Skip to content

Would an interactive vLLM learning resource fit this inference list? #196

Description

@lora-sys

Hi! I maintain nano-vLLM Interactive Guide, a MIT-licensed, browser-based learning project for understanding vLLM-style inference internals.

It provides Chinese-first concept labs, source anchors, teaching traces, and an optional CUDA runtime path for:

  • PagedAttention and KV-block allocation
  • continuous batching and scheduler decisions
  • Prefix Cache reuse boundