + - [使用 Windows GPU 创建带有 Phi-3.5-Instruct ONNX 的 Prompt flow 解决方案](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [使用微软 Phi-3.5 tflite 创建 Android 应用](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [使用 Microsoft.ML.OnnxRuntime 的本地 ONNX Phi-3 模型的问答 .NET 示例](../../md/04.HOL/dotnet/src/LabsPhi301) + - [使用 Semantic Kernel 和 Phi-3 的控制台聊天 .NET 应用](../../md/04.HOL/dotnet/src/LabsPhi302) - Azure AI 推理 SDK 代码示例 - Phi-4 示例 - [📓] [使用 Phi-4-multimodal 生成项目代码](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 示例 - - [使用 Microsoft Phi-3 家族构建属于你的 Visual Studio Code GitHub Copilot 聊天](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [使用 GitHub 模型和 Phi-3.5 创建自己的 Visual Studio Code 聊天 Copilot 代理](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [使用微软 Phi-3 家族构建你自己的 Visual Studio Code GitHub Copilot 聊天](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [使用 GitHub 模型和 Phi-3.5 创建你自己的 Visual Studio Code 聊天 Copilot 代理](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - 高级推理示例 - Phi-4 示例 - [📓] [Phi-4-mini-推理或 Phi-4-推理示例](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [使用 Microsoft Olive 对 Phi-4-mini-推理进行微调](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [使用 Apple MLX 对 Phi-4-mini-推理进行微调](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [使用 GitHub 模型的 Phi-4-mini-推理](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [使用 Microsoft Foundry 模型的 Phi-4-mini-推理](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - [📓] [用 Microsoft Olive 微调 Phi-4-mini-推理](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [用 Apple MLX 微调 Phi-4-mini-推理](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [与 GitHub 模型的 Phi-4-mini-推理](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [与 Microsoft Foundry 模型的 Phi-4-mini-推理](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - 演示 - [Phi-4-mini 演示，托管于 Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal 演示，托管于 Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal 演示，托管于 Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - 视觉示例 - Phi-4 示例 - [📓] [使用 Phi-4-multimodal 读取图像并生成代码](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 示例 - - [📓][Phi-3-vision-图像文字到文字](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-图像文本转换](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP 嵌入](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [演示：Phi-3 回收](https://github.com/jennifermarsman/PhiRecycling/) + - [DEMO: Phi-3 回收](https://github.com/jennifermarsman/PhiRecycling/) - [Phi-3-vision - 视觉语言助手 - 结合 Phi3-Vision 和 OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 视觉 Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 视觉 OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - [📓][Phi-3.5 视觉多帧或多图像示例](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 视觉本地 ONNX 模型，使用 Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [基于菜单的 Phi-3 视觉本地 ONNX 模型，使用 Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [使用 Microsoft.ML.OnnxRuntime .NET 的 Phi-3 视觉本地 ONNX 模型](../../md/04.HOL/dotnet/src/LabsPhi303) + - [基于菜单的使用 Microsoft.ML.OnnxRuntime .NET 的 Phi-3 视觉本地 ONNX 模型](../../md/04.HOL/dotnet/src/LabsPhi304) - 推理-视觉示例 - Phi-4-Reasoning-Vision-15B - - [📓] [使用 Phi-4-Reasoning-Vision-15B 进行乱穿马路检测](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [使用 Phi-4-Reasoning-Vision-15B 进行数学推理](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [使用 Phi-4-Reasoning-Vision-15B 检测乱穿马路](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [使用 Phi-4-Reasoning-Vision-15B 进行数学运算](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - [📓] [使用 Phi-4-Reasoning-Vision-15B 进行 UI 检测](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - 数学示例 - - Phi-4-Mini-Flash-Reasoning-Instruct 示例 [使用 Phi-4-Mini-Flash-Reasoning-Instruct 的数学演示](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-Mini-Flash-Reasoning-Instruct 示例 [使用 Phi-4-Mini-Flash-Reasoning-Instruct 的数学演示](./md/02.Application/09.Math/MathDemo.ipynb) - 音频示例 - Phi-4 示例 - - [📓] [使用 Phi-4-multimodal 提取音频转录](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [使用 Phi-4-multimodal 提取音频转录文本](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - [📓] [Phi-4-multimodal 音频示例](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Phi-4-multimodal 语音翻译示例](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET 控制台应用，使用 Phi-4-multimodal 分析音频文件并生成转录](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [.NET 控制台应用，使用 Phi-4-multimodal 分析音频文件并生成转录文本](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE 示例 - Phi-3 / 3.5 示例 - [📓] [Phi-3.5 专家混合模型 (MoEs) 社交媒体示例](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [构建基于 NVIDIA NIM Phi-3 MOE、Azure AI 搜索和 LlamaIndex 的检索增强生成 (RAG) 流水线](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [使用 NVIDIA NIM Phi-3 MOE、Azure AI 搜索和 LlamaIndex 构建检索增强生成（RAG）流水线](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - 函数调用示例 - Phi-4 示例 🆕 - - [📓] [使用 Phi-4-mini 进行函数调用](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [使用函数调用创建多代理，基于 Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [使用 Ollama 进行函数调用](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [使用 ONNX 进行函数调用](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - [📓] [使用 Phi-4-mini 的函数调用](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [使用 Phi-4-mini 的函数调用创建多代理](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [使用函数调用和 Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [使用函数调用和 ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - 多模态混合示例 - Phi-4 示例 🆕 - - [📓] [作为科技记者使用 Phi-4-multimodal](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [📓] [使用 Phi-4-multimodal 作为科技记者](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - [.NET 控制台应用，使用 Phi-4-multimodal 分析图像](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) -- 微调 Phi 示例 +- Phi 微调示例 - [微调场景](./md/03.FineTuning/FineTuning_Scenarios.md) - [微调与 RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [微调让 Phi-3 成为行业专家](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [使用 AI Toolkit for VS Code 微调 Phi-3](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [使用 Azure 机器学习服务微调 Phi-3](./md/03.FineTuning/Introduce_AzureML.md) - - [使用 Lora 微调 Phi-3](./md/03.FineTuning/FineTuning_Lora.md) - - [使用 QLora 微调 Phi-3](./md/03.FineTuning/FineTuning_Qlora.md) - - [使用 Microsoft Foundry 微调 Phi-3](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [使用 Azure ML CLI/SDK 微调 Phi-3](./md/03.FineTuning/FineTuning_MLSDK.md) - - [使用 Microsoft Olive 微调](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive 实战微调实验室](./md/03.FineTuning/olive-lab/readme.md) + - [微调 Phi-3 使其成为行业专家](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [使用 AI Toolkit for VS Code 对 Phi-3 进行微调](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [使用 Azure 机器学习服务对 Phi-3 进行微调](./md/03.FineTuning/Introduce_AzureML.md) + - [使用 Lora 对 Phi-3 进行微调](./md/03.FineTuning/FineTuning_Lora.md) + - [使用 QLora 对 Phi-3 进行微调](./md/03.FineTuning/FineTuning_Qlora.md) + - [使用 Microsoft Foundry 对 Phi-3 进行微调](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [使用 Azure ML CLI/SDK 对 Phi-3 进行微调](./md/03.FineTuning/FineTuning_MLSDK.md) + - [使用 Microsoft Olive 进行微调](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Microsoft Olive 实践训练营](./md/03.FineTuning/olive-lab/readme.md) - [使用 Weights and Bias 微调 Phi-3-vision](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [使用 Apple MLX 框架微调 Phi-3](./md/03.FineTuning/FineTuning_MLX.md) - [微调 Phi-3-vision（官方支持）](./md/03.FineTuning/FineTuning_Vision.md) - - [使用Kaito AKS和Azure容器微调Phi-3（官方支持）](./md/03.FineTuning/FineTuning_Kaito.md) - - [微调Phi-3和3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + - [使用 Kaito AKS 和 Azure 容器微调 Phi-3（官方支持）](./md/03.FineTuning/FineTuning_Kaito.md) + - [微调 Phi-3 和 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - 实操实验室 - - [探索前沿模型：LLMs、SLMs、本地开发及更多](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [释放NLP潜力：使用Microsoft Olive微调](https://github.com/azure/Ignite_FineTuning_workshop) + - [探索前沿模型：LLMs、SLMs、本地开发等](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [释放 NLP 潜力：使用 Microsoft Olive 微调](https://github.com/azure/Ignite_FineTuning_workshop) -- 学术研究论文与出版物 - - [Textbooks Are All You Need II：phi-1.5 技术报告](https://arxiv.org/abs/2309.05463) - - [Phi-3 技术报告：一款极具能力的本地手机语言模型](https://arxiv.org/abs/2404.14219) +- 学术研究论文和出版物 + - [教材就是全部你所需 II: phi-1.5 技术报告](https://arxiv.org/abs/2309.05463) + - [Phi-3 技术报告：在您的手机上运行的高能力语言模型](https://arxiv.org/abs/2404.14219) - [Phi-4 技术报告](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini 技术报告：通过LoRAs混合实现紧凑而强大的多模态语言模型](https://arxiv.org/abs/2503.01743) - - [优化车载功能调用的小型语言模型](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) 针对多项选择题微调PHI-3的方法论、结果与挑战](https://arxiv.org/abs/2501.01588) - - [Phi-4-推理 技术报告](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - - [Phi-4-mini-推理 技术报告](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) + - [Phi-4-Mini 技术报告：通过 LoRA 混合实现紧凑而强大的多模态语言模型](https://arxiv.org/abs/2503.01743) + - [优化用于车内功能调用的小型语言模型](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) 微调 PHI-3 进行多项选择题问答：方法论、结果和挑战](https://arxiv.org/abs/2501.01588) + - [Phi-4 推理技术报告](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) + - [Phi-4-mini 推理技术报告](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) -## 使用Phi模型 +## 使用 Phi 模型 -### Microsoft Foundry上的Phi +### Microsoft Foundry 上的 Phi -您可以了解如何使用Microsoft Phi以及如何在不同硬件设备上构建端到端解决方案。要亲自体验Phi，可以先使用模型并根据您的场景定制Phi，使用[Microsoft Foundry Azure AI模型目录](https://aka.ms/phi3-azure-ai)。您可以在[Microsoft Foundry快速入门](/md/02.QuickStart/AzureAIFoundry_QuickStart.md)了解更多。 +您可以学习如何使用 Microsoft Phi 以及如何在不同硬件设备上构建端到端解决方案。想要亲自体验 Phi，可以从玩转模型和针对您的场景定制 Phi 开始，使用 [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) 。您可以在[Microsoft Foundry 入门指南](/md/02.QuickStart/AzureAIFoundry_QuickStart.md)了解更多信息。 游乐场 -每个模型都有专门的游乐场用于测试模型[Azure AI Playground](https://aka.ms/try-phi3)。 +每个模型都有专门的游乐场供您测试模型，访问 [Azure AI Playground](https://aka.ms/try-phi3)。 -### GitHub模型上的Phi +### GitHub 上的 Phi 模型 -您可以了解如何使用Microsoft Phi以及如何在不同硬件设备上构建端到端解决方案。要亲自体验Phi，可以先使用模型并根据您的场景定制Phi，使用[GitHub模型目录](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo)。您可以在[GitHub模型目录快速入门](/md/02.QuickStart/GitHubModel_QuickStart.md)了解更多。 +您可以学习如何使用 Microsoft Phi 以及如何在不同硬件设备上构建端到端解决方案。想要亲自体验 Phi，可以从玩转模型和针对您的场景定制 Phi 开始，使用 [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) 。您可以在[GitHub Model Catalog 入门指南](/md/02.QuickStart/GitHubModel_QuickStart.md)了解更多。 游乐场 -每个模型都有专门的[游乐场用于测试模型](/md/02.QuickStart/GitHubModel_QuickStart.md)。 +每个模型都有专门的[游乐场供测试模型](/md/02.QuickStart/GitHubModel_QuickStart.md)。 -### Hugging Face上的Phi +### Hugging Face 上的 Phi -您也可以在[Hugging Face](https://huggingface.co/microsoft)找到该模型。 +您还可以在 [Hugging Face](https://huggingface.co/microsoft) 上找到该模型。 -游乐场 -[Hugging Chat游乐场](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +游乐场 + [Hugging Chat 游乐场](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 其他课程 + ## 🎒 其他课程 -我们的团队还制作了其他课程！请查看： +我们团队还制作了其他课程！请查看： ### LangChain -[![LangChain4j入门](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js入门](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain入门](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain4j 初学者](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js 初学者](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain 初学者](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- ### Azure / Edge / MCP / Agents -[![AZD入门](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI入门](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP入门](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents入门](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AZD 初学者](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI 初学者](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP 初学者](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents 初学者](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- -### 生成式AI系列 -[![生成式AI入门](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![生成式AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![生成式AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![生成式AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) +### 生成式 AI 系列 +[![生成式 AI 初学者](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![生成式 AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![生成式 AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![生成式 AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- ### 核心学习 -[![机器学习入门](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![数据科学入门](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![人工智能入门](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![网络安全入门](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web开发入门](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![物联网入门](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR开发入门](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) +[![机器学习初学者](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![数据科学初学者](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![人工智能初学者](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![网络安全初学者](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![Web 开发初学者](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![物联网初学者](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR 开发初学者](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- ### Copilot 系列 -[![AI配对编程Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![C#/.NET Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![面向 AI 配对编程的 Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![面向 C#/.NET 的 Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot 冒险](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) -## 负责任的AI +## 负责任的 AI + +微软致力于帮助我们的客户负责任地使用我们的 AI 产品，分享我们的经验，并通过透明度说明和影响评估等工具建立基于信任的合作伙伴关系。许多资源可以在[https://aka.ms/RAI](https://aka.ms/RAI)找到。 +微软的负责任 AI 方法基于我们的 AI 原则，包括公平性、可靠性和安全性、隐私和安全性、包容性、透明度和问责制。 + +大规模自然语言、图像和语音模型——如本示例中使用的模型——可能会表现出不公平、不可靠或冒犯性的行为，从而导致伤害。请参阅[Azure OpenAI 服务透明度说明](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text)以了解风险和限制。 -微软致力于帮助客户负责任地使用我们的AI产品，分享我们的经验，并通过透明度说明和影响评估等工具建立基于信任的合作伙伴关系。许多资源可在[https://aka.ms/RAI](https://aka.ms/RAI)找到。 -微软的负责任AI方法基于我们的AI原则：公平性、可靠性与安全、隐私与安全性、包容性、透明度和问责制。 -大规模自然语言、图像和语音模型——如本示例中使用的模型——可能表现出不公平、不可靠或冒犯性的行为，从而导致损害。请查阅[Azure OpenAI服务透明度说明](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text)，了解相关风险和限制。 -减轻这些风险的推荐方法是在您的架构中包含一个安全系统，该系统能够检测和防止有害行为。[Azure AI 内容安全](https://learn.microsoft.com/azure/ai-services/content-safety/overview) 提供了一个独立的保护层，能够检测应用程序和服务中有害的用户生成内容和 AI 生成内容。Azure AI 内容安全包括文本和图像 API，允许您检测有害材料。在 Microsoft Foundry 中，内容安全服务允许您查看、探索并试用检测不同模态中有害内容的示例代码。以下[快速入门文档](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest)将指导您如何向该服务发出请求。 +减轻这些风险的推荐方法是在您的架构中包含一个安全系统，该系统能够检测并阻止有害行为。[Azure AI 内容安全](https://learn.microsoft.com/azure/ai-services/content-safety/overview) 提供了一个独立的保护层，能够检测应用和服务中由用户生成和 AI 生成的有害内容。Azure AI 内容安全包含文本和图像 API，允许您检测有害材料。在 Microsoft Foundry 中，内容安全服务允许您查看、探索并试用不同模态下检测有害内容的示例代码。以下[快速入门文档](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) 将指导您如何向该服务发起请求。 -另一个需要考虑的方面是整体应用性能。对于多模态和多模型应用，我们认为性能意味着系统的表现符合您和用户的期望，包括不生成有害输出。重要的是使用[性能和质量与风险和安全评估器](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)来评估您整体应用的性能。您还可以使用[自定义评估器](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators)创建和进行评估。 +另一个需要考虑的方面是整体应用性能。对于多模态和多模型应用，我们认为性能意味着系统按您和用户的预期运行，包括不会生成有害的输出。评估整体应用性能时，使用[性能与质量及风险和安全评估器](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) 非常重要。您还可以创建和评估[自定义评估器](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators)。 -您可以使用[Azure AI 评估 SDK](https://microsoft.github.io/promptflow/index.html)在开发环境中评估您的 AI 应用。通过测试数据集或目标，使用内置评估器或您选择的自定义评估器对您的生成式 AI 应用生成内容进行定量测量。若要开始使用 azure ai evaluation sdk 评估您的系统，您可以按照[快速入门指南](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)操作。一旦执行评估运行，您可以在[Microsoft Foundry 中可视化结果](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)。 +您可以使用[Azure AI 评估 SDK](https://microsoft.github.io/promptflow/index.html)在开发环境中评估您的 AI 应用。无论是给定测试数据集还是目标，您的生成式 AI 应用生成内容都可以通过内置评估器或您选择的自定义评估器进行定量测量。要开始使用 azure ai 评估 sdk 评估您的系统，您可以参考[快速入门指南](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)。执行评估运行后，您可以在[Microsoft Foundry 中可视化结果](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)。 ## 商标 -本项目可能包含项目、产品或服务的商标或徽标。微软商标或徽标的授权使用须遵守并必须遵守[微软商标与品牌指南](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)。 -修改版本中使用微软商标或徽标不得引起混淆或暗示微软赞助。任何使用第三方商标或徽标的行为均须遵守相关第三方的政策。 +本项目可能包含项目、产品或服务的商标或徽标。微软商标或徽标的授权使用须遵守并遵循[微软商标与品牌指南](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)。 +修改版本中使用微软商标或徽标不得引起混淆或暗示微软的赞助。对第三方商标或徽标的任何使用均须遵守相应第三方的政策。 ## 获取帮助 -如果遇到困难或对构建 AI 应用有任何疑问，请加入： +如果您遇到困难或对构建 AI 应用有任何疑问，请加入： [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -如果您有产品反馈或在构建过程中遇到错误，请访问： +如果您在构建过程中有产品反馈或错误，欢迎访问： [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**免责声明**： -本文档由 AI 翻译服务 [Co-op Translator](https://github.com/Azure/co-op-translator) 生成翻译。虽然我们努力确保准确性，但请注意自动翻译可能包含错误或不准确之处。以文档的原始语言版本为权威来源。对于重要信息，建议采用专业人工翻译。对于因使用本翻译而产生的任何误解或误释，我们不承担任何责任。 +**免责声明**： +本文件由 AI 翻译服务 [Co-op Translator](https://github.com/Azure/co-op-translator) 翻译完成。尽管我们力求准确，但请注意，自动翻译可能包含错误或不准确之处。原始语言版文件应视为权威来源。对于重要信息，建议使用专业人工翻译。我们对因使用本翻译而产生的任何误解或误释不承担责任。 \ No newline at end of file diff --git a/translations/zh-CN/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/zh-CN/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..aa3a5e2bb --- /dev/null +++ b/translations/zh-CN/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 使用 Microsoft phi-4-mini 和 SQLite FTS5 的混合 RAG（零云 SLM）\n", + "\n", + "> **作者：** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **重点：** 小型语言模型 (SLMs)、SQLite FTS5 BM25、密集嵌入、逆序等级融合 (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. 动机：本地 SLM 中的关键词召回困境\n", + "仅依赖密集向量嵌入的标准 RAG 架构通常无法检索精确的数字标记（例如 `2,340,000 TL`、合同代码、账号）。 \n", + "相反，稀疏词汇搜索（BM25）会遗失语义同义词和意译问题。\n", + "\n", + "本手册演示如何实现一个 **高速、内存中的混合检索引擎**，结合：\n", + "1. 密集向量（余弦相似度）\n", + "2. 稀疏词汇搜索（SQLite FTS5 BM25）\n", + "3. **逆序等级融合 (RRF, $k=60$)**\n", + "4. 使用 Microsoft `phi-4-mini` 的 有据可查引文生成 (`[1]`, `[2]`)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. 双SQLite模式（稠密向量 + 虚拟FTS5 BM25表）\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. 示例摄取与执行基准测试\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. 用于 Microsoft phi-4-mini 的有依据提示制定\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**免责声明**：\n本文件由 AI 翻译服务 [Co-op Translator](https://github.com/Azure/co-op-translator) 翻译完成。尽管我们力求准确，但请注意，自动翻译可能包含错误或不准确之处。原始语言版文件应视为权威来源。对于重要信息，建议使用专业人工翻译。我们对因使用本翻译而产生的任何误解或误释不承担责任。\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/zh-HK/.co-op-translator.json b/translations/zh-HK/.co-op-translator.json index 8c8dd2a54..e3ec6f087 100644 --- a/translations/zh-HK/.co-op-translator.json +++ b/translations/zh-HK/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "zh-HK" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:33:25+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T16:22:44+00:00", "source_file": "README.md", "language_code": "zh-HK" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "zh-HK" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T16:27:22+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "zh-HK" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:29:56+00:00", diff --git a/translations/zh-HK/README.md b/translations/zh-HK/README.md index 528bb050e..a5283fd53 100644 --- a/translations/zh-HK/README.md +++ b/translations/zh-HK/README.md @@ -1,29 +1,29 @@ -# Phi Cookbook: 使用 Microsoft Phi 模型的實作範例 +# Phi 菜譜：使用微軟 Phi 模型的實作範例 -[![Open and use the samples in GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![Open in Dev Containers](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![在 GitHub Codespaces 開啟並使用範例](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![在 Dev Containers 中開啟](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) -[![GitHub contributors](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub issues](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub pull-requests](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 貢獻者](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 問題](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 拉取請求](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) +[![歡迎 PR](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub watchers](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub stars](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 觀察者](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 派生](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 星標](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi 是由 Microsoft 開發的一系列開源 AI 模型。 +Phi 是由微軟開發的一系列開源 AI 模型。 -Phi 目前是最強大且具成本效益的小型語言模型（SLM），在多語言、推理、文本／聊天生成、程式碼、影像、音訊及其他場景均有出色的基準表現。 +Phi 目前是最強大且具成本效益的小型語言模型（SLM），在多語言、推理、文本／聊天生成、程式碼、影像、音訊及其他場景中表現卓越。 -您可以將 Phi 部署到雲端或邊緣裝置，且能輕鬆以有限的運算能力構建生成式 AI 應用程式。 +您可以將 Phi 部署到雲端或邊緣裝置，並且能夠輕鬆以有限的運算能力構建生成式 AI 應用程式。 -請依照以下步驟開始使用這些資源： -1. **Fork 儲存庫**：點擊 [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Clone 儲存庫**：`git clone https://github.com/microsoft/PhiCookBook.git` -3. [**加入 Microsoft AI Discord 社群，與專家及開發者交流**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +按照以下步驟開始使用這些資源： +1. **Fork 倉庫**：點擊 [![GitHub 派生](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. 複製倉庫：`git clone https://github.com/microsoft/PhiCookBook.git` +3. [**加入 Microsoft AI Discord 社群，與專家和其他開發者交流**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/zh-HK/cover.eb18d1b9605d754b.webp) @@ -32,11 +32,11 @@ Phi 目前是最強大且具成本效益的小型語言模型（SLM），在多 #### 透過 GitHub Action 支援（自動且持續更新） -[阿拉伯語](../ar/README.md) | [孟加拉語](../bn/README.md) | [保加利亞語](../bg/README.md) | [緬甸語 (Myanmar)](../my/README.md) | [中文（簡體）](../zh-CN/README.md) | [中文（繁體，香港）](./README.md) | [中文（繁體，澳門）](../zh-MO/README.md) | [中文（繁體，台灣）](../zh-TW/README.md) | [克羅地亞語](../hr/README.md) | [捷克語](../cs/README.md) | [丹麥語](../da/README.md) | [荷蘭語](../nl/README.md) | [愛沙尼亞語](../et/README.md) | [芬蘭語](../fi/README.md) | [法語](../fr/README.md) | [德語](../de/README.md) | [希臘語](../el/README.md) | [希伯來語](../he/README.md) | [印地語](../hi/README.md) | [匈牙利語](../hu/README.md) | [印尼語](../id/README.md) | [義大利語](../it/README.md) | [日語](../ja/README.md) | [卡納達語](../kn/README.md) | [高棉語](../km/README.md) | [韓語](../ko/README.md) | [立陶宛語](../lt/README.md) | [馬來語](../ms/README.md) | [馬拉雅拉姆語](../ml/README.md) | [馬拉地語](../mr/README.md) | [尼泊爾語](../ne/README.md) | [奈及利亞洋涇浜語](../pcm/README.md) | [挪威語](../no/README.md) | [波斯語 (法爾西語)](../fa/README.md) | [波蘭語](../pl/README.md) | [巴西葡萄牙語](../pt-BR/README.md) | [葡萄牙語 (葡萄牙)](../pt-PT/README.md) | [旁遮普語 (Gurmukhi)](../pa/README.md) | [羅馬尼亞語](../ro/README.md) | [俄語](../ru/README.md) | [塞爾維亞語 (西里爾字母)](../sr/README.md) | [斯洛伐克語](../sk/README.md) | [斯洛維尼亞語](../sl/README.md) | [西班牙語](../es/README.md) | [斯瓦希里語](../sw/README.md) | [瑞典語](../sv/README.md) | [他加祿語 (菲律賓語)](../tl/README.md) | [泰米爾語](../ta/README.md) | [泰盧固語](../te/README.md) | [泰語](../th/README.md) | [土耳其語](../tr/README.md) | [烏克蘭語](../uk/README.md) | [烏爾都語](../ur/README.md) | [越南語](../vi/README.md) +[阿拉伯語](../ar/README.md) | [孟加拉語](../bn/README.md) | [保加利亞語](../bg/README.md) | [緬甸語](../my/README.md) | [中文（簡體）](../zh-CN/README.md) | [中文（繁體，香港）](./README.md) | [中文（繁體，澳門）](../zh-MO/README.md) | [中文（繁體，台灣）](../zh-TW/README.md) | [克羅埃西亞語](../hr/README.md) | [捷克語](../cs/README.md) | [丹麥語](../da/README.md) | [荷蘭語](../nl/README.md) | [愛沙尼亞語](../et/README.md) | [芬蘭語](../fi/README.md) | [法語](../fr/README.md) | [德語](../de/README.md) | [希臘語](../el/README.md) | [希伯來語](../he/README.md) | [印地語](../hi/README.md) | [匈牙利語](../hu/README.md) | [印尼語](../id/README.md) | [義大利語](../it/README.md) | [日語](../ja/README.md) | [卡納達語](../kn/README.md) | [高棉語](../km/README.md) | [韓語](../ko/README.md) | [立陶宛語](../lt/README.md) | [馬來語](../ms/README.md) | [馬拉雅拉姆語](../ml/README.md) | [馬拉地語](../mr/README.md) | [尼泊爾語](../ne/README.md) | [奈及利亞皮欽語](../pcm/README.md) | [挪威語](../no/README.md) | [波斯語（法爾西語）](../fa/README.md) | [波蘭語](../pl/README.md) | [葡萄牙語（巴西）](../pt-BR/README.md) | [葡萄牙語（葡萄牙）](../pt-PT/README.md) | [旁遮普語（古爾穆奇文）](../pa/README.md) | [羅馬尼亞語](../ro/README.md) | [俄語](../ru/README.md) | [塞爾維亞語（西里爾字母）](../sr/README.md) | [斯洛伐克語](../sk/README.md) | [斯洛維尼亞語](../sl/README.md) | [西班牙語](../es/README.md) | [史瓦希里語](../sw/README.md) | [瑞典語](../sv/README.md) | [他加祿語（菲律賓語）](../tl/README.md) | [泰米爾語](../ta/README.md) | [泰盧固語](../te/README.md) | [泰語](../th/README.md) | [土耳其語](../tr/README.md) | [烏克蘭語](../uk/README.md) | [烏爾都語](../ur/README.md) | [越南語](../vi/README.md) -> **喜歡本地端 Clone？** +> **偏好本地複製？** > -> 本儲存庫包含 50 多種語言的翻譯，會大幅增加下載大小。若想不下載翻譯內容，請使用 sparse checkout： +> 此倉庫包含超過 50 種語言翻譯，導致下載大小大幅增加。若想克隆時不包含翻譯，請使用稀疏檢出（sparse checkout）： > > **Bash / macOS / Linux：** > ```bash @@ -52,86 +52,88 @@ Phi 目前是最強大且具成本效益的小型語言模型（SLM），在多 > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> 這樣能讓您以更快速度下載，並取得完成課程所需的所有內容。 +> 這樣能讓你以更快的速度下載所需全部資源，完成課程。 ## 目錄 -- 介紹 +- 簡介 - [歡迎加入 Phi 家族](./md/01.Introduction/01/01.PhiFamily.md) - - [設定您的開發環境](./md/01.Introduction/01/01.EnvironmentSetup.md) - - [理解主要技術](./md/01.Introduction/01/01.Understandingtech.md) - - [Phi 模型的 AI 安全性](./md/01.Introduction/01/01.AISafety.md) + - [環境設定](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [關鍵技術認識](./md/01.Introduction/01/01.Understandingtech.md) + - [Phi 模型的 AI 安全](./md/01.Introduction/01/01.AISafety.md) - [Phi 硬體支援](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi 模型及各平台可用性](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Phi 模型及跨平台可用性](./md/01.Introduction/01/01.Edgeandcloud.md) - [使用 Guidance-ai 與 Phi](./md/01.Introduction/01/01.Guidance.md) - - [GitHub Marketplace 模型](https://github.com/marketplace/models) + - [GitHub 市集模型](https://github.com/marketplace/models) - [Azure AI 模型目錄](https://ai.azure.com) -- 在不同環境推論 Phi - - [Hugging face](./md/01.Introduction/02/01.HF.md) +- 不同環境下的 Phi 推論 + - [Hugging Face](./md/01.Introduction/02/01.HF.md) - [GitHub 模型](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry 模型目錄](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [AI 工具包 VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) -- 推論 Phi 家族 - - [在 iOS 推論 Phi](./md/01.Introduction/03/iOS_Inference.md) - - [在 Android 推論 Phi](./md/01.Introduction/03/Android_Inference.md) - - [在 Jetson 推論 Phi](./md/01.Introduction/03/Jetson_Inference.md) - - [在 AI PC 推論 Phi](./md/01.Introduction/03/AIPC_Inference.md) - - [使用 Apple MLX Framework 推論 Phi](./md/01.Introduction/03/MLX_Inference.md) - - [在本地伺服器推論 Phi](./md/01.Introduction/03/Local_Server_Inference.md) - - [使用 AI Toolkit 在遠端伺服器推論 Phi](./md/01.Introduction/03/Remote_Interence.md) - - [使用 Rust 推論 Phi](./md/01.Introduction/03/Rust_Inference.md) - - [本地端推論 Phi--Vision](./md/01.Introduction/03/Vision_Inference.md) - - [使用 Kaito AKS、Azure Containers (官方支援) 推論 Phi](./md/01.Introduction/03/Kaito_Inference.md) -- [Phi 家族量化](./md/01.Introduction/04/QuantifyingPhi.md) +- Phi Family 推論 + - [iOS 上的 Phi 推論](./md/01.Introduction/03/iOS_Inference.md) + - [Android 上的 Phi 推論](./md/01.Introduction/03/Android_Inference.md) + - [Jetson 上的 Phi 推論](./md/01.Introduction/03/Jetson_Inference.md) + - [AI PC 上的 Phi 推論](./md/01.Introduction/03/AIPC_Inference.md) + - [使用 Apple MLX 框架的 Phi 推論](./md/01.Introduction/03/MLX_Inference.md) + - [本地伺服器上的 Phi 推論](./md/01.Introduction/03/Local_Server_Inference.md) + - [使用 AI 工具包在遠端伺服器進行 Phi 推論](./md/01.Introduction/03/Remote_Interence.md) + - [使用 Rust 的 Phi 推論](./md/01.Introduction/03/Rust_Inference.md) + - [本地的 Phi--視覺推論](./md/01.Introduction/03/Vision_Inference.md) + - [使用 Kaito AKS，Azure 容器（官方支援）的 Phi 推論](./md/01.Introduction/03/Kaito_Inference.md) +- [Phi Family 量化](./md/01.Introduction/04/QuantifyingPhi.md) - [使用 llama.cpp 量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [使用 onnxruntime 的生成式 AI 擴充功能量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [使用 onnxruntime 生成式 AI 擴充量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [使用 Intel OpenVINO 量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [使用 Apple MLX Framework 量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [使用 Apple MLX 框架量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- Phi 評估 - - [負責任 AI](./md/01.Introduction/05/ResponsibleAI.md) +- Phi 評估 + - [負責任的 AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry 評估](./md/01.Introduction/05/AIFoundry.md) - [使用 Promptflow 進行評估](./md/01.Introduction/05/Promptflow.md) -- 使用 Azure AI Search 進行 RAG - - [如何結合 Phi-4-mini 和 Phi-4-multimodal(RAG) 與 Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- 與 Azure AI 搜尋的 RAG + - [如何使用 Phi-4-mini 和 Phi-4-multimodal (RAG) 搭配 Azure AI 搜尋](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [零雲端本地混合 RAG，使用 SQLite FTS5 與 phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- Phi 應用開發範例 - - 文本及聊天應用 - - Phi-4 範例 +- Phi 應用程式開發範例 + - 文本與聊天應用 + - Phi-4 範例 - [📓] [與 Phi-4-mini ONNX 模型聊天](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [使用 Phi-4 本地 ONNX 模型於 .NET 建立聊天](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [使用 Semantic Kernel 和 Phi-4 ONNX 的 .NET 主控台聊天應用](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - [使用本地 Phi-4 ONNX 模型聊天（.NET）](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [使用 Sementic Kernel 的 Phi-4 ONNX .NET 控制台應用](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 範例 - - [在瀏覽器使用 Phi3、ONNX Runtime Web 及 WebGPU 的本地聊天機器人](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [使用 Phi3、ONNX Runtime Web 與 WebGPU 於瀏覽器本地聊天機械人](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino 聊天](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [多模型 - 互動式 Phi-3-mini 與 OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - 建立封裝及使用 Phi-3 與 MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [模型優化 - 如何使用 Olive 為 ONNX Runtime Web 優化 Phi-3-min 模型](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 應用程式與 Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 多模型 AI 便箋應用範例](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [微調及整合自訂 Phi-3 模型與 Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [在 Microsoft Foundry 中微調及整合自訂 Phi-3 模型與 Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [在 Microsoft Foundry 評估微調後的 Phi-3 / Phi-3.5 模型，聚焦於微軟的負責任 AI 原則](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct 語言預測範例（中英）](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG 聊天機器人](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [使用 Windows GPU 建立 Phi-3.5-Instruct ONNX Prompt flow 解決方案](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [使用 Microsoft Phi-3.5 tflite 建立 Android 應用程式](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [使用 Microsoft.ML.OnnxRuntime 的本地 ONNX Phi-3 模型的 Q&A .NET 範例](../../md/04.HOL/dotnet/src/LabsPhi301) - - [結合 Semantic Kernel 及 Phi-3 的 .NET 控制台聊天應用程式](../../md/04.HOL/dotnet/src/LabsPhi302) - - - Azure AI 推理 SDK 代碼範例 + - [MLFlow - 建立包裝器並使用 Phi-3 與 MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [模型優化 - 如何使用 Olive 優化 ONNX Runtime Web 的 Phi-3-min 模型](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [WinUI3 應用程式示範，使用 Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 多模型 AI 助力筆記應用示範](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [微調及集成自訂 Phi-3 模型與 Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [在 Microsoft Foundry 使用 Prompt flow 微調及集成自訂 Phi-3 模型](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [於 Microsoft Foundry 評估已微調的 Phi-3 / Phi-3.5 模型，聚焦於微軟責任 AI 原則](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct 語言預測範例（中文/英文）](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [Phi-3.5-Instruct WebGPU RAG 聊天機械人](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [使用 Windows GPU 建立 Prompt flow 解決方案，搭配 Phi-3.5-Instruct ONNX](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [使用微軟 Phi-3.5 tflite 建立 Android 應用程式](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [使用 Microsoft.ML.OnnxRuntime 利用本地 ONNX Phi-3 模型的問答 .NET 範例](../../md/04.HOL/dotnet/src/LabsPhi301) + - [結合 Semantic Kernel 與 Phi-3 的控制臺聊天 .NET 應用程式](../../md/04.HOL/dotnet/src/LabsPhi302) + + - Azure AI 推理 SDK 程式碼範例 - Phi-4 範例 - - [📓] [使用 Phi-4-multimodal 生成專案代碼](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [📓] [使用 Phi-4-multimodal 生成專案程式碼](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 範例 - - [使用 Microsoft Phi-3 家族建立您的 Visual Studio Code GitHub Copilot Chat](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [使用 GitHub 模型及 Phi-3.5 創建您的 Visual Studio Code 聊天 Copilot 代理](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [使用 Microsoft Phi-3 系列打造您自己的 Visual Studio Code GitHub Copilot 聊天助理](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [利用 GitHub 模型及 Phi-3.5 建立您自己的 Visual Studio Code 聊天助理代理人](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - 進階推理範例 - Phi-4 範例 @@ -140,184 +142,188 @@ Phi 目前是最強大且具成本效益的小型語言模型（SLM），在多 - [📓] [使用 Apple MLX 微調 Phi-4-mini-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - [📓] [使用 GitHub 模型的 Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - [📓] [使用 Microsoft Foundry 模型的 Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - 展示 - - [在 Hugging Face Spaces 上託管的 Phi-4-mini 展示](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [在 Hugging Face Spaces 上託管的 Phi-4-multimodal 展示](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - 示範 + - [Phi-4-mini 示範，托管於 Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal 示範，托管於 Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - 視覺範例 - Phi-4 範例 - - [📓] [使用 Phi-4-multimodal 讀取圖像並生成代碼](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [📓] [使用 Phi-4-multimodal 讀取圖片並生成程式碼](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 範例 - - [📓][Phi-3-vision-圖像文字到文本](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-圖片文字到文字](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP 嵌入](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [DEMO: Phi-3 回收利用](https://github.com/jennifermarsman/PhiRecycling/) + - [展示：Phi-3 回收](https://github.com/jennifermarsman/PhiRecycling/) - [Phi-3-vision - 視覺語言助理 - 使用 Phi3-Vision 與 OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 視覺 Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 視覺 OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 視覺多幀或多圖像範例](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 視覺本地 ONNX 模型，使用 Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [基於選單的 Phi-3 視覺本地 ONNX 模型，使用 Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision 多幀或多圖像範例](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [使用 Microsoft.ML.OnnxRuntime .NET 的本地 Phi-3 Vision ONNX 模型](../../md/04.HOL/dotnet/src/LabsPhi303) + - [基於選單的本地 Phi-3 Vision ONNX 模型，使用 Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - 推理-視覺範例 - Phi-4-Reasoning-Vision-15B - - [📓] [使用 Phi-4-Reasoning-Vision-15B 偵測違規穿越馬路](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [使用 Phi-4-Reasoning-Vision-15B 進行數學運算](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [使用 Phi-4-Reasoning-Vision-15B 偵測介面設計](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [使用 Phi-4-Reasoning-Vision-15B 偵測闖紅燈行人](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [使用 Phi-4-Reasoning-Vision-15B 解數學題](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [使用 Phi-4-Reasoning-Vision-15B 偵測 UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - 數學範例 - - Phi-4-Mini-Flash-Reasoning-Instruct 範例 [Phi-4-Mini-Flash-Reasoning-Instruct 數學演示](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-Mini-Flash-Reasoning-Instruct 範例 [使用 Phi-4-Mini-Flash-Reasoning-Instruct 的數學示範](./md/02.Application/09.Math/MathDemo.ipynb) - 音訊範例 - Phi-4 範例 - - [📓] [使用 Phi-4-multimodal 擷取音訊轉錄](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal 音訊範例](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - - [📓] [Phi-4-multimodal 語音翻譯範例](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET 控制台應用程式使用 Phi-4-multimodal 音訊分析音訊檔案並生成轉錄](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [📓] [使用 Phi-4-multimodal 擷取音訊逐字稿](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal 音訊示範](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [Phi-4-multimodal 語音翻譯示範](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) + - [.NET 控制臺應用程式使用 Phi-4-multimodal 分析音訊檔並生成逐字稿](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE 範例 - Phi-3 / 3.5 範例 - - [📓] [Phi-3.5 混合專家模型 (MoEs) 社群媒體範例](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [使用 NVIDIA NIM Phi-3 MOE、Azure AI Search 及 LlamaIndex 建立檢索增強生成（RAG）管線](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 專家混合模型 (MoEs) 社群媒體示範](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [結合 NVIDIA NIM Phi-3 MOE、Azure AI 搜尋與 LlamaIndex 建立檢索增強生成 (RAG) 管線](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - 函數調用範例 + - 函數呼叫範例 - Phi-4 範例 🆕 - - [📓] [使用 Phi-4-mini 的函數調用](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [使用函數調用建立 Phi-4-mini 多代理](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [與 Ollama 使用函數調用](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [與 ONNX 使用函數調用](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - [📓] [使用 Phi-4-mini 的函數呼叫範例](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [使用函數呼叫構建多代理人，搭配 Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [在 Ollama 中使用函數呼叫](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [在 ONNX 中使用函數呼叫](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - 多模態混合範例 - Phi-4 範例 🆕 - - [📓] [使用 Phi-4-multimodal 作為科技記者](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET 控制台應用程式使用 Phi-4-multimodal 分析圖像](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [📓] [將 Phi-4-multimodal 用作科技記者](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [.NET 控制臺應用程式使用 Phi-4-multimodal 分析影像](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - Phi 微調範例 - [微調情境](./md/03.FineTuning/FineTuning_Scenarios.md) - [微調與 RAG 比較](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [微調讓 Phi-3 成為產業專家](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [使用 AI Toolkit for VS Code 微調 Phi-3](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [使用 Azure Machine Learning Service 微調 Phi-3](./md/03.FineTuning/Introduce_AzureML.md) + - [微調 Phi-3 讓其成為行業專家](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [使用 AI 工具組在 VS Code 微調 Phi-3](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [使用 Azure 機器學習服務微調 Phi-3](./md/03.FineTuning/Introduce_AzureML.md) - [使用 Lora 微調 Phi-3](./md/03.FineTuning/FineTuning_Lora.md) - [使用 QLora 微調 Phi-3](./md/03.FineTuning/FineTuning_Qlora.md) - [使用 Microsoft Foundry 微調 Phi-3](./md/03.FineTuning/FineTuning_AIFoundry.md) - [使用 Azure ML CLI/SDK 微調 Phi-3](./md/03.FineTuning/FineTuning_MLSDK.md) - [使用 Microsoft Olive 微調](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [Microsoft Olive 實作實驗室微調](./md/03.FineTuning/olive-lab/readme.md) + - [Microsoft Olive 實作實驗室微調示範](./md/03.FineTuning/olive-lab/readme.md) - [使用 Weights and Bias 微調 Phi-3-vision](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [使用 Apple MLX 框架微調 Phi-3](./md/03.FineTuning/FineTuning_MLX.md) + + - [使用 Apple MLX Framework 微調 Phi-3](./md/03.FineTuning/FineTuning_MLX.md) - [微調 Phi-3-vision（官方支援）](./md/03.FineTuning/FineTuning_Vision.md) - - [使用 Kaito AKS、Azure Containers（官方支援）微調 Phi-3](./md/03.FineTuning/FineTuning_Kaito.md) + - [使用 Kaito AKS、Azure Containers 微調 Phi-3（官方支援）](./md/03.FineTuning/FineTuning_Kaito.md) - [微調 Phi-3 和 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - 實作實驗室 - - [探索尖端模型：LLMs、SLMs、本地開發及更多](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [釋放 NLP 潛力：使用 Microsoft Olive 進行微調](https://github.com/azure/Ignite_FineTuning_workshop) + - [探索前沿模型：大型語言模型(LLMs)、小型語言模型(SLMs)、本地開發等](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [解鎖自然語言處理潛力：使用 Microsoft Olive 微調](https://github.com/azure/Ignite_FineTuning_workshop) -- 學術研究論文與出版物 - - [Textbooks Are All You Need II: phi-1.5 技術報告](https://arxiv.org/abs/2309.05463) - - [Phi-3 技術報告：在你的手機上本地運行的高性能語言模型](https://arxiv.org/abs/2404.14219) +- 學術研究論文與出版品 + - [Textbooks Are All You Need II：phi-1.5 技術報告](https://arxiv.org/abs/2309.05463) + - [Phi-3 技術報告：高性能的本地手機語言模型](https://arxiv.org/abs/2404.14219) - [Phi-4 技術報告](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini 技術報告：透過混合 LoRA 打造體積小但強大的多模態語言模型](https://arxiv.org/abs/2503.01743) + - [Phi-4-Mini 技術報告：透過混合 LoRAs 實現的緊湊強大多模態語言模型](https://arxiv.org/abs/2503.01743) - [優化車載功能呼叫的小型語言模型](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) 微調 PHI-3 用於多重選擇題問答：方法論、結果與挑戰](https://arxiv.org/abs/2501.01588) + - [(WhyPHI) 微調 PHI-3 用於多選題問答：方法、結果與挑戰](https://arxiv.org/abs/2501.01588) - [Phi-4 推理技術報告](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Phi-4-mini 推理技術報告](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## 使用 Phi 模型 -### Microsoft Foundry 上的 Phi +### 在 Microsoft Foundry 上使用 Phi -你可以學習如何使用 Microsoft Phi 以及如何在不同硬件設備上建置端到端解決方案。想親身體驗 Phi，請先嘗試操作模型並根據你的需求自訂 Phi，使用 [Microsoft Foundry Azure AI 模型目錄](https://aka.ms/phi3-azure-ai) 你可以在 Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) 中了解更多。 +你可以學習如何使用 Microsoft Phi 以及如何在不同硬件設備上構建端到端解決方案。若想親自體驗 Phi，請先試用模型並根據你的場景自訂 Phi，使用 [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) ，也可以在 [Microsoft Foundry 入門指南](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) 學習更多。 -**Playground** -每個模型都有專屬的遊戲場用來測試模型 [Azure AI Playground](https://aka.ms/try-phi3)。 +遊樂場 +每個模型都有專屬的遊樂場來測試模型，[Azure AI 遊樂場](https://aka.ms/try-phi3)。 -### GitHub 模型上的 Phi +### 在 GitHub 模型庫上使用 Phi -你可以學習如何使用 Microsoft Phi 以及如何在不同硬件設備上建置端到端解決方案。想親身體驗 Phi，請先嘗試操作模型並根據你的需求自訂 Phi，使用 [GitHub 模型目錄](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) 你可以在 Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) 了解更多。 +你可以學習如何使用 Microsoft Phi 以及如何在不同硬件設備上構建端到端解決方案。若想親自體驗 Phi，請先試用模型並根據你的場景自訂 Phi，使用 [GitHub 模型庫](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) ，也可以在 [GitHub 模型庫入門指南](/md/02.QuickStart/GitHubModel_QuickStart.md) 學習更多。 -**Playground** -每個模型都有專屬的 [遊戲場用來測試模型](/md/02.QuickStart/GitHubModel_QuickStart.md)。 +遊樂場 +每個模型都有專屬的[遊樂場來測試模型](/md/02.QuickStart/GitHubModel_QuickStart.md)。 -### Hugging Face 上的 Phi +### 在 Hugging Face 上使用 Phi -你也可以在 [Hugging Face](https://huggingface.co/microsoft) 找到該模型。 +你也可以在 [Hugging Face](https://huggingface.co/microsoft) 找到模型。 -**Playground** -[Hugging Chat playground](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +遊樂場 + [Hugging Chat 遊樂場](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 其他課程 + ## 🎒 其他課程 -我們團隊還製作了其他課程！請參考： +我們團隊還製作了其他課程！歡迎查看： -### LangChain -[![LangChain4j 入門](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js 入門](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain 入門](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +### LangChain +[![LangChain4j 入門](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js 入門](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain 入門](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / 邊緣 / MCP / Agents -[![AZD 入門](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI 入門](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP 入門](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents 入門](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agents +[![Azure Developer 入門](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI 入門](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP 入門](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI 代理人入門](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### 生成式 AI 系列 -[![生成式 AI 入門](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![生成式 AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![生成式 AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![生成式 AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### 生成式 AI 系列 +[![生成式 AI 入門](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![生成式 AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![生成式 AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![生成式 AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### 核心學習 -[![機器學習入門](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![資料科學入門](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![人工智能入門](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![資安入門](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![網頁開發入門](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![物聯網入門](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR 開發入門](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) + +### 核心學習 +[![機器學習入門](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![數據科學入門](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![人工智能入門](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![網絡安全入門](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![網頁開發入門](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![物聯網入門](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR 開發入門](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Copilot 系列 -[![AI 搭檔程式設計 Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![C#/.NET Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot 冒險之旅](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + +### Copilot 系列 +[![AI 協同編程 Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![C#/.NET Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot 冒險](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## 負責任的 AI -微軟致力於幫助客戶負責任地使用我們的 AI 產品，分享我們的經驗，並透過透明度說明與影響評估等工具建立基於信任的合作夥伴關係。許多相關資源可於 [https://aka.ms/RAI](https://aka.ms/RAI) 找到。 -微軟負責任 AI 的方法基於我們的 AI 原則：公平性、可靠性與安全性、隱私與安全性、包容性、透明性以及問責制。 +Microsoft 致力於幫助客戶負責任地使用我們的 AI 產品，分享我們的經驗，並通過透明度說明和影響評估等工具建立基於信任的合作關係。許多相關資源可於 [https://aka.ms/RAI](https://aka.ms/RAI) 獲得。 +Microsoft 對負責任 AI 的方法建立於我們的 AI 原則：公平性、可靠性與安全性、隱私與安全性、包容性、透明度和問責制。 + +大型自然語言、影像和語音模型——如本範例中使用的模型——可能存在不公平、不可靠或冒犯性的行為，並可能帶來傷害。請查閱 [Azure OpenAI 服務透明度說明](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) 以瞭解風險與限制。 + -大型自然語言、影像和語音模型——如本範例中所用，可能會表現出不公平、不可靠或冒犯性的行為，從而導致傷害。請參閱 [Azure OpenAI 服務透明度說明](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) 以瞭解風險與限制。 -減輕這些風險的建議方法是在您的架構中包含一個安全系統，能夠檢測和防止有害行為。[Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) 提供獨立的保護層，能夠偵測應用程式和服務中用戶生成及 AI 生成的有害內容。Azure AI Content Safety 包含文字與圖片的 API，讓您能偵測有害的素材。在 Microsoft Foundry 中，Content Safety 服務允許您查看、探索並嘗試偵測不同模態有害內容的範例程式碼。以下的[快速入門文件](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) 導引您如何向服務發出請求。 +減輕這些風險的建議方法是在您的架構中包含一個安全系統，能夠檢測並防止有害行為。 [Azure AI Content Safety](https://learn.microsoft.com/azure/ai-services/content-safety/overview) 提供獨立的保護層，能夠檢測應用程式和服務中的用戶生成及 AI 生成的有害內容。Azure AI Content Safety 包含文字和圖片 API，讓您能夠檢測有害材料。在 Microsoft Foundry 中，Content Safety 服務允許您查看、探索並試用不同模態下檢測有害內容的範例代碼。以下的 [快速入門文件](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) 指導您如何向該服務發出請求。 -另一個需要考慮的方面是整體應用程式的效能。對於多模態及多模型應用，我們將效能視為系統符合您與使用者的預期，包括不產生有害輸出。評估整體應用程式的效能很重要，建議使用[效能與品質及風險與安全評估器](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)。您也可以使用[自訂評估器](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) 來建立和評估。 +另一個需要考慮的方面是整體應用程式效能。對於多模態和多模型應用程式，我們認為效能是指系統如您和您的使用者所期望地執行，包括不產生有害輸出。評估整體應用程式效能時，使用 [效能和品質以及風險和安全評估器](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in) 非常重要。您也可以使用 [自訂評估器](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) 來建立和評估。 -您可以使用[Azure AI 評估 SDK](https://microsoft.github.io/promptflow/index.html) 在開發環境中評估您的 AI 應用程式。只要有測試數據集或目標，您的生成式 AI 應用程式生成內容便會透過內建評估器或您選擇的自訂評估器進行定量測量。要開始使用 Azure AI 評估 SDK 評估系統，您可以參考[快速入門指南](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)。執行評估後，您還可以在[Microsoft Foundry 中視覺化結果](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)。 +您可以使用 [Azure AI 評估 SDK](https://microsoft.github.io/promptflow/index.html) 在開發環境中評估您的 AI 應用程式。無論是給定測試資料集或目標，您的生成型 AI 應用程式生成結果都會以內建評估器或您選擇的自訂評估器進行量化測量。想要開始使用 Azure AI 評估 SDK 來評估您的系統，您可以參考 [快速入門指南](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)。執行完一次評估後，您可以在 [Microsoft Foundry 中視覺化結果](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)。 ## 商標 -本專案可能包含用於專案、產品或服務的商標或標誌。授權使用 Microsoft 商標或標誌須符合並遵守[Microsoft 商標與品牌指導方針](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)。修改版本中對 Microsoft 商標或標誌的使用不得造成混淆或暗示 Microsoft 之贊助。任何第三方商標或標誌的使用均需遵守該第三方的政策。 +本專案可能包含專案、產品或服務的商標或標誌。微軟商標或標誌的授權使用須遵守且依據 [Microsoft 的商標與品牌指南](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)。 +在此專案的修改版本中使用微軟商標或標誌時，必須避免造成混淆或暗示微軟贊助。任何第三方商標或標誌的使用均須遵循該第三方的政策。 ## 尋求協助 -如果您遇到困難或對建立 AI 應用有任何疑問，請加入： +如果您遇到困難或有關建構 AI 應用程式的問題，請加入： [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -如果您在開發過程中有產品回饋或錯誤，請訪問： +若您在建構過程中有產品反饋或遇到錯誤，請造訪： [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**免責聲明**： -本文件使用 AI 翻譯服務 [Co-op Translator](https://github.com/Azure/co-op-translator) 進行翻譯。雖然我們致力於確保準確性，但請注意自動翻譯可能包含錯誤或不準確之處。原始文件的母語版本應被視為權威來源。對於重要資訊，建議採用專業人工翻譯。本公司不對因使用本翻譯而導致的任何誤解或誤譯承擔責任。 +**免責聲明**： +本文件由 AI 翻譯服務 [Co-op Translator](https://github.com/Azure/co-op-translator) 翻譯而成。雖然我們致力於確保準確性，但請注意，機器自動翻譯可能包含錯誤或不準確之處。原始文件的母語版本應被視為權威來源。對於重要資訊，建議進行專業人工翻譯。我們不對因使用本翻譯而產生的任何誤解或誤釋承擔責任。 \ No newline at end of file diff --git a/translations/zh-HK/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/zh-HK/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..def4716fa --- /dev/null +++ b/translations/zh-HK/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 結合 Microsoft phi-4-mini 與 SQLite FTS5 的混合 RAG （零雲端 SLM）\n", + "\n", + "> **作者：** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **專注：** 小型語言模型 (SLMs)、SQLite FTS5 BM25、稠密嵌入、互惠排序融合 (RRF)\n", + "\n", + "---\n", + "\n", + "## 📌 1. 動機：本地 SLM 中的關鍵字召回困境\n", + "純粹依賴稠密向量嵌入的標準 RAG 架構，經常無法檢索到精確的數值標記（例如 `2,340,000 TL`、合約代碼、帳號）。 \n", + "相反地，稀疏詞彙搜尋（BM25）則會錯過語意同義詞與意譯過的問題。\n", + "\n", + "本手冊示範如何實作一個高速、記憶體內的混合檢索引擎，結合：\n", + "1. 稠密向量（餘弦相似度）\n", + "2. 稀疏詞彙搜尋（SQLite FTS5 BM25）\n", + "3. **互惠排序融合（RRF，$k=60$）**\n", + "4. 以 Microsoft `phi-4-mini` 進行有據引用生成（`[1]`、`[2]`）\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. 雙 SQLite 架構（密集向量 + 虛擬 FTS5 BM25 表）\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. 樣本攝取及執行基準測試\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. Microsoft phi-4-mini 的有根據提示形成\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**免責聲明**：\n本文件由 AI 翻譯服務 [Co-op Translator](https://github.com/Azure/co-op-translator) 翻譯而成。雖然我們致力於確保準確性，但請注意，機器自動翻譯可能包含錯誤或不準確之處。原始文件的母語版本應被視為權威來源。對於重要資訊，建議進行專業人工翻譯。我們不對因使用本翻譯而產生的任何誤解或誤釋承擔責任。\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/zh-MO/.co-op-translator.json b/translations/zh-MO/.co-op-translator.json index cc80838c6..e1cd48cff 100644 --- a/translations/zh-MO/.co-op-translator.json +++ b/translations/zh-MO/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "zh-MO" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:31:36+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T16:18:49+00:00", "source_file": "README.md", "language_code": "zh-MO" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "zh-MO" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T16:27:03+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "zh-MO" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:29:52+00:00", diff --git a/translations/zh-MO/README.md b/translations/zh-MO/README.md index 2e44c8910..d1f983a36 100644 --- a/translations/zh-MO/README.md +++ b/translations/zh-MO/README.md @@ -8,35 +8,35 @@ [![GitHub 拉取請求](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) [![歡迎 PR](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub 觀察者](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) -[![GitHub 派生](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 監看者](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) +[![GitHub 分支](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub 星標](https://img.shields.io/github/stars/microsoft/phicookbook?style=social&label=Star)](https://GitHub.com/microsoft/phicookbook/stargazers/?WT.mc_id=aiml-137032-kinfeylo) [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi 是由 Microsoft 開發的一系列開源人工智能模型。 +Phi 是 Microsoft 開發的一系列開源 AI 模型。 -Phi 目前是功能最強大且最具成本效益的小型語言模型（SLM），在多語言、推理、文字／聊天生成、編碼、影像、音訊及其他場景中均有非常優異的基準表現。 +Phi 目前是最強大且具成本效益的小型語言模型（SLM），在多語言、推理、文本/聊天生成、編碼、圖像、音頻及其他場景中都擁有很好的基準表現。 -您可以將 Phi 部署到雲端或邊緣裝置，並能夠輕鬆使用有限的計算能力建構生成式 AI 應用。 +您可以將 Phi 部署到雲端或邊緣裝置，並能輕鬆建立計算能力有限下的生成式 AI 應用程式。 -請依照以下步驟開始使用這些資源： -1. 派生此倉庫：點擊 [![GitHub 派生](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. 克隆此倉庫：`git clone https://github.com/microsoft/PhiCookBook.git` -3. [**加入 Microsoft AI Discord 社群，與專家及開發者交流**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +按照以下步驟開始使用這些資源： +1. 派生儲存庫：點擊 [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. 複製儲存庫：`git clone https://github.com/microsoft/PhiCookBook.git` +3. [**加入 Microsoft AI Discord 社群，認識專家與其他開發者**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/zh-MO/cover.eb18d1b9605d754b.webp) ### 🌐 多語言支援 -#### 透過 GitHub Action 支援（自動且持續更新） +#### 透過 GitHub Action 支援（自動化且持續更新） -[阿拉伯文](../ar/README.md) | [孟加拉文](../bn/README.md) | [保加利亞語](../bg/README.md) | [緬甸語 (緬甸)](../my/README.md) | [中文（簡體）](../zh-CN/README.md) | [中文（繁體，香港）](../zh-HK/README.md) | [中文（繁體，澳門）](./README.md) | [中文（繁體，臺灣）](../zh-TW/README.md) | [克羅地亞語](../hr/README.md) | [捷克語](../cs/README.md) | [丹麥語](../da/README.md) | [荷蘭語](../nl/README.md) | [愛沙尼亞語](../et/README.md) | [芬蘭語](../fi/README.md) | [法語](../fr/README.md) | [德語](../de/README.md) | [希臘語](../el/README.md) | [希伯來語](../he/README.md) | [印地語](../hi/README.md) | [匈牙利語](../hu/README.md) | [印尼語](../id/README.md) | [義大利語](../it/README.md) | [日語](../ja/README.md) | [坎納達語](../kn/README.md) | [高棉語](../km/README.md) | [韓語](../ko/README.md) | [立陶宛語](../lt/README.md) | [馬來語](../ms/README.md) | [馬拉雅拉姆語](../ml/README.md) | [馬拉地語](../mr/README.md) | [尼泊爾語](../ne/README.md) | [奈及利亞洋蔥語](../pcm/README.md) | [挪威語](../no/README.md) | [波斯語 (法爾西語)](../fa/README.md) | [波蘭語](../pl/README.md) | [葡萄牙語 (巴西)](../pt-BR/README.md) | [葡萄牙語 (葡萄牙)](../pt-PT/README.md) | [旁遮普語 (谷魯穆奇)](../pa/README.md) | [羅馬尼亞語](../ro/README.md) | [俄語](../ru/README.md) | [塞爾維亞語 (西里爾字母)](../sr/README.md) | [斯洛伐克語](../sk/README.md) | [斯洛文尼亞語](../sl/README.md) | [西班牙語](../es/README.md) | [斯瓦希里語](../sw/README.md) | [瑞典語](../sv/README.md) | [塔加洛語 (菲律賓語)](../tl/README.md) | [泰米爾語](../ta/README.md) | [泰盧固語](../te/README.md) | [泰語](../th/README.md) | [土耳其語](../tr/README.md) | [烏克蘭語](../uk/README.md) | [烏爾都語](../ur/README.md) | [越南語](../vi/README.md) +[Arabic](../ar/README.md) | [Bengali](../bn/README.md) | [Bulgarian](../bg/README.md) | [Burmese (Myanmar)](../my/README.md) | [Chinese (Simplified)](../zh-CN/README.md) | [Chinese (Traditional, Hong Kong)](../zh-HK/README.md) | [Chinese (Traditional, Macau)](./README.md) | [Chinese (Traditional, Taiwan)](../zh-TW/README.md) | [Croatian](../hr/README.md) | [Czech](../cs/README.md) | [Danish](../da/README.md) | [Dutch](../nl/README.md) | [Estonian](../et/README.md) | [Finnish](../fi/README.md) | [French](../fr/README.md) | [German](../de/README.md) | [Greek](../el/README.md) | [Hebrew](../he/README.md) | [Hindi](../hi/README.md) | [Hungarian](../hu/README.md) | [Indonesian](../id/README.md) | [Italian](../it/README.md) | [Japanese](../ja/README.md) | [Kannada](../kn/README.md) | [Khmer](../km/README.md) | [Korean](../ko/README.md) | [Lithuanian](../lt/README.md) | [Malay](../ms/README.md) | [Malayalam](../ml/README.md) | [Marathi](../mr/README.md) | [Nepali](../ne/README.md) | [Nigerian Pidgin](../pcm/README.md) | [Norwegian](../no/README.md) | [Persian (Farsi)](../fa/README.md) | [Polish](../pl/README.md) | [Portuguese (Brazil)](../pt-BR/README.md) | [Portuguese (Portugal)](../pt-PT/README.md) | [Punjabi (Gurmukhi)](../pa/README.md) | [Romanian](../ro/README.md) | [Russian](../ru/README.md) | [Serbian (Cyrillic)](../sr/README.md) | [Slovak](../sk/README.md) | [Slovenian](../sl/README.md) | [Spanish](../es/README.md) | [Swahili](../sw/README.md) | [Swedish](../sv/README.md) | [Tagalog (Filipino)](../tl/README.md) | [Tamil](../ta/README.md) | [Telugu](../te/README.md) | [Thai](../th/README.md) | [Turkish](../tr/README.md) | [Ukrainian](../uk/README.md) | [Urdu](../ur/README.md) | [Vietnamese](../vi/README.md) -> **偏好本地克隆嗎？** +> **偏好本地複製？** > -> 本倉庫包含超過 50 種語言的翻譯，會大幅增加下載大小。若不包含翻譯克隆，請使用稀疏檢出功能： +> 本儲存庫包含 50 多種語言的翻譯，會大幅增加下載大小。若想複製但不包含翻譯，請使用稀疏檢出： > > **Bash / macOS / Linux:** > ```bash @@ -45,280 +45,285 @@ Phi 目前是功能最強大且最具成本效益的小型語言模型（SLM） > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` > -> **CMD (Windows):** +> **CMD（Windows）：** > ```cmd > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> 這樣您就會取得完成課程所需的所有內容，且下載速度更快。 +> 這會讓您以更快的速度下載完成課程所需的一切。 ## 目錄 - 介紹 - [歡迎加入 Phi 家族](./md/01.Introduction/01/01.PhiFamily.md) - - [設置您的開發環境](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [設定您的環境](./md/01.Introduction/01/01.EnvironmentSetup.md) - [理解關鍵技術](./md/01.Introduction/01/01.Understandingtech.md) - - [Phi 模型的 AI 安全](./md/01.Introduction/01/01.AISafety.md) + - [Phi 模型的 AI 安全性](./md/01.Introduction/01/01.AISafety.md) - [Phi 硬體支援](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi 模型及跨平台可用性](./md/01.Introduction/01/01.Edgeandcloud.md) - - [使用 Guidance-ai 和 Phi](./md/01.Introduction/01/01.Guidance.md) - - [GitHub 市場模型](https://github.com/marketplace/models) + - [Phi 模型與跨平台可用性](./md/01.Introduction/01/01.Edgeandcloud.md) + - [使用 Guidance-ai 與 Phi](./md/01.Introduction/01/01.Guidance.md) + - [GitHub 市集模型](https://github.com/marketplace/models) - [Azure AI 模型目錄](https://ai.azure.com) -- 在不同環境中推論 Phi +- 在不同環境中推理 Phi - [Hugging face](./md/01.Introduction/02/01.HF.md) - [GitHub 模型](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry 模型目錄](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - [AI 工具包 VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry 本地](./md/01.Introduction/02/07.FoundryLocal.md) - -- 推論 Phi 家族 - - [在 iOS 上推論 Phi](./md/01.Introduction/03/iOS_Inference.md) - - [在 Android 上推論 Phi](./md/01.Introduction/03/Android_Inference.md) - - [在 Jetson 上推論 Phi](./md/01.Introduction/03/Jetson_Inference.md) - - [在 AI PC 上推論 Phi](./md/01.Introduction/03/AIPC_Inference.md) - - [使用 Apple MLX 框架推論 Phi](./md/01.Introduction/03/MLX_Inference.md) - - [在本地伺服器推論 Phi](./md/01.Introduction/03/Local_Server_Inference.md) - - [使用 AI 工具包於遠端伺服器推論 Phi](./md/01.Introduction/03/Remote_Interence.md) - - [使用 Rust 推論 Phi](./md/01.Introduction/03/Rust_Inference.md) - - [本地推論 Phi–視覺](./md/01.Introduction/03/Vision_Inference.md) - - [使用 Kaito AKS 和 Azure Containers（官方支援）推論 Phi](./md/01.Introduction/03/Kaito_Inference.md) + - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) + +- Phi 家族推理 + - [iOS 推理 Phi](./md/01.Introduction/03/iOS_Inference.md) + - [Android 推理 Phi](./md/01.Introduction/03/Android_Inference.md) + - [Jetson 推理 Phi](./md/01.Introduction/03/Jetson_Inference.md) + - [AI PC 推理 Phi](./md/01.Introduction/03/AIPC_Inference.md) + - [使用 Apple MLX 框架推理 Phi](./md/01.Introduction/03/MLX_Inference.md) + - [本地伺服器推理 Phi](./md/01.Introduction/03/Local_Server_Inference.md) + - [使用 AI 工具包在遠端伺服器推理 Phi](./md/01.Introduction/03/Remote_Interence.md) + - [使用 Rust 推理 Phi](./md/01.Introduction/03/Rust_Inference.md) + - [本地推理 Phi--影像](./md/01.Introduction/03/Vision_Inference.md) + - [使用 Kaito AKS、Azure 容器推理 Phi（官方支援）](./md/01.Introduction/03/Kaito_Inference.md) - [Phi 家族量化](./md/01.Introduction/04/QuantifyingPhi.md) - [使用 llama.cpp 量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [使用 onnxruntime 的生成 AI 擴充功能量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [使用 onnxruntime 生成式 AI 擴充套件量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [使用 Intel OpenVINO 量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - [使用 Apple MLX 框架量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) - Phi 評估 - [負責任的 AI](./md/01.Introduction/05/ResponsibleAI.md) - - [使用 Microsoft Foundry 進行評估](./md/01.Introduction/05/AIFoundry.md) + - [Microsoft Foundry 評估](./md/01.Introduction/05/AIFoundry.md) - [使用 Promptflow 進行評估](./md/01.Introduction/05/Promptflow.md) + +- 與 Azure AI 搜尋的 RAG + - [如何使用 Phi-4-mini 與 Phi-4-multimodal (RAG) 搭配 Azure AI 搜尋](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [零雲端本地混合 RAG，使用 SQLite FTS5 與 phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) -- 結合 Azure AI 搜尋的 RAG - - [如何使用 Phi-4-mini 和 Phi-4-multimodal (RAG) 搭配 Azure AI 搜尋](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- Phi 應用開發範例 + - 文字與聊天應用程式 + - Phi-4 範例 + - [📓] [與 Phi-4-mini ONNX 模型聊天](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) + - [使用 Phi-4 本地 ONNX 模型聊天 .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [使用 Sementic Kernel 的 Phi-4 ONNX .NET 控制台應用程式聊天](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) -- Phi 應用程式開發範例 - - 文字與聊天應用 - - Phi-4 範例 - - [📓] [使用 Phi-4-mini ONNX 模型聊天](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [使用 Phi-4 本地 ONNX .NET 模型聊天](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [使用 Sementic Kernel 的 Phi-4 ONNX .NET 主控台應用聊天](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) - Phi-3 / 3.5 範例 - - [使用 Phi3、ONNX Runtime Web 與 WebGPU 的本地瀏覽器聊天機器人](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [在瀏覽器中使用 Phi3、ONNX Runtime Web 和 WebGPU 的本地聊天機械人](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino 聊天](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [多模型 - 互動式 Phi-3-mini 與 OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - [MLFlow - 建立封裝並使用 Phi-3 與 MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [模型優化 - 如何為 ONNX Runtime Web 使用 Olive 優化 Phi-3-min 模型](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 應用程式搭配 Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 多模型 AI 功能筆記應用程式範例](https://github.com/microsoft/ai-powered-notes-winui3-sample) - - [微調並整合自訂 Phi-3 模型與 Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - - [在 Microsoft Foundry 中使用 Prompt flow 微調並整合自訂 Phi-3 模型](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [評估在 Microsoft Foundry 中微調的 Phi-3 / Phi-3.5 模型，聚焦於微軟負責任的 AI 原則](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [模型優化 - 如何使用 Olive 優化 Phi-3-min 模型以用於 ONNX Runtime Web](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [用 Phi-3 mini-4k-instruct-onnx 製作的 WinUI3 應用程式](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 多模型 AI 助力筆記應用範例](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [微調及整合定制 Phi-3 模型與 Prompt flow](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) + - [在 Microsoft Foundry 使用 Prompt flow 微調及整合定制 Phi-3 模型](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) + - [在 Microsoft Foundry 評估微調後的 Phi-3 / Phi-3.5 模型，重點為 Microsoft 的負責任 AI 原則](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - [📓] [Phi-3.5-mini-instruct 語言預測範例（中文/英文）](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - - [Phi-3.5-Instruct WebGPU RAG 聊天機器人](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [使用 Windows GPU 與 Phi-3.5-Instruct ONNX 建立 Prompt flow 解決方案](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [使用微軟 Phi-3.5 tflite 建立 Android 應用程式](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [使用本地 ONNX Phi-3 模型和 Microsoft.ML.OnnxRuntime 的 Q&A .NET 範例](../../md/04.HOL/dotnet/src/LabsPhi301) - - [使用 Semantic Kernel 與 Phi-3 的主控台聊天 .NET 應用程式](../../md/04.HOL/dotnet/src/LabsPhi302) + - [Phi-3.5-Instruct WebGPU RAG 聊天機械人](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) + - [使用 Windows GPU 建構帶有 Phi-3.5-Instruct ONNX 的 Prompt flow 解決方案](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [使用 Microsoft Phi-3.5 tflite 建立 Android 應用](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [利用本地 ONNX Phi-3 模型的 .NET 問答範例，使用 Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) + - [用 Semantic Kernel 與 Phi-3 製作的 .NET 控制台聊天應用](../../md/04.HOL/dotnet/src/LabsPhi302) - - Azure AI 推理 SDK 代碼範例 + - Azure AI 推論 