安装完成后测试出现报错

[ckcfcc]

用 CMAKE_ARGS="-DGGML_CUDA=on" pip install --no-cache-dir git+https://github.com/JamePeng/llama-cpp-python.git
安装完成后，执行测试脚本出现：

(comfy) xxx@cf-ai:~$ python -c "from llama_cpp.llama_chat_format import Gemma4ChatHandler; print('OK')"
[llama-cpp-python].provided_path: loaded library from /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libggml-base.so
[llama-cpp-python].provided_path: loaded library from /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libggml.so
[llama-cpp-python].provided_path: loaded library from /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_graph_reserve, ?llama_graph_reserve@@YAPEAUggml_cgraph@@PEAUllama_context@@iii@Z, __Z19llama_graph_reserveP13llama_contextjjj
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_ftype_get_default_type, ?llama_ftype_get_default_type@@ya?AW4ggml_type@@W4llama_ftype@@@z, __Z28llama_ftype_get_default_type11llama_ftype
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_model_n_expert, ?llama_model_n_expert@@YAHPEBUllama_model@@@z, __Z20llama_model_n_expertPK11llama_model
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_model_n_devices, ?llama_model_n_devices@@YAHPEBUllama_model@@@z, __Z21llama_model_n_devicesPK11llama_model
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_model_get_device, ?llama_model_get_device@@YAPEAUggml_backend_device@@PEBUllama_model@@h@Z, __Z22llama_model_get_devicePK11llama_modeli
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_set_embeddings_nextn, ?llama_set_embeddings_nextn@@YAXPEAUllama_context@@_N1@Z, __Z26llama_set_embeddings_nextnP13llama_contextbb
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_set_nextn_layer_offset, ?llama_set_nextn_layer_offset@@YAXPEAUllama_context@@h@Z, __Z28llama_set_nextn_layer_offsetP13llama_contexti
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_get_embeddings_nextn, ?llama_get_embeddings_nextn@@YAPEAMPEAUllama_context@@@z, __Z26llama_get_embeddings_nextnP13llama_context
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_get_embeddings_nextn_ith, ?llama_get_embeddings_nextn_ith@@YAPEAMPEAUllama_context@@h@Z, __Z30llama_get_embeddings_nextn_ithP13llama_contexti
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_set_embeddings_layer_inp, ?llama_set_embeddings_layer_inp@@YAXPEAUllama_context@@I_N@Z, __Z30llama_set_embeddings_layer_inpP13llama_contextjb
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_get_embeddings_layer_inp, ?llama_get_embeddings_layer_inp@@YAPEAMPEAUllama_context@@i@Z, __Z30llama_get_embeddings_layer_inpP13llama_contextj
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_get_ctx_other, ?llama_get_ctx_other@@YAPEAUllama_context@@PEAU1@@z, __Z19llama_get_ctx_otherP13llama_context
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_model_target_layer_ids, ?llama_model_target_layer_ids@@YAPEBHPEBUllama_model@@@z, __Z28llama_model_target_layer_idsPK11llama_model
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
[llama-cpp-python].ctypes_function: WARNING! optional API unavailable
symbols: llama_model_target_layer_ids_n, ?llama_model_target_layer_ids_n@@YAIPEBUllama_model@@@z, __Z30llama_model_target_layer_ids_nPK11llama_model
library: /home/xxx/miniconda3/envs/comfy/lib/python3.12/site-packages/llama_cpp/lib/libllama.so
OK

请问这个问题该如何处理？

[ckcfcc]

直接使用 pip install https://github.com/JamePeng/llama-cpp-python/releases/download/v0.3.44-cu128-linux-20260721/llama_cpp_python-0.3.44+cu128-cp312-cp312-linux_x86_64.whl

安装与编辑好的轮子也不行，也是报上面的错。

[JamePeng]

这个是llama-ext.h的可选API找不到ABI命名的警告提示而已，这个库名会受到不同编译器的ABI命名影响，目前测试Windows和macos没问题，github action也没有报错，你是在什么系统下运行的。

其实只是找不到llama-ext.h可选库的提示而已，不会影响到功能使用。

[JamePeng]

用 CMAKE_ARGS="-DGGML_CUDA=on" pip install --no-cache-dir git+https://github.com/JamePeng/llama-cpp-python.git

@ckcfcc 你再试试看，我提交了gcc的abi名称补充