SDK 程式碼範例 - Phi-4 範例 - - [📓] [使用 Phi-4-multimodal 生成專案程式碼](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [📓] [使用 Phi-4-multimodal 產生專案程式碼](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 範例 - - [使用微軟 Phi-3 家族建立您自己的 Visual Studio Code GitHub Copilot 聊天](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [使用 GitHub 模型搭配 Phi-3.5 建立您自己的 Visual Studio Code 聊天 Copilot 代理](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [用 Microsoft Phi-3 系列建立自己的 Visual Studio Code GitHub Copilot 聊天](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [用 GitHub 模型與 Phi-3.5 創建你自己的 Visual Studio Code 聊天 Copilot 代理](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - 進階推理範例 - Phi-4 範例 - - [📓] [Phi-4-mini-reasoning 或 Phi-4-reasoning 範例](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [使用微軟 Olive 微調 Phi-4-mini-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [使用 Apple MLX 微調 Phi-4-mini-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [使用 GitHub 模型的 Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [使用微軟 Foundry 模型的 Phi-4-mini-reasoning](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - 演示 - - [托管於 Hugging Face Spaces 的 Phi-4-mini 演示](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [托管於 Hugginge Face Spaces 的 Phi-4-multimodal 演示](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [📓] [Phi-4-mini 推理或 Phi-4 推理範例](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [用 Microsoft Olive 微調 Phi-4-mini 推理](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [用 Apple MLX 微調 Phi-4-mini 推理](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [用 GitHub 模型推理 Phi-4-mini](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [用 Microsoft Foundry 模型推理 Phi-4-mini](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - 示範 + - [Phi-4-mini 示範托管於 Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Phi-4-multimodal 示範托管於 Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - 視覺範例 - Phi-4 範例 - - [📓] [使用 Phi-4-multimodal 讀取圖片並生成程式碼](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [📓] [使用 Phi-4-multimodal 閱讀圖像與生成程式碼](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 範例 - - [📓][Phi-3-vision-圖片文本轉文本](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-圖像文本轉文字](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [📓][Phi-3-vision CLIP 嵌入](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [演示: Phi-3 回收](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - 視覺語言助理 - 使用 Phi3-Vision 和 OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [示範：Phi-3 回收](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - 視覺語言助手 - 使用 Phi3-Vision 與 OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - [Phi-3 視覺 Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - [Phi-3 視覺 OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - [📓][Phi-3.5 視覺多幀或多圖像範例](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - [Phi-3 視覺本地 ONNX 模型，使用 Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [選單式 Phi-3 視覺本地 ONNX 模型，使用 Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [基於選單的 Phi-3 視覺本地 ONNX 模型，使用 Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) - 推理-視覺範例 - Phi-4-Reasoning-Vision-15B - [📓] [使用 Phi-4-Reasoning-Vision-15B 偵測闖紅燈](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [使用 Phi-4-Reasoning-Vision-15B 解數學題](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [使用 Phi-4-Reasoning-Vision-15B 偵測使用者界面](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [使用 Phi-4-Reasoning-Vision-15B 進行數學運算](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [使用 Phi-4-Reasoning-Vision-15B 偵測使用者介面](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - 數學範例 - - Phi-4-Mini-Flash-Reasoning-Instruct 範例 [使用 Phi-4-Mini-Flash-Reasoning-Instruct 的數學演示](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-Mini-Flash-Reasoning-Instruct 範例 [Phi-4-Mini-Flash-Reasoning-Instruct 的數學示範](./md/02.Application/09.Math/MathDemo.ipynb) - 音訊範例 - Phi-4 範例 - - [📓] [使用 Phi-4-multimodal 擷取語音文字轉錄](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal 音頻範例](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [使用 Phi-4-multimodal 提取音訊文字記錄](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal 音訊範例](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Phi-4-multimodal 語音翻譯範例](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET 主控台應用程式，使用 Phi-4-multimodal 分析音訊檔案並生成文字轉錄](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [.NET 控制台應用程序使用 Phi-4-multimodal 音訊分析音訊檔並生成文字記錄](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - MOE 範例 - Phi-3 / 3.5 範例 - - [📓] [Phi-3.5 專家混合模型 (MoEs) 社交媒體範例](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [使用 NVIDIA NIM Phi-3 MOE、Azure AI 搜尋及 LlamaIndex 建立檢索增強生成 (RAG) 管線](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 混合專家模型 (MoEs) 社交媒體範例](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [建立一個使用 NVIDIA NIM Phi-3 MOE、Azure AI 搜尋與 LlamaIndex 的檢索增強生成 (RAG) 管線](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - 函式調用範例 + - 函數調用範例 - Phi-4 範例 🆕 - - [📓] [使用 Phi-4-mini 的函式調用](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [使用函式調用建立多代理，搭配 Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [搭配 Ollama 的函式調用](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [搭配 ONNX 的函式調用](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - [📓] [使用 Phi-4-mini 進行函數調用](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [使用函數調用與 Phi-4-mini 創建多代理](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [使用函數調用與 Ollama](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [使用函數調用與 ONNX](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - 多模態混合範例 - Phi-4 範例 🆕 - [📓] [使用 Phi-4-multimodal 作為科技記者](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET 主控台應用程式，使用 Phi-4-multimodal 分析圖片](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [.NET 控制台應用程式使用 Phi-4-multimodal 分析圖像](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - Phi 微調範例 - - [微調場景](./md/03.FineTuning/FineTuning_Scenarios.md) - - [微調與 RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [微調讓 Phi-3 成為產業專家](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [使用 AI Toolkit for VS Code 微調 Phi-3](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - - [使用 Azure 機器學習服務微調 Phi-3](./md/03.FineTuning/Introduce_AzureML.md) + - [微調情境](./md/03.FineTuning/FineTuning_Scenarios.md) + - [微調與 RAG 的比較](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [微調讓 Phi-3 成為業界專家](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [利用 AI 工具組於 VS Code 微調 Phi-3](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [利用 Azure 機器學習服務微調 Phi-3](./md/03.FineTuning/Introduce_AzureML.md) - [使用 Lora 微調 Phi-3](./md/03.FineTuning/FineTuning_Lora.md) - [使用 QLora 微調 Phi-3](./md/03.FineTuning/FineTuning_Qlora.md) - - [使用微軟 Foundry 微調 Phi-3](./md/03.FineTuning/FineTuning_AIFoundry.md) + - [利用 Microsoft Foundry 微調 Phi-3](./md/03.FineTuning/FineTuning_AIFoundry.md) - [使用 Azure ML CLI/SDK 微調 Phi-3](./md/03.FineTuning/FineTuning_MLSDK.md) - - [與微軟 Olive 共同微調](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - - [微軟 Olive 實務操作實驗室](./md/03.FineTuning/olive-lab/readme.md) + - [使用 Microsoft Olive 微調](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) + - [Microsoft Olive 微調實作實驗室](./md/03.FineTuning/olive-lab/readme.md) - [使用 Weights and Bias 微調 Phi-3-vision](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) + - [使用 Apple MLX 框架微調 Phi-3](./md/03.FineTuning/FineTuning_MLX.md) - [微調 Phi-3-vision（官方支援）](./md/03.FineTuning/FineTuning_Vision.md) - - [使用 Kaito AKS、Azure Containers（官方支援）微調 Phi-3](./md/03.FineTuning/FineTuning_Kaito.md) + - [使用 Kaito AKS 與 Azure 容器微調 Phi-3（官方支援）](./md/03.FineTuning/FineTuning_Kaito.md) - [微調 Phi-3 與 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) - 實作實驗室 - - [探索尖端模型：LLMs、SLMs、本地開發及更多](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [解鎖 NLP 潛力：使用 Microsoft Olive 進行微調](https://github.com/azure/Ignite_FineTuning_workshop) + - [探索尖端模型：LLM、SLM、本地開發及更多](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [解鎖 NLP 潛能：使用 Microsoft Olive 微調](https://github.com/azure/Ignite_FineTuning_workshop) -- 學術研究論文與出版品 - - [Textbooks Are All You Need II：phi-1.5 技術報告](https://arxiv.org/abs/2309.05463) - - [Phi-3 技術報告：在您的手機上運行的高度能力語言模型](https://arxiv.org/abs/2404.14219) +- 學術研究論文與出版物 + - [教材即萬能 II：phi-1.5 技術報告](https://arxiv.org/abs/2309.05463) + - [Phi-3 技術報告：一款功能強大的本機手機語言模型](https://arxiv.org/abs/2404.14219) - [Phi-4 技術報告](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini 技術報告：通過 LoRA 混合實現小巧但強大的多模態語言模型](https://arxiv.org/abs/2503.01743) - - [優化車載功能呼叫用的小型語言模型](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) 用於多選題問答的 PHI-3 微調：方法論、結果與挑戰](https://arxiv.org/abs/2501.01588) + - [Phi-4-Mini 技術報告：透過混合 LoRA 的緊湊而強大的多模態語言模型](https://arxiv.org/abs/2503.01743) + - [優化用於車載功能呼叫的小型語言模型](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) 微調 PHI-3 用於多項選擇題答題：方法、結果與挑戰](https://arxiv.org/abs/2501.01588) - [Phi-4 推理技術報告](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Phi-4-mini 推理技術報告](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) ## 使用 Phi 模型 -### Microsoft Foundry 上的 Phi +### 在 Microsoft Foundry 上使用 Phi -您可以學習如何使用 Microsoft Phi 以及如何在不同硬件裝置上構建端到端解決方案。想親自體驗 Phi，請先試玩模型並使用 [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai) 根據您的場景自訂 Phi。更多資訊請參考 [Microsoft Foundry 入門指南](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) +你可以學習如何使用 Microsoft Phi 以及如何在你的各類硬體設備上建立端到端解決方案。要親身體驗 Phi，先從試用模型並根據你的場景自訂 Phi 入手，使用 [Microsoft Foundry Azure AI 模型目錄](https://aka.ms/phi3-azure-ai)，更多資訊請參閱快速入門的 [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) -遊樂場 -每個模型都有專屬的遊樂場測試模型 [Azure AI Playground](https://aka.ms/try-phi3)。 +遊樂場 +每個模型都有專屬的遊樂場可供測試模型，[Azure AI 遊樂場](https://aka.ms/try-phi3)。 -### GitHub 模型上的 Phi +### GitHub 上的 Phi 模型 -您可以學習如何使用 Microsoft Phi 以及如何在不同硬件裝置上構建端到端解決方案。想親自體驗 Phi，請先試玩模型並使用 [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) 根據您的場景自訂 Phi。更多資訊請參考 [GitHub Model Catalog 入門指南](/md/02.QuickStart/GitHubModel_QuickStart.md) +你可以學習如何使用 Microsoft Phi 以及如何在你的各類硬體設備上建立端到端解決方案。要親身體驗 Phi，先從試用模型並根據你的場景自訂 Phi 入手，使用 [GitHub 模型目錄](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo)，更多資訊請參閱[GitHub 模型目錄快速入門](/md/02.QuickStart/GitHubModel_QuickStart.md) -遊樂場 -每個模型皆有專屬的[遊樂場以測試模型](/md/02.QuickStart/GitHubModel_QuickStart.md)。 +遊樂場 +每個模型都有專屬的[遊樂場用於測試模型](/md/02.QuickStart/GitHubModel_QuickStart.md)。 ### Hugging Face 上的 Phi -您亦可在 [Hugging Face](https://huggingface.co/microsoft) 平台找到此模型。 +你亦可在 [Hugging Face](https://huggingface.co/microsoft) 上找到模型 -遊樂場 -[Hugging Chat 遊樂場](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) +遊樂場 + [Hugging Chat 遊樂場](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) -## 🎒 其他課程 + ## 🎒 其他課程 -我們的團隊亦製作了其他課程！歡迎查看： +我們的團隊還提供其他課程！歡迎查看： ### LangChain -[![LangChain4j for Beginners](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) -[![LangChain.js for Beginners](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) -[![LangChain for Beginners](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain4j 初學者](https://img.shields.io/badge/LangChain4j%20for%20Beginners-22C55E?style=for-the-badge&&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchain4j-for-beginners) +[![LangChain.js 初學者](https://img.shields.io/badge/LangChain.js%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://aka.ms/langchainjs-for-beginners?WT.mc_id=m365-94501-dwahlin) +[![LangChain 初學者](https://img.shields.io/badge/LangChain%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=0553D6)](https://github.com/microsoft/langchain-for-beginners?WT.mc_id=m365-94501-dwahlin) --- -### Azure / Edge / MCP / Agents -[![AZD for Beginners](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Edge AI for Beginners](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![MCP for Beginners](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) -[![AI Agents for Beginners](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) +### Azure / Edge / MCP / Agents +[![AZD 初學者](https://img.shields.io/badge/AZD%20for%20Beginners-0078D4?style=for-the-badge&labelColor=E5E7EB&color=0078D4)](https://github.com/microsoft/AZD-for-beginners?WT.mc_id=academic-105485-koreyst) +[![Edge AI 初學者](https://img.shields.io/badge/Edge%20AI%20for%20Beginners-00B8E4?style=for-the-badge&labelColor=E5E7EB&color=00B8E4)](https://github.com/microsoft/edgeai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![MCP 初學者](https://img.shields.io/badge/MCP%20for%20Beginners-009688?style=for-the-badge&labelColor=E5E7EB&color=009688)](https://github.com/microsoft/mcp-for-beginners?WT.mc_id=academic-105485-koreyst) +[![AI Agents 初學者](https://img.shields.io/badge/AI%20Agents%20for%20Beginners-00C49A?style=for-the-badge&labelColor=E5E7EB&color=00C49A)](https://github.com/microsoft/ai-agents-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### 生成式 AI 系列 -[![Generative AI for Beginners](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) -[![Generative AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) -[![Generative AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) -[![Generative AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) + +### 生成式 AI 系列 +[![生成式 AI 初學者](https://img.shields.io/badge/Generative%20AI%20for%20Beginners-8B5CF6?style=for-the-badge&labelColor=E5E7EB&color=8B5CF6)](https://github.com/microsoft/generative-ai-for-beginners?WT.mc_id=academic-105485-koreyst) +[![生成式 AI (.NET)](https://img.shields.io/badge/Generative%20AI%20(.NET)-9333EA?style=for-the-badge&labelColor=E5E7EB&color=9333EA)](https://github.com/microsoft/Generative-AI-for-beginners-dotnet?WT.mc_id=academic-105485-koreyst) +[![生成式 AI (Java)](https://img.shields.io/badge/Generative%20AI%20(Java)-C084FC?style=for-the-badge&labelColor=E5E7EB&color=C084FC)](https://github.com/microsoft/generative-ai-for-beginners-java?WT.mc_id=academic-105485-koreyst) +[![生成式 AI (JavaScript)](https://img.shields.io/badge/Generative%20AI%20(JavaScript)-E879F9?style=for-the-badge&labelColor=E5E7EB&color=E879F9)](https://github.com/microsoft/generative-ai-with-javascript?WT.mc_id=academic-105485-koreyst) --- - -### 核心學習 -[![ML for Beginners](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) -[![Data Science for Beginners](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) -[![AI for Beginners](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) -[![Cybersecurity for Beginners](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![Web Dev for Beginners](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) -[![IoT for Beginners](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) -[![XR Development for Beginners](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) + +### 核心學習 +[![機器學習初學者](https://img.shields.io/badge/ML%20for%20Beginners-22C55E?style=for-the-badge&labelColor=E5E7EB&color=22C55E)](https://aka.ms/ml-beginners?WT.mc_id=academic-105485-koreyst) +[![數據科學初學者](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) +[![人工智能初學者](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) +[![網絡安全初學者](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) +[![網頁開發初學者](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![物聯網初學者](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) +[![XR 開發初學者](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- - -### Copilot 系列 -[![Copilot for AI Paired Programming](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![Copilot for C#/.NET](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) -[![Copilot Adventure](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) + +### Copilot 系列 +[![AI 配對程式設計的 Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![C#/.NET Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![Copilot 冒險](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## 負責任的 AI -Microsoft 致力於幫助客戶負責任地使用我們的 AI 產品，分享我們的學習成果，並通過透明備註和影響評估等工具建立基於信任的合作夥伴關係。許多這些資源可以在 [https://aka.ms/RAI](https://aka.ms/RAI) 找到。 -Microsoft 的負責任 AI 方法基於我們的 AI 原則：公平性、可靠性與安全性、隱私與安全、包容性、透明度及問責制。 +微軟致力協助客戶負責任地使用我們的 AI 產品，透過透明度說明與影響評估等工具分享我們的學習，並建立基於信任的合作夥伴關係。許多相關資源可在 [https://aka.ms/RAI](https://aka.ms/RAI) 找到。 +微軟的負責任 AI 方法根植於我們的 AI 原則：公平、可靠與安全、隱私與安全性、包容性、透明度與問責制。 + +大型的自然語言、影像與語音模型 — 如本範例中使用的 — 可能會出現不公平、不可靠或冒犯性的行為，進而造成傷害。請參閱 [Azure OpenAI 服務透明度說明](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) 以了解風險與限制。 + -大規模的自然語言、影像和語音模型——例如本範例中使用的模型——有可能表現出不公平、不可靠或冒犯性的行為，繼而造成傷害。請參閱 [Azure OpenAI 服務透明備註](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) 以了解相關風險與限制。 -推薦的減輕這些風險的方法是在您的架構中包含一個安全系統，以檢測和防止有害行為。[Azure AI 內容安全](https://learn.microsoft.com/azure/ai-services/content-safety/overview) 提供一個獨立的保護層，能夠在應用程式和服務中檢測用戶生成和 AI 生成的有害內容。Azure AI 內容安全包含文字和圖片 API，讓您能夠檢測有害的材料。在 Microsoft Foundry 中，內容安全服務允許您查看、探索並嘗試跨不同模態檢測有害內容的範例程式碼。以下的[快速入門文件](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) 將指引您如何向該服務發出請求。 +減輕這些風險的建議方法是在您的架構中包含一個安全系統，能夠檢測並防止有害行為。[Azure AI 內容安全](https://learn.microsoft.com/azure/ai-services/content-safety/overview) 提供獨立的保護層，能夠檢測應用程式和服務中的用戶生成及 AI 生成的有害內容。Azure AI 內容安全包括文本和圖像 API，允許您檢測有害資料。在 Microsoft Foundry 中，內容安全服務允許您檢視、探索並嘗試用於檢測不同模態有害內容的範例程式碼。以下[入門文件](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) 導引您如何對該服務發出請求。 -另一個需要考慮的方面是整體應用程式的效能。對於多模態和多模型應用程式，我們認為效能意指系統的表現符合您和您的用戶的期望，包括不產生有害輸出。評估整體應用程式的效能非常重要，您可以使用[效能和品質以及風險和安全評估器](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)來完成評估。您也可以使用[自訂評估器](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators)來建立和評估。 +另一個需要考量的方面是整體應用程式的效能。在多模態和多模型的應用程式中，我們認為效能是系統能如您與使用者預期地運作，包括不產生有害輸出。評估您的整體應用程式效能很重要，您可使用[效能與品質以及風險與安全評估機制](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)。您亦有能力建立並使用[自訂評估機制](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators) 進行評估。 -您可以在開發環境中使用[Azure AI 評估 SDK](https://microsoft.github.io/promptflow/index.html)評估您的 AI 應用程式。透過測試資料集或目標，系統可以使用內建或您選擇的自訂評估器對生成的 AI 輸出進行量化測量。要開始使用 Azure AI 評估 SDK 評估系統，您可以參考[快速入門指南](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)。執行評估後，您可以在[Microsoft Foundry 中視覺化結果](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)。 +您可以在開發環境中使用[Azure AI 評估 SDK](https://microsoft.github.io/promptflow/index.html)來評估您的 AI 應用。針對測試資料集或目標，您的生成式 AI 應用程式的產出可透過內建或您選擇的自訂評估者進行量化測量。若要開始使用 azure ai 評估 sdk 評估您的系統，可以參考[快速入門指南](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)。完成評估執行後，您可在[Microsoft Foundry 中視覺化結果](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)。 ## 商標 -此專案可能包含專案、產品或服務的商標或標誌。授權使用 Microsoft 商標或標誌需遵守並遵循[Microsoft 商標及品牌指引](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)。 -在修改版本中使用 Microsoft 商標或標誌不得造成混淆或暗示 Microsoft 的贊助。任何第三方商標或標誌的使用皆須遵守該第三方的政策。 +本專案可能包含專案、產品或服務的商標或標誌。授權使用 Microsoft 商標或標誌，必須遵守並跟隨[Microsoft 商標與品牌指南](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)。 +在此專案修改版本中使用 Microsoft 商標或標誌，必須不得導致混淆或暗示 Microsoft 贊助。任何第三方商標或標誌的使用均須遵守該第三方的政策。 ## 尋求協助 -如果您遇到困難或對建立 AI 應用程式有任何疑問，請加入： +若您遇到問題或對建立 AI 應用有任何問題，歡迎加入： [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -如果您在開發過程中有產品反饋或錯誤，請造訪： +若您在建立過程中有產品反饋或錯誤，請造訪： [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**免責聲明**： -本文件已使用 AI 翻譯服務 [Co-op Translator](https://github.com/Azure/co-op-translator) 進行翻譯。雖然我們致力於提高準確性，但請注意，自動翻譯可能包含錯誤或不準確之處。原始文件的母語版本應視為權威來源。對於關鍵資訊，建議採用專業人工翻譯。我們不對因使用本翻譯而產生的任何誤解或誤譯承擔責任。 +**免責聲明**： +本文件使用 AI 翻譯服務 [Co-op Translator](https://github.com/Azure/co-op-translator) 進行翻譯。雖然我們力求準確，但請注意，自動翻譯可能包含錯誤或不準確之處。原始文件的母語版本應被視為權威來源。對於重要資訊，建議尋求專業人工翻譯。我們不對因使用本翻譯而引起的任何誤解或曲解承擔責任。 \ No newline at end of file diff --git a/translations/zh-MO/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/zh-MO/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..cc4fd5437 --- /dev/null +++ b/translations/zh-MO/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 混合式 RAG 結合 Microsoft phi-4-mini 及 SQLite FTS5（零雲端 SLM）\n", + "\n", + "> **作者：** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **焦點：** 小型語言模型（SLM）、SQLite FTS5 BM25、密集嵌入、相互排序融合（RRF）\n", + "\n", + "---\n", + "\n", + "## 📌 1. 動機：本地 SLM 中的關鍵字召回兩難\n", + "僅依賴密集向量嵌入的標準 RAG 架構常常無法準確檢索數值型標記（例如 `2,340,000 TL`、合約編號、帳戶號碼）。 \n", + "反之，稀疏詞彙搜尋（BM25）則會遺漏語義同義詞和同義改寫的問題。\n", + "\n", + "本手冊示範如何實作一個高速、記憶體內混合檢索引擎，結合：\n", + "1. 密集向量（餘弦相似度）\n", + "2. 稀疏詞彙搜尋（SQLite FTS5 BM25）\n", + "3. **相互排序融合（RRF, $k=60$）**\n", + "4. 使用 Microsoft `phi-4-mini` 進行 **具依據的引文生成（`[1]`、`[2]`）**。\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. 雙重 SQLite 架構（密集向量 + 虛擬 FTS5 BM25 表）\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. 樣本攝取及執行基準測試\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. 為 Microsoft phi-4-mini 制定有根據的提示語\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**免責聲明**：\n本文件使用 AI 翻譯服務 [Co-op Translator](https://github.com/Azure/co-op-translator) 進行翻譯。雖然我們力求準確，但請注意，自動翻譯可能包含錯誤或不準確之處。原始文件的母語版本應被視為權威來源。對於重要資訊，建議尋求專業人工翻譯。我們不對因使用本翻譯而引起的任何誤解或曲解承擔責任。\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file diff --git a/translations/zh-TW/.co-op-translator.json b/translations/zh-TW/.co-op-translator.json index 3a191e726..4b40a1981 100644 --- a/translations/zh-TW/.co-op-translator.json +++ b/translations/zh-TW/.co-op-translator.json @@ -12,8 +12,8 @@ "language_code": "zh-TW" }, "README.md": { - "original_hash": "a49252a2706935f21c8abf8471481553", - "translation_date": "2026-04-21T00:35:18+00:00", + "original_hash": "e1043c69855a4bf9bb560a087f3f0bff", + "translation_date": "2026-08-26T16:26:32+00:00", "source_file": "README.md", "language_code": "zh-TW" }, @@ -53,6 +53,12 @@ "source_file": "code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb", "language_code": "zh-TW" }, + "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb": { + "original_hash": "71f9303a811084cf162acf977625cda9", + "translation_date": "2026-08-26T16:27:39+00:00", + "source_file": "code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb", + "language_code": "zh-TW" + }, "code/07.Lab/01/AIPC/extensions/phi3ext/CHANGELOG.md": { "original_hash": "dbb0b6218ce5f9cf0ede8f4201f6ad58", "translation_date": "2025-07-16T16:30:00+00:00", diff --git a/translations/zh-TW/README.md b/translations/zh-TW/README.md index 11959fb14..d91f682aa 100644 --- a/translations/zh-TW/README.md +++ b/translations/zh-TW/README.md @@ -1,12 +1,12 @@ -# Phi Cookbook：使用 Microsoft Phi 模型的實作範例 +# Phi 食譜：Microsoft Phi 模型的實作範例 -[![在 GitHub Codespaces 打開並使用範例](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) -[![在開發容器中打開](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) +[![在 GitHub Codespaces 中開啟並使用範例](https://github.com/codespaces/badge.svg)](https://codespaces.new/microsoft/phicookbook) +[![在 Dev Containers 中開啟](https://img.shields.io/static/v1?style=for-the-badge&label=Dev%20Containers&message=Open&color=blue&logo=visualstudiocode)](https://vscode.dev/redirect?url=vscode://ms-vscode-remote.remote-containers/cloneInVolume?url=https://github.com/microsoft/phicookbook) [![GitHub 貢獻者](https://img.shields.io/github/contributors/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/graphs/contributors/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub 問題](https://img.shields.io/github/issues/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/issues/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub 拉取請求](https://img.shields.io/github/issues-pr/microsoft/phicookbook.svg)](https://GitHub.com/microsoft/phicookbook/pulls/?WT.mc_id=aiml-137032-kinfeylo) -[![歡迎 PR](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) +[![歡迎 Pull Request](https://img.shields.io/badge/PRs-welcome-brightgreen.svg?style=flat-square)](http://makeapullrequest.com?WT.mc_id=aiml-137032-kinfeylo) [![GitHub 觀察者](https://img.shields.io/github/watchers/microsoft/phicookbook.svg?style=social&label=Watch)](https://GitHub.com/microsoft/phicookbook/watchers/?WT.mc_id=aiml-137032-kinfeylo) [![GitHub 分支](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) @@ -14,29 +14,29 @@ [![Microsoft Foundry Discord](https://dcbadge.limes.pink/api/server/ByRwuEEgH4)](https://discord.com/invite/ByRwuEEgH4) -Phi 是由 Microsoft 開發的一系列開源 AI 模型。 +Phi 是 Microsoft 開發的一系列開源 AI 模型。 -Phi 目前是最強大且具有成本效益的小型語言模型 (SLM)，在多語言、推理、文本/聊天生成、程式碼、影像、音訊及其他應用場景中有出色的基準表現。 +Phi 目前是功能最強大且具成本效益的小型語言模型（SLM），在多語言、推理、文字/聊天生成、程式編碼、影像、音訊等多種場景中表現優異。 -你可以將 Phi 部署於雲端或邊緣設備，並且可以輕鬆地以有限的計算能力建構生成式 AI 應用。 +您可以將 Phi 部署到雲端或邊緣裝置，並能輕鬆利用有限的計算資源打造生成式 AI 應用。 -請按照以下步驟開始使用這些資源： -1. **Fork 本倉庫**：點擊 [![GitHub 分支](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) -2. **Clone 本倉庫**：`git clone https://github.com/microsoft/PhiCookBook.git` -3. [**加入 Microsoft AI Discord 社群，與專家和其他開發者交流**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) +按照以下步驟開始使用這些資源： +1. **Fork 此存放庫**：點選 [![GitHub forks](https://img.shields.io/github/forks/microsoft/phicookbook.svg?style=social&label=Fork)](https://GitHub.com/microsoft/phicookbook/network/?WT.mc_id=aiml-137032-kinfeylo) +2. **Clone 此存放庫**：`git clone https://github.com/microsoft/PhiCookBook.git` +3. [**加入 Microsoft AI Discord 社群，與專家及開發者交流**](https://discord.com/invite/ByRwuEEgH4?WT.mc_id=aiml-137032-kinfeylo) ![cover](../../translated_images/zh-TW/cover.eb18d1b9605d754b.webp) ### 🌐 多語言支援 -#### 透過 GitHub Action 支援（自動且始終保持最新） +#### 透過 GitHub Action 支援（自動且隨時更新） -[阿拉伯語](../ar/README.md) | [孟加拉語](../bn/README.md) | [保加利亞語](../bg/README.md) | [緬甸語 (緬甸)](../my/README.md) | [中文 (簡體)](../zh-CN/README.md) | [中文 (繁體，香港)](../zh-HK/README.md) | [中文 (繁體，澳門)](../zh-MO/README.md) | [中文 (繁體，臺灣)](./README.md) | [克羅埃西亞語](../hr/README.md) | [捷克語](../cs/README.md) | [丹麥語](../da/README.md) | [荷蘭語](../nl/README.md) | [愛沙尼亞語](../et/README.md) | [芬蘭語](../fi/README.md) | [法語](../fr/README.md) | [德語](../de/README.md) | [希臘語](../el/README.md) | [希伯來語](../he/README.md) | [印地語](../hi/README.md) | [匈牙利語](../hu/README.md) | [印尼語](../id/README.md) | [義大利語](../it/README.md) | [日語](../ja/README.md) | [坎那達語](../kn/README.md) | [高棉語](../km/README.md) | [韓語](../ko/README.md) | [立陶宛語](../lt/README.md) | [馬來語](../ms/README.md) | [馬拉雅拉姆語](../ml/README.md) | [馬拉地語](../mr/README.md) | [尼泊爾語](../ne/README.md) | [奈及利亞皮欽語](../pcm/README.md) | [挪威語](../no/README.md) | [波斯語 (法爾西)](../fa/README.md) | [波蘭語](../pl/README.md) | [葡萄牙語 (巴西)](../pt-BR/README.md) | [葡萄牙語 (葡萄牙)](../pt-PT/README.md) | [旁遮普語 (古爾穆奇)](../pa/README.md) | [羅馬尼亞語](../ro/README.md) | [俄語](../ru/README.md) | [塞爾維亞語 (西里爾字母)](../sr/README.md) | [斯洛伐克語](../sk/README.md) | [斯洛維尼亞語](../sl/README.md) | [西班牙語](../es/README.md) | [斯瓦希里語](../sw/README.md) | [瑞典語](../sv/README.md) | [他加祿語 (菲律賓語)](../tl/README.md) | [泰米爾語](../ta/README.md) | [泰盧固語](../te/README.md) | [泰語](../th/README.md) | [土耳其語](../tr/README.md) | [烏克蘭語](../uk/README.md) | [烏爾都語](../ur/README.md) | [越南語](../vi/README.md) +[阿拉伯語](../ar/README.md) | [孟加拉語](../bn/README.md) | [保加利亞語](../bg/README.md) | [緬甸語 (Myanmar)](../my/README.md) | [中文 (簡體)](../zh-CN/README.md) | [中文 (繁體，香港)](../zh-HK/README.md) | [中文 (繁體，澳門)](../zh-MO/README.md) | [中文 (繁體，台灣)](./README.md) | [克羅埃西亞語](../hr/README.md) | [捷克語](../cs/README.md) | [丹麥語](../da/README.md) | [荷蘭語](../nl/README.md) | [愛沙尼亞語](../et/README.md) | [芬蘭語](../fi/README.md) | [法語](../fr/README.md) | [德語](../de/README.md) | [希臘語](../el/README.md) | [希伯來語](../he/README.md) | [印地語](../hi/README.md) | [匈牙利語](../hu/README.md) | [印尼語](../id/README.md) | [義大利語](../it/README.md) | [日語](../ja/README.md) | [卡納達語](../kn/README.md) | [高棉語](../km/README.md) | [韓語](../ko/README.md) | [立陶宛語](../lt/README.md) | [馬來語](../ms/README.md) | [馬拉雅拉姆語](../ml/README.md) | [馬拉地語](../mr/README.md) | [尼泊爾語](../ne/README.md) | [奈及利亞洋拉語](../pcm/README.md) | [挪威語](../no/README.md) | [波斯語 (法爾西語)](../fa/README.md) | [波蘭語](../pl/README.md) | [葡萄牙語 (巴西)](../pt-BR/README.md) | [葡萄牙語 (葡萄牙)](../pt-PT/README.md) | [旁遮普語 (古魯穆奇體)](../pa/README.md) | [羅馬尼亞語](../ro/README.md) | [俄語](../ru/README.md) | [塞爾維亞語 (西里爾字母)](../sr/README.md) | [斯洛伐克語](../sk/README.md) | [斯洛文尼亞語](../sl/README.md) | [西班牙語](../es/README.md) | [斯瓦希里語](../sw/README.md) | [瑞典語](../sv/README.md) | [他加祿語 (菲律賓語)](../tl/README.md) | [泰米爾語](../ta/README.md) | [泰盧固語](../te/README.md) | [泰語](../th/README.md) | [土耳其語](../tr/README.md) | [烏克蘭語](../uk/README.md) | [烏爾都語](../ur/README.md) | [越南語](../vi/README.md) -> **偏好本機 Clone？** +> **偏好本地 Clone 嗎？** > -> 本倉庫包含 50 多種語言翻譯，這大幅增加下載大小。若只需核心內容，請使用稀疏檢出（sparse checkout）： +> 本存放庫包含 50 多種語言翻譯，會大幅增加下載大小。如需不帶翻譯版本，請使用稀疏檢出： > > **Bash / macOS / Linux：** > ```bash @@ -45,180 +45,183 @@ Phi 目前是最強大且具有成本效益的小型語言模型 (SLM)，在多 > git sparse-checkout set --no-cone '/*' '!translations' '!translated_images' > ``` > -> **CMD (Windows)：** +> **CMD（Windows）：** > ```cmd > git clone --filter=blob:none --sparse https://github.com/microsoft/PhiCookBook.git > cd PhiCookBook > git sparse-checkout set --no-cone "/*" "!translations" "!translated_images" > ``` > -> 這樣可以更快速地下載所需內容，完成學習課程。 +> 這會讓您在更快的下載速度下取得完成課程所需的全部內容。 ## 目錄 -- 介紹 +- 簡介 - [歡迎加入 Phi 家族](./md/01.Introduction/01/01.PhiFamily.md) - - [設定你的開發環境](./md/01.Introduction/01/01.EnvironmentSetup.md) + - [環境設定](./md/01.Introduction/01/01.EnvironmentSetup.md) - [理解關鍵技術](./md/01.Introduction/01/01.Understandingtech.md) - [Phi 模型的 AI 安全性](./md/01.Introduction/01/01.AISafety.md) - [Phi 硬體支援](./md/01.Introduction/01/01.Hardwaresupport.md) - - [Phi 模型與跨平台可用性](./md/01.Introduction/01/01.Edgeandcloud.md) + - [Phi 模型及跨平台可用性](./md/01.Introduction/01/01.Edgeandcloud.md) - [使用 Guidance-ai 與 Phi](./md/01.Introduction/01/01.Guidance.md) - - [GitHub Marketplace 模型](https://github.com/marketplace/models) + - [GitHub 市集模型](https://github.com/marketplace/models) - [Azure AI 模型目錄](https://ai.azure.com) -- 在不同環境中使用 Phi 推論 - - [Hugging face](./md/01.Introduction/02/01.HF.md) +- 在不同環境中推論 Phi + - [Hugging Face](./md/01.Introduction/02/01.HF.md) - [GitHub 模型](./md/01.Introduction/02/02.GitHubModel.md) - [Microsoft Foundry 模型目錄](./md/01.Introduction/02/03.AzureAIFoundry.md) - [Ollama](./md/01.Introduction/02/04.Ollama.md) - - [AI 工具組 VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) + - [AI Toolkit VSCode (AITK)](./md/01.Introduction/02/05.AITK.md) - [NVIDIA NIM](./md/01.Introduction/02/06.NVIDIA.md) - - [Foundry 本地端](./md/01.Introduction/02/07.FoundryLocal.md) + - [Foundry Local](./md/01.Introduction/02/07.FoundryLocal.md) - Phi 家族推論 - [iOS 推論 Phi](./md/01.Introduction/03/iOS_Inference.md) - [Android 推論 Phi](./md/01.Introduction/03/Android_Inference.md) - [Jetson 推論 Phi](./md/01.Introduction/03/Jetson_Inference.md) - [AI PC 推論 Phi](./md/01.Introduction/03/AIPC_Inference.md) - - [使用 Apple MLX 框架推論 Phi](./md/01.Introduction/03/MLX_Inference.md) + - [使用 Apple MLX Framework 推論 Phi](./md/01.Introduction/03/MLX_Inference.md) - [本地伺服器推論 Phi](./md/01.Introduction/03/Local_Server_Inference.md) - - [使用 AI 工具組於遠端伺服器推論 Phi](./md/01.Introduction/03/Remote_Interence.md) + - [使用 AI Toolkit 在遠端伺服器推論 Phi](./md/01.Introduction/03/Remote_Interence.md) - [使用 Rust 推論 Phi](./md/01.Introduction/03/Rust_Inference.md) - - [本地端 Phi 視覺推論](./md/01.Introduction/03/Vision_Inference.md) - - [配合 Kaito AKS、Azure 容器推論 Phi（官方支援）](./md/01.Introduction/03/Kaito_Inference.md) + - [本地推論 Phi 視覺](./md/01.Introduction/03/Vision_Inference.md) + - [使用 Kaito AKS、Azure Containers（官方支援）推論 Phi](./md/01.Introduction/03/Kaito_Inference.md) - [Phi 家族量化](./md/01.Introduction/04/QuantifyingPhi.md) - [使用 llama.cpp 量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingLlamacppQuantifyingPhi.md) - - [使用 onnxruntime 生成式 AI 擴充功能量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) + - [使用 onnxruntime 的生成式 AI 擴充功能量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingORTGenAIQuantifyingPhi.md) - [使用 Intel OpenVINO 量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingIntelOpenVINOQuantifyingPhi.md) - - [使用 Apple MLX 框架量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) + - [使用 Apple MLX Framework 量化 Phi-3.5 / 4](./md/01.Introduction/04/UsingAppleMLXQuantifyingPhi.md) -- Phi 評估 +- Phi 評估 - [負責任的 AI](./md/01.Introduction/05/ResponsibleAI.md) - [Microsoft Foundry 評估](./md/01.Introduction/05/AIFoundry.md) - [使用 Promptflow 進行評估](./md/01.Introduction/05/Promptflow.md) -- 結合 Azure AI Search 的 RAG - - [如何使用 Phi-4-mini 與 Phi-4 多模態 (RAG) 配合 Azure AI Search](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) +- 使用 RAG 與 Azure AI 搜尋 + - [如何使用 Phi-4-mini 及 Phi-4-multimodal (RAG) 與 Azure AI 搜尋](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-RAG-Azure-AI-Search.ipynb) + - [零雲端本地混合 RAG 搭配 SQLite FTS5 和 phi-4-mini](https://github.com/microsoft/PhiCookBook/blob/main/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb) - Phi 應用開發範例 - 文字與聊天應用 - Phi-4 範例 - [📓] [與 Phi-4-mini ONNX 模型聊天](./md/02.Application/01.TextAndChat/Phi4/ChatWithPhi4ONNX/README.md) - - [使用 Phi-4 本地 ONNX 模型的聊天 .NET 範例](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) - - [使用 Phi-4 ONNX 與 Sementic Kernel 的 .NET 控制台聊天應用](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - [使用 Phi-4 本地 ONNX 模型聊天室 .NET](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-01OnnxRuntime) + - [使用 Sementic Kernel 的 Phi-4 ONNX 聊天 .NET 控制台應用](../../md/04.HOL/dotnet/src/LabsPhi4-Chat-02SK) + - Phi-3 / 3.5 範例 - - [在瀏覽器中使用 Phi3、ONNX Runtime Web 及 WebGPU 的本地聊天機器人](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) + - [使用 Phi3、ONNX Runtime Web 和 WebGPU 在瀏覽器中本地聊天機器人](https://github.com/microsoft/onnxruntime-inference-examples/tree/main/js/chat) - [OpenVino 聊天](./md/02.Application/01.TextAndChat/Phi3/E2E_OpenVino_Chat.md) - [多模型 - 互動式 Phi-3-mini 與 OpenAI Whisper](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-mini_with_whisper.md) - - [MLFlow - 建立包裝並使用 Phi-3 與 MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) - - [模型優化 - 如何使用 Olive 為 ONNX Runtime Web 優化 Phi-3-mini 模型](https://github.com/microsoft/Olive/tree/main/examples/phi3) - - [WinUI3 App 搭配 Phi-3 mini-4k-instruct-onnx](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) - -[WinUI3 多模型 AI 助力筆記應用程式範例](https://github.com/microsoft/ai-powered-notes-winui3-sample) + - [MLFlow - 建構 Wrapper 並使用 Phi-3 搭配 MLFlow](./md//02.Application/01.TextAndChat/Phi3/E2E_Phi-3-MLflow.md) + - [模型優化 - 如何使用 Olive 對 ONNX Runtime Web 優化 Phi-3-min 模型](https://github.com/microsoft/Olive/tree/main/examples/phi3) + - [帶有 Phi-3 mini-4k-instruct-onnx 的 WinUI3 應用程式](https://github.com/microsoft/Phi3-Chat-WinUI3-Sample/) + -[WinUI3 多模型 AI 助力筆記應用範例](https://github.com/microsoft/ai-powered-notes-winui3-sample) - [使用 Prompt flow 微調並整合自訂 Phi-3 模型](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration.md) - [在 Microsoft Foundry 中使用 Prompt flow 微調並整合自訂 Phi-3 模型](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-FineTuning_PromptFlow_Integration_AIFoundry.md) - - [在 Microsoft Foundry 中評估微調的 Phi-3 / Phi-3.5 模型，重點關注微軟的負責任 AI 原則](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) - - [📓] [Phi-3.5-mini-instruct 語言預測範例（中文／英文）](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) + - [在 Microsoft Foundry 中評估微調後的 Phi-3 / Phi-3.5 模型，專注於微軟責任 AI 原則](./md/02.Application/01.TextAndChat/Phi3/E2E_Phi-3-Evaluation_AIFoundry.md) + - [📓] [Phi-3.5-mini-instruct 語言預測範例（中文/英文）](./md/02.Application/01.TextAndChat/Phi3/phi3-instruct-demo.ipynb) - [Phi-3.5-Instruct WebGPU RAG 聊天機器人](./md/02.Application/01.TextAndChat/Phi3/WebGPUWithPhi35Readme.md) - - [使用 Windows GPU 建立搭配 Phi-3.5-Instruct ONNX 的 Prompt flow 解決方案](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) - - [使用 Microsoft Phi-3.5 tflite 建立 Android 應用程式](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) - - [Q&A .NET 範例，使用本地 ONNX Phi-3 模型搭配 Microsoft.ML.OnnxRuntime](../../md/04.HOL/dotnet/src/LabsPhi301) - - [Console 聊天 .NET 應用程式，搭配 Semantic Kernel 與 Phi-3](../../md/04.HOL/dotnet/src/LabsPhi302) + - [使用 Windows GPU 建立 Phi-3.5-Instruct ONNX 的 Prompt flow 解決方案](./md/02.Application/01.TextAndChat/Phi3/UsingPromptFlowWithONNX.md) + - [使用微軟 Phi-3.5 tflite 建立 Android 應用程式](./md/02.Application/01.TextAndChat/Phi3/UsingPhi35TFLiteCreateAndroidApp.md) + - [使用 Microsoft.ML.OnnxRuntime 以本地 ONNX Phi-3 模型的 .NET Q&A 範例](../../md/04.HOL/dotnet/src/LabsPhi301) + - [帶有語義核心和 Phi-3 的控制台聊天 .NET 應用程式](../../md/04.HOL/dotnet/src/LabsPhi302) - Azure AI 推論 SDK 代碼範例 - Phi-4 範例 - - [📓] [使用 Phi-4-multimodal 產生專案代碼](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) + - [📓] [使用 Phi-4-multimodal 生成專案代碼](./md/02.Application/02.Code/Phi4/GenProjectCode/README.md) - Phi-3 / 3.5 範例 - - [使用 Microsoft Phi-3 家族打造您自己的 Visual Studio Code GitHub Copilot Chat](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) - - [使用 GitHub 模型建立您自己的 Visual Studio Code 聊天 Copilot 代理人 Phi-3.5](./md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) + - [使用微軟 Phi-3 系列建立您自己的 Visual Studio Code GitHub Copilot 聊天](./md/02.Application/02.Code/Phi3/VSCodeExt/README.md) + - [使用 GitHub 模型與 Phi-3.5 創建您自己的 Visual Studio Code 聊天 Copilot 代理](/md/02.Application/02.Code/Phi3/CreateVSCodeChatAgentWithGitHubModels.md) - 進階推理範例 - Phi-4 範例 - - [📓] [Phi-4-mini-reasoning 或 Phi-4-reasoning 範例](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) - - [📓] [使用 Microsoft Olive 微調 Phi-4-mini-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [使用 Apple MLX 微調 Phi-4-mini-reasoning](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) - - [📓] [Phi-4-mini-reasoning 搭配 GitHub 模型](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) - - [📓] [Phi-4-mini-reasoning 搭配 Microsoft Foundry 模型](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) - - 演示 - - [Phi-4-mini 演示，托管於 Hugging Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) - - [Phi-4-multimodal 演示，托管於 Hugginge Face Spaces](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) + - [📓] [Phi-4-mini-推理或 Phi-4-推理範例](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/README.md) + - [📓] [使用 Microsoft Olive 微調 Phi-4-mini-推理](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/olive_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [使用 Apple MLX 微調 Phi-4-mini-推理](./md/02.Application/03.AdvancedReasoning/Phi4/AdvancedResoningPhi4mini/mlx_ft_phi_4_reasoning_with_medicaldata.ipynb) + - [📓] [使用 GitHub 模型的 Phi-4-mini-推理](./md/02.Application/02.Code/Phi4r/github_models_inference.ipynb) + - [📓] [使用 Microsoft Foundry 模型的 Phi-4-mini-推理](./md/02.Application/02.Code/Phi4r/azure_models_inference.ipynb) + - 範例演示 + - [Hugging Face Spaces 上的 Phi-4-mini 演示](https://huggingface.co/spaces/microsoft/phi-4-mini?WT.mc_id=aiml-137032-kinfeylo) + - [Hugging Face Spaces 上的 Phi-4-multimodal 演示](https://huggingface.co/spaces/microsoft/phi-4-multimodal?WT.mc_id=aiml-137032-kinfeylo) - 視覺範例 - Phi-4 範例 - - [📓] [使用 Phi-4-multimodal 讀取影像並生成代碼](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) + - [📓] [使用 Phi-4-multimodal 讀取圖片並生成代碼](./md/02.Application/04.Vision/Phi4/CreateFrontend/README.md) - Phi-3 / 3.5 範例 - - [📓][Phi-3-視覺-影像文字到文字](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [📓][Phi-3-vision-影像文字轉文字](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - [Phi-3-vision-ONNX](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - - [📓][Phi-3-視覺 CLIP 向量](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) - - [演示：Phi-3 回收](https://github.com/jennifermarsman/PhiRecycling/) - - [Phi-3-vision - 視覺語言助理 - 搭配 Phi3-Vision 與 OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) - - [Phi-3 視覺 Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) - - [Phi-3 視覺 OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) - - [📓][Phi-3.5 視覺多幀或多影像範例](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) - - [Phi-3 視覺本地 ONNX 模型，使用 Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi303) - - [菜單式 Phi-3 視覺本地 ONNX 模型，使用 Microsoft.ML.OnnxRuntime .NET](../../md/04.HOL/dotnet/src/LabsPhi304) + - [📓][Phi-3-vision CLIP 嵌入](./md/02.Application/04.Vision/Phi3/E2E_Phi-3-vision-image-text-to-text-online-endpoint.ipynb) + - [DEMO: Phi-3 回收](https://github.com/jennifermarsman/PhiRecycling/) + - [Phi-3-vision - 視覺語言助理 - 包含 Phi3-Vision 及 OpenVINO](https://docs.openvino.ai/nightly/notebooks/phi-3-vision-with-output.html) + - [Phi-3 Vision Nvidia NIM](./md/02.Application/04.Vision/Phi3/E2E_Nvidia_NIM_Vision.md) + - [Phi-3 Vision OpenVino](./md/02.Application/04.Vision/Phi3/E2E_OpenVino_Phi3Vision.md) + - [📓][Phi-3.5 Vision 多幀或多影像範例](./md/02.Application/04.Vision/Phi3/phi3-vision-demo.ipynb) + - [使用 Microsoft.ML.OnnxRuntime .NET 的 Phi-3 Vision 本地 ONNX 模型](../../md/04.HOL/dotnet/src/LabsPhi303) + - [使用 Microsoft.ML.OnnxRuntime .NET 的選單式 Phi-3 Vision 本地 ONNX 模型](../../md/04.HOL/dotnet/src/LabsPhi304) - 推理-視覺範例 - Phi-4-Reasoning-Vision-15B - - [📓] [使用 Phi-4-Reasoning-Vision-15B 進行闖紅燈檢測](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) - - [📓] [使用 Phi-4-Reasoning-Vision-15B 進行數學推理](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) - - [📓] [使用 Phi-4-Reasoning-Vision-15B 進行介面檢測](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) + - [📓] [使用 Phi-4-Reasoning-Vision-15B 偵測闖紅燈行人](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Jaywalking.ipynb) + - [📓] [使用 Phi-4-Reasoning-Vision-15B 進行數學運算](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_Math.ipynb) + - [📓] [使用 Phi-4-Reasoning-Vision-15B 偵測 UI](./md/02.Application/10.ReasoningVision/Phi_4_reasoning_vision_15b_ui.ipynb) - 數學範例 - - Phi-4-Mini-Flash-Reasoning-Instruct 範例 [Phi-4-Mini-Flash-Reasoning-Instruct 數學演示](./md/02.Application/09.Math/MathDemo.ipynb) + - Phi-4-Mini-Flash-Reasoning-Instruct 範例 [使用 Phi-4-Mini-Flash-Reasoning-Instruct 的數學演示](./md/02.Application/09.Math/MathDemo.ipynb) - - 語音範例 + - 音訊範例 - Phi-4 範例 - - [📓] [使用 Phi-4-multimodal 擷取音檔文字記錄](./md/02.Application/05.Audio/Phi4/Transciption/README.md) - - [📓] [Phi-4-multimodal 語音範例](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) + - [📓] [使用 Phi-4-multimodal 擷取音訊文字記錄](./md/02.Application/05.Audio/Phi4/Transciption/README.md) + - [📓] [Phi-4-multimodal 音訊範例](./md/02.Application/05.Audio/Phi4/Siri/demo.ipynb) - [📓] [Phi-4-multimodal 語音翻譯範例](./md/02.Application/05.Audio/Phi4/Translate/demo.ipynb) - - [.NET 控制台應用程式使用 Phi-4-multimodal 音頻分析與匯出文字記錄](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) + - [使用 Phi-4-multimodal 的 .NET 控制台應用程式分析音訊檔並生成文字記錄](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-02Audio) - - MOE 範例 + - MoE 範例 - Phi-3 / 3.5 範例 - - [📓] [Phi-3.5 專家混合模型 (MoEs) 社交媒體範例](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) - - [📓] [使用 NVIDIA NIM Phi-3 MOE、Azure AI 搜尋與 LlamaIndex 建立檢索增強生成 (RAG) 流程](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) + - [📓] [Phi-3.5 專家混合模型 (MoEs) 社群媒體範例](./md/02.Application/06.MoE/Phi3/phi3_moe_demo.ipynb) + - [📓] [使用 NVIDIA NIM Phi-3 MOE、Azure AI 搜尋及 LlamaIndex 建立檢索增強生成 (RAG) 管線](./md/02.Application/06.MoE/Phi3/azure-ai-search-nvidia-rag.ipynb) - - - 函數呼叫範例 + - 函式呼叫範例 - Phi-4 範例 🆕 - - [📓] [使用 Phi-4-mini 的函數呼叫](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) - - [📓] [使用函數呼叫創建多代理人，搭配 Phi-4-mini](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) - - [📓] [使用 Ollama 的函數呼叫](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) - - [📓] [使用 ONNX 的函數呼叫](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) - - 多模態混合範例 + - [📓] [在 Phi-4-mini 中使用函式呼叫](./md/02.Application/07.FunctionCalling/Phi4/FunctionCallingBasic/README.md) + - [📓] [使用函式呼叫建構 Phi-4-mini 的多重代理](./md/02.Application/07.FunctionCalling/Phi4/Multiagents/Phi_4_mini_multiagent.ipynb) + - [📓] [與 Ollama 使用函式呼叫](./md/02.Application/07.FunctionCalling/Phi4/Ollama/ollama_functioncalling.ipynb) + - [📓] [與 ONNX 使用函式呼叫](./md/02.Application/07.FunctionCalling/Phi4/ONNX/onnx_parallel_functioncalling.ipynb) + - 多模態混和範例 - Phi-4 範例 🆕 - - [📓] [使用 Phi-4-multimodal 擔任科技記者](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) - - [.NET 控制台應用程式，使用 Phi-4-multimodal 分析影像](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) + - [📓] [將 Phi-4-multimodal 用作科技記者](./md/02.Application/08.Multimodel/Phi4/TechJournalist/phi_4_mm_audio_text_publish_news.ipynb) + - [使用 Phi-4-multimodal 的 .NET 控制台應用程式來分析影像](../../md/04.HOL/dotnet/src/LabsPhi4-MultiModal-01Images) - Phi 微調範例 - - [微調情境](./md/03.FineTuning/FineTuning_Scenarios.md) - - [微調與 RAG 的差異](./md/03.FineTuning/FineTuning_vs_RAG.md) - - [微調讓 Phi-3 成為行業專家](./md/03.FineTuning/LetPhi3gotoIndustriy.md) - - [搭配 AI Toolkit for VS Code 微調 Phi-3](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) + - [微調場景](./md/03.FineTuning/FineTuning_Scenarios.md) + - [微調與 RAG](./md/03.FineTuning/FineTuning_vs_RAG.md) + - [微調讓 Phi-3 成為產業專家](./md/03.FineTuning/LetPhi3gotoIndustriy.md) + - [使用 AI Toolkit for VS Code 微調 Phi-3](./md/03.FineTuning/Finetuning_VSCodeaitoolkit.md) - [使用 Azure 機器學習服務微調 Phi-3](./md/03.FineTuning/Introduce_AzureML.md) - [使用 Lora 微調 Phi-3](./md/03.FineTuning/FineTuning_Lora.md) - [使用 QLora 微調 Phi-3](./md/03.FineTuning/FineTuning_Qlora.md) - [使用 Microsoft Foundry 微調 Phi-3](./md/03.FineTuning/FineTuning_AIFoundry.md) - - [透過 Azure ML CLI/SDK 微調 Phi-3](./md/03.FineTuning/FineTuning_MLSDK.md) + - [使用 Azure ML CLI/SDK 微調 Phi-3](./md/03.FineTuning/FineTuning_MLSDK.md) - [使用 Microsoft Olive 微調](./md/03.FineTuning/FineTuning_MicrosoftOlive.md) - [Microsoft Olive 實作實驗室微調](./md/03.FineTuning/olive-lab/readme.md) - [使用 Weights and Bias 微調 Phi-3-vision](./md/03.FineTuning/FineTuning_Phi-3-visionWandB.md) - - [使用 Apple MLX Framework 微調 Phi-3](./md/03.FineTuning/FineTuning_MLX.md) - - [Phi-3-vision 微調（官方支持）](./md/03.FineTuning/FineTuning_Vision.md) - - [使用 Kaito AKS 進行 Phi-3 微調，Azure Containers（官方支援）](./md/03.FineTuning/FineTuning_Kaito.md) - - [Phi-3 和 3.5 Vision 微調](https://github.com/2U1/Phi3-Vision-Finetune) -- 實作實驗室 - - [探索最前沿模型：LLMs、SLMs、本地開發及更多](https://github.com/microsoft/aitour-exploring-cutting-edge-models) - - [開啟 NLP 潛能：使用 Microsoft Olive 進行微調](https://github.com/azure/Ignite_FineTuning_workshop) + - [使用 Apple MLX 框架微調 Phi-3](./md/03.FineTuning/FineTuning_MLX.md) + - [微調 Phi-3-vision（官方支援）](./md/03.FineTuning/FineTuning_Vision.md) + - [使用 Kaito AKS 及 Azure 容器微調 Phi-3（官方支援）](./md/03.FineTuning/FineTuning_Kaito.md) + - [微調 Phi-3 與 3.5 Vision](https://github.com/2U1/Phi3-Vision-Finetune) + +- 動手實作實驗室 + - [探索前沿模型：大型語言模型、專業語言模型、本地開發等](https://github.com/microsoft/aitour-exploring-cutting-edge-models) + - [釋放 NLP 潛力：使用 Microsoft Olive 進行微調](https://github.com/azure/Ignite_FineTuning_workshop) -- 學術研究論文與出版物 +- 學術研究論文與出版品 - [Textbooks Are All You Need II：phi-1.5 技術報告](https://arxiv.org/abs/2309.05463) - - [Phi-3 技術報告：在您的手機上本地運行的高度強大語言模型](https://arxiv.org/abs/2404.14219) + - [Phi-3 技術報告：高性能語言模型在您的手機上本地運行](https://arxiv.org/abs/2404.14219) - [Phi-4 技術報告](https://arxiv.org/abs/2412.08905) - - [Phi-4-Mini 技術報告：透過 Mixture-of-LoRAs 的小巧但強大的多模態語言模型](https://arxiv.org/abs/2503.01743) - - [針對車載功能調用優化的小型語言模型](https://arxiv.org/abs/2501.02342) - - [(WhyPHI) 針對多選擇題回答微調 PHI-3：方法、結果與挑戰](https://arxiv.org/abs/2501.01588) + - [Phi-4-Mini 技術報告：通過 LoRA 混合實現緊湊而強大的多模態語言模型](https://arxiv.org/abs/2503.01743) + - [優化小型語言模型用於車載功能呼叫](https://arxiv.org/abs/2501.02342) + - [(WhyPHI) 微調 PHI-3 用於多選題問答：方法學、結果與挑戰](https://arxiv.org/abs/2501.01588) - [Phi-4 推理技術報告](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/04/phi_4_reasoning.pdf) - [Phi-4-mini 推理技術報告](https://huggingface.co/microsoft/Phi-4-mini-reasoning/blob/main/Phi-4-Mini-Reasoning.pdf) @@ -226,28 +229,28 @@ Phi 目前是最強大且具有成本效益的小型語言模型 (SLM)，在多 ### Microsoft Foundry 上的 Phi -您可以學習如何使用 Microsoft Phi 以及如何在不同硬體裝置上構建端到端解決方案。想要親身體驗 Phi，請從使用模型開始，並利用 [Microsoft Foundry Azure AI 模型目錄](https://aka.ms/phi3-azure-ai) 為您的場景自訂 Phi。您可以在 [Microsoft Foundry 入門指南](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) 了解更多。 +您可以學習如何使用 Microsoft Phi 以及如何在不同硬體裝置上構建端對端解決方案。若要親自體驗 Phi，請先從試玩模型並為您的場景自定義 Phi 開始，透過 [Microsoft Foundry Azure AI Model Catalog](https://aka.ms/phi3-azure-ai)。您也可以參考 Getting Started with [Microsoft Foundry](/md/02.QuickStart/AzureAIFoundry_QuickStart.md) 了解更多資訊。 遊樂場 -每個模型都有專屬的遊樂場來測試模型 [Azure AI Playground](https://aka.ms/try-phi3)。 +每個模型都有專屬遊樂場可供測試模型，[Azure AI Playground](https://aka.ms/try-phi3)。 -### GitHub 模型上的 Phi +### GitHub 上的 Phi 模型 -您可以學習如何使用 Microsoft Phi 以及如何在不同硬體裝置上構建端到端解決方案。想要親身體驗 Phi，請從使用模型開始，並利用 [GitHub 模型目錄](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo) 為您的場景自訂 Phi。您可以在 [GitHub 模型目錄入門指南](/md/02.QuickStart/GitHubModel_QuickStart.md) 了解更多。 +您可以學習如何使用 Microsoft Phi 以及如何在不同硬體裝置上構建端對端解決方案。若要親自體驗 Phi，請先試玩模型並為您的場景自定義 Phi，透過 [GitHub Model Catalog](https://github.com/marketplace/models?WT.mc_id=aiml-137032-kinfeylo)。您可透過 Getting Started with [GitHub Model Catalog](/md/02.QuickStart/GitHubModel_QuickStart.md) 瞭解更多。 遊樂場 每個模型都有專屬的[遊樂場來測試模型](/md/02.QuickStart/GitHubModel_QuickStart.md)。 ### Hugging Face 上的 Phi -您也可以在 [Hugging Face](https://huggingface.co/microsoft) 上找到該模型。 +您也可以在 [Hugging Face](https://huggingface.co/microsoft) 找到該模型 遊樂場 [Hugging Chat 遊樂場](https://huggingface.co/chat/models/microsoft/Phi-3-mini-4k-instruct) ## 🎒 其他課程 -我們團隊還製作了其他課程！請查看： +我們團隊還製作了其他課程！請參考： ### LangChain @@ -277,49 +280,50 @@ Phi 目前是最強大且具有成本效益的小型語言模型 (SLM)，在多 [![資料科學初學者](https://img.shields.io/badge/Data%20Science%20for%20Beginners-84CC16?style=for-the-badge&labelColor=E5E7EB&color=84CC16)](https://aka.ms/datascience-beginners?WT.mc_id=academic-105485-koreyst) [![人工智慧初學者](https://img.shields.io/badge/AI%20for%20Beginners-A3E635?style=for-the-badge&labelColor=E5E7EB&color=A3E635)](https://aka.ms/ai-beginners?WT.mc_id=academic-105485-koreyst) [![資安初學者](https://img.shields.io/badge/Cybersecurity%20for%20Beginners-F97316?style=for-the-badge&labelColor=E5E7EB&color=F97316)](https://github.com/microsoft/Security-101?WT.mc_id=academic-96948-sayoung) -[![網站開發初學者](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) +[![網頁開發初學者](https://img.shields.io/badge/Web%20Dev%20for%20Beginners-EC4899?style=for-the-badge&labelColor=E5E7EB&color=EC4899)](https://aka.ms/webdev-beginners?WT.mc_id=academic-105485-koreyst) [![物聯網初學者](https://img.shields.io/badge/IoT%20for%20Beginners-14B8A6?style=for-the-badge&labelColor=E5E7EB&color=14B8A6)](https://aka.ms/iot-beginners?WT.mc_id=academic-105485-koreyst) [![XR 開發初學者](https://img.shields.io/badge/XR%20Development%20for%20Beginners-38BDF8?style=for-the-badge&labelColor=E5E7EB&color=38BDF8)](https://github.com/microsoft/xr-development-for-beginners?WT.mc_id=academic-105485-koreyst) --- ### Copilot 系列 -[![AI 配對程式設計 Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) -[![C#/.NET Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) +[![AI 配對程式設計的 Copilot](https://img.shields.io/badge/Copilot%20for%20AI%20Paired%20Programming-FACC15?style=for-the-badge&labelColor=E5E7EB&color=FACC15)](https://aka.ms/GitHubCopilotAI?WT.mc_id=academic-105485-koreyst) +[![C#/.NET 的 Copilot](https://img.shields.io/badge/Copilot%20for%20C%23/.NET-FBBF24?style=for-the-badge&labelColor=E5E7EB&color=FBBF24)](https://github.com/microsoft/mastering-github-copilot-for-dotnet-csharp-developers?WT.mc_id=academic-105485-koreyst) [![Copilot 冒險](https://img.shields.io/badge/Copilot%20Adventure-FDE68A?style=for-the-badge&labelColor=E5E7EB&color=FDE68A)](https://github.com/microsoft/CopilotAdventures?WT.mc_id=academic-105485-koreyst) ## 負責任的 AI -微軟致力於幫助我們的客戶負責任地使用 AI 產品，分享我們的學習經驗，並透過像是透明度說明和影響評估等工具建立基於信任的合作關係。許多相關資源可於 [https://aka.ms/RAI](https://aka.ms/RAI) 找到。 -微軟的負責任 AI 方法基於我們的 AI 原則，包括公平性、可靠性與安全性、隱私與安全、包容性、透明度與問責。 +微軟致力於幫助我們的客戶負責任地使用 AI 產品，分享我們的學習經驗，並透過像是透明度說明和影響評估等工具建立以信任為基礎的夥伴關係。許多這些資源可在 [https://aka.ms/RAI](https://aka.ms/RAI) 找到。 +微軟在負責任 AI 的方法基於我們的 AI 原則：公平、可靠與安全、隱私與安全性、包容性、透明度與問責制。 + +大型自然語言、圖像與語音模型——如本範例中使用的——可能會表現出不公平、不可靠或冒犯性的行為，進而造成損害。請查閱 [Azure OpenAI 服務透明度說明](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text)，以了解風險與限制。 + -大規模的自然語言、影像及語音模型——如本範例中使用的模型——可能會有表現不公平、不可靠或冒犯性的行為，進而造成傷害。請參考 [Azure OpenAI 服務透明度說明](https://learn.microsoft.com/legal/cognitive-services/openai/transparency-note?tabs=text) 以了解風險與限制。 -減輕這些風險的建議方法是在您的架構中包含能夠偵測並防止有害行為的安全系統。[Azure AI 內容安全](https://learn.microsoft.com/azure/ai-services/content-safety/overview) 提供獨立的保護層，能夠偵測應用程式和服務中的有害使用者生成內容及 AI 生成內容。Azure AI 內容安全包含文字和圖片 API，讓您能夠偵測有害素材。在 Microsoft Foundry 中，內容安全服務允許您檢視、探索並嘗試各種跨模態偵測有害內容的範例程式碼。以下的[快速入門文件](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest) 將引導您如何向該服務發送請求。 +減輕這些風險的建議方法是在您的架構中包含一個安全系統，該系統能夠檢測和防止有害行為。[Azure AI 內容安全](https://learn.microsoft.com/azure/ai-services/content-safety/overview) 提供獨立的保護層，能夠在應用程式和服務中檢測有害的用戶生成和 AI 生成的內容。Azure AI 內容安全包括文字和圖像 API，允許您檢測有害的材料。在 Microsoft Foundry 中，內容安全服務允許您查看、探索並嘗試不同模態中檢測有害內容的範例程式碼。以下的[快速入門文件](https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-text?tabs=visual-studio%2Clinux&pivots=programming-language-rest)將引導您如何向服務發出請求。 -另一個需要考慮的方面是整體應用程式的效能。對於多模態和多模型的應用程式，我們將效能視為系統能如您和使用者所期望地運作，包括不產生有害輸出。使用[效能與品質及風險與安全評估工具](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)評估您整體應用程式的效能非常重要。您也可以使用[自訂評估器](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators)來建立和進行評估。 +另一個需要考慮的方面是整體應用程式的效能。對於多模態和多模型的應用程式，我們認為效能是指系統能如您和使用者所期望地運作，包括不產生有害輸出。評估您整體應用程式的效能，請使用[效能與品質以及風險與安全評估器](https://learn.microsoft.com/azure/ai-studio/concepts/evaluation-metrics-built-in)。您也可以使用[自訂評估器](https://learn.microsoft.com/azure/ai-studio/how-to/develop/evaluate-sdk#custom-evaluators)建立並執行評估。 -您可以在您的開發環境中，透過[Azure AI 評估 SDK](https://microsoft.github.io/promptflow/index.html)評估您的 AI 應用程式。無論是給定測試數據集或目標，您的生成式 AI 應用程式產生結果都可以透過內建或您選擇的自訂評估器進行量化衡量。若要開始使用 Azure AI 評估 SDK 評估系統，您可以遵循[快速入門指南](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)。執行評估流程後，您可以在[Microsoft Foundry 中視覺化結果](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)。 +您可以使用[Azure AI 評估 SDK](https://microsoft.github.io/promptflow/index.html)，在開發環境中評估您的 AI 應用程式。給定測試資料集或目標，您可使用內建評估器或您所選擇的自訂評估器來定量衡量生成式 AI 應用程式的產出。若要開始使用 Azure AI 評估 SDK 評估您的系統，可以參考[快速入門指南](https://learn.microsoft.com/azure/ai-studio/how-to/develop/flow-evaluate-sdk)。執行評估運行後，您可在[Microsoft Foundry 中視覺化結果](https://learn.microsoft.com/azure/ai-studio/how-to/evaluate-flow-results)。 ## 商標 -本專案可能包含專案、產品或服務的商標或標誌。微軟商標或標誌的授權使用需遵守並依循[微軟商標與品牌指引](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)。 -在本專案的修改版本中使用微軟商標或標誌，必須避免引起混淆或暗示微軟贊助。 -任何使用第三方商標或標誌，均需遵守該第三方的相關政策。 +本專案可能包含專案、產品或服務的商標或標誌。授權使用 Microsoft 商標或標誌必須遵循且受限於[Microsoft 的商標與品牌指南](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general)。 +在修改版本中使用 Microsoft 商標或標誌時，必須避免造成混淆或暗示 Microsoft 的贊助。任何第三方商標或標誌的使用，皆需遵守該第三方的政策。 ## 尋求協助 -如果您遇到困難或對構建 AI 應用程式有任何疑問，歡迎加入： +如果您遇到困難或對構建 AI 應用程式有任何疑問，請加入： [![Microsoft Foundry Discord](https://img.shields.io/badge/Discord-Microsoft_Foundry_Community_Discord-blue?style=for-the-badge&logo=discord&color=5865f2&logoColor=fff)](https://aka.ms/foundry/discord) -如果您有產品反饋或構建時發生錯誤，請造訪： +如果您在開發過程中有產品反饋或錯誤，請造訪： [![Microsoft Foundry Developer Forum](https://img.shields.io/badge/GitHub-Microsoft_Foundry_Developer_Forum-blue?style=for-the-badge&logo=github&color=000000&logoColor=fff)](https://aka.ms/foundry/forum) --- -**免責聲明**： -本文件係使用 AI 翻譯服務 [Co-op Translator](https://github.com/Azure/co-op-translator) 翻譯完成。雖然我們力求準確，但請注意自動翻譯可能包含錯誤或不準確之處。原文文件之母語版本應被視為權威來源。關於重要資訊，建議採用專業人工翻譯。我們對因使用本翻譯所產生之任何誤解或曲解概不負責。 +**免責聲明**： +此文件已使用 AI 翻譯服務 [Co-op Translator](https://github.com/Azure/co-op-translator) 進行翻譯。雖然我們努力追求準確性，但請注意自動翻譯可能包含錯誤或不準確之處。原始文件的母語版本應視為權威來源。對於關鍵資訊，建議採用專業人工翻譯。我們不對因使用此翻譯所產生的任何誤解或誤譯承擔責任。 \ No newline at end of file diff --git a/translations/zh-TW/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb b/translations/zh-TW/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb new file mode 100644 index 000000000..e346aca26 --- /dev/null +++ b/translations/zh-TW/code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb @@ -0,0 +1,244 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# 🚀 使用 Microsoft phi-4-mini 與 SQLite FTS5 的混合式 RAG（零雲端 SLM）\n", + "\n", + "> **作者：** Çağrı Giray Keşan ([@Cagrik34](https://github.com/Cagrik34)) \n", + "> **焦點：** 小型語言模型（SLMs）、SQLite FTS5 BM25、密集向量嵌入、互惠排名融合（RRF）\n", + "\n", + "---\n", + "\n", + "## 📌 1. 動機：本地 SLM 中的關鍵字召回困境\n", + "純依賴密集向量嵌入的標準 RAG 架構常常無法精確檢索數字標記（例如，`2,340,000 TL`、合約代碼、帳號）。 \n", + "相反地，稀疏詞彙檢索（BM25）則容易錯過語義同義詞與改寫問題。\n", + "\n", + "本食譜示範如何實作一個高速、記憶體內混合檢索引擎，結合：\n", + "1. 密集向量（餘弦相似度）\n", + "2. 稀疏詞彙檢索（SQLite FTS5 BM25）\n", + "3. **互惠排名融合（RRF，$k=60$）**\n", + "4. **基於事實的引用生成（`[1]`, `[2]`）**，使用 Microsoft `phi-4-mini`。\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import sqlite3\n", + "import numpy as np\n", + "from typing import List, Tuple, Dict, Any\n", + "\n", + "RRF_K = 60\n", + "TOP_K = 2\n", + "print(\"✅ Core dependencies loaded successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 🏗️ 2. 雙 SQLite 架構（稠密向量 + 虛擬 FTS5 BM25 表）\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "class LocalHybridRAGStore:\n", + " def __init__(self, db_path: str = \":memory:\"):\n", + " self.conn = sqlite3.connect(db_path)\n", + " self._init_schema()\n", + "\n", + " def _init_schema(self) -> None:\n", + " with self.conn:\n", + " self.conn.execute(\"\"\"\n", + " CREATE TABLE IF NOT EXISTS document_chunks (\n", + " id INTEGER PRIMARY KEY AUTOINCREMENT,\n", + " source_file TEXT NOT NULL,\n", + " chunk_index INTEGER NOT NULL,\n", + " content TEXT NOT NULL,\n", + " embedding BLOB NOT NULL\n", + " )\n", + " \"\"\")\n", + " self.conn.execute(\"\"\"\n", + " CREATE VIRTUAL TABLE IF NOT EXISTS document_chunks_fts USING fts5(\n", + " content,\n", + " source_file UNINDEXED,\n", + " chunk_index UNINDEXED,\n", + " tokenize='unicode61'\n", + " )\n", + " \"\"\")\n", + "\n", + " def insert_chunk(self, source_file: str, chunk_index: int, content: str, embedding: List[float]) -> None:\n", + " vec = np.array(embedding, dtype=np.float32)\n", + " norm = np.linalg.norm(vec)\n", + " if norm > 0:\n", + " vec = vec / norm\n", + "\n", + " with self.conn:\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks (source_file, chunk_index, content, embedding) VALUES (?, ?, ?, ?)\",\n", + " (source_file, chunk_index, content, vec.tobytes())\n", + " )\n", + " self.conn.execute(\n", + " \"INSERT INTO document_chunks_fts (content, source_file, chunk_index) VALUES (?, ?, ?)\",\n", + " (content, source_file, str(chunk_index))\n", + " )\n", + "\n", + " def search_dense(self, query_embedding: List[float], top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " q_vec = np.array(query_embedding, dtype=np.float32)\n", + " q_norm = np.linalg.norm(q_vec)\n", + " if q_norm > 0:\n", + " q_vec = q_vec / q_norm\n", + "\n", + " cursor = self.conn.execute(\"SELECT id, source_file, content, embedding FROM document_chunks\")\n", + " results = []\n", + " for doc_id, src, content, blob in cursor.fetchall():\n", + " doc_vec = np.frombuffer(blob, dtype=np.float32)\n", + " similarity = float(np.dot(q_vec, doc_vec))\n", + " results.append((doc_id, src, content, similarity))\n", + " results.sort(key=lambda x: x[3], reverse=True)\n", + " return results[:top_k]\n", + "\n", + " def search_sparse_bm25(self, query_text: str, top_k: int = 5) -> List[Tuple[int, str, str, float]]:\n", + " clean_tokens = [t for t in query_text.replace(\"'\", \"\").replace('\"', '').split() if len(t) > 1]\n", + " if not clean_tokens:\n", + " return []\n", + " fts_query = \" OR \".join(f'\"{t}\"' for t in clean_tokens)\n", + " cursor = self.conn.execute(\n", + " \"SELECT rowid, source_file, content, rank FROM document_chunks_fts WHERE document_chunks_fts MATCH ? ORDER BY rank LIMIT ?\",\n", + " (fts_query, top_k)\n", + " )\n", + " results = []\n", + " for doc_id, src, content, bm25_rank in cursor.fetchall():\n", + " bm25_score = 1.0 / (1.0 + abs(float(bm25_rank)))\n", + " results.append((doc_id, src, content, bm25_score))\n", + " return results\n", + "\n", + " def hybrid_search(self, query_text: str, query_embedding: List[float], top_k: int = TOP_K) -> List[Dict[str, Any]]:\n", + " dense_hits = self.search_dense(query_embedding, top_k=10)\n", + " sparse_hits = self.search_sparse_bm25(query_text, top_k=10)\n", + " fused_scores = {}\n", + " chunk_map = {}\n", + "\n", + " for rank, (doc_id, src, content, sim) in enumerate(dense_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " chunk_map[key] = (src, content, \"vector\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " for rank, (doc_id, src, content, bm25) in enumerate(sparse_hits, start=1):\n", + " key = f\"{src}::{content[:50]}\"\n", + " if key not in chunk_map:\n", + " chunk_map[key] = (src, content, \"bm25\")\n", + " else:\n", + " chunk_map[key] = (src, content, \"hybrid\")\n", + " fused_scores[key] = fused_scores.get(key, 0.0) + (1.0 / (RRF_K + rank))\n", + "\n", + " sorted_keys = sorted(fused_scores.keys(), key=lambda k: fused_scores[k], reverse=True)[:top_k]\n", + " output = []\n", + " for citation_idx, key in enumerate(sorted_keys, start=1):\n", + " src, content, match_type = chunk_map[key]\n", + " output.append({\n", + " \"citation_index\": citation_idx,\n", + " \"source_file\": src,\n", + " \"content\": content,\n", + " \"rrf_score\": fused_scores[key],\n", + " \"match_type\": match_type\n", + " })\n", + " return output\n", + "\n", + "print(\"✅ LocalHybridRAGStore class compiled successfully.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📊 3. 範例資料匯入與執行效能基準測試\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "store = LocalHybridRAGStore()\n", + "\n", + "sample_docs = [\n", + " (\"q3_financial_report.pdf\", 0, \"CodePulse engineering project total Q3 budget was allocated at 2,340,000 TL with 15 active developers.\", [0.8, 0.1, 0.2] + [0.0] * 1021),\n", + " (\"architecture_specs.md\", 0, \"Zenith AI leverages Microsoft phi-4-mini (3.8B parameters) for local zero-cloud inference.\", [0.2, 0.9, 0.1] + [0.0] * 1021),\n", + " (\"hr_policy_2026.docx\", 0, \"Remote work expense allowance is capped at 15,000 TL per employee quarterly.\", [0.1, 0.1, 0.8] + [0.0] * 1021)\n", + "]\n", + "\n", + "for src, idx, content, emb in sample_docs:\n", + " store.insert_chunk(src, idx, content, emb)\n", + "\n", + "query = \"What is the total allocated budget for the CodePulse project?\"\n", + "query_vec = [0.75, 0.15, 0.25] + [0.0] * 1021\n", + "\n", + "results = store.hybrid_search(query, query_vec, top_k=2)\n", + "for res in results:\n", + " print(f\"[{res['citation_index']}] {res['source_file']} ({res['match_type'].upper()}) -> Score: {res['rrf_score']:.4f}\")\n", + " print(f\" Content: {res['content']}\\n\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## 📝 4. 為 Microsoft phi-4-mini 制定具根據性的提示語\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "def construct_grounded_prompt(query: str, retrieved_chunks: List[Dict[str, Any]]) -> str:\n", + " context_blocks = []\n", + " for chunk in retrieved_chunks:\n", + " context_blocks.append(f\"[{chunk['citation_index']}] (Source: {chunk['source_file']})\\n{chunk['content']}\")\n", + " context_str = \"\\n\\n\".join(context_blocks)\n", + "\n", + " return f\"\"\"You are Zenith AI, an enterprise-grade local assistant.\n", + "Answer the user query strictly based on the provided context below.\n", + "Every factual claim must cite its source index like [1] or [2].\n", + "If the context does not contain the answer, respond: 'This information is not present in the indexed documents.'\n", + "\n", + "--- CONTEXT ---\n", + "{context_str}\n", + "--- END CONTEXT ---\n", + "\n", + "User Query: {query}\n", + "Answer:\"\"\"\n", + "\n", + "prompt = construct_grounded_prompt(query, results)\n", + "print(prompt)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "---\n\n\n**免責聲明**：\n此文件已使用 AI 翻譯服務 [Co-op Translator](https://github.com/Azure/co-op-translator) 進行翻譯。雖然我們努力追求準確性，但請注意自動翻譯可能包含錯誤或不準確之處。原始文件的母語版本應視為權威來源。對於關鍵資訊，建議採用專業人工翻譯。我們不對因使用此翻譯所產生的任何誤解或誤譯承擔責任。\n\n" + ] + } + ], + "metadata": { + "language_info": { + "name": "python", + "version": "3.11.0" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} \ No newline at end of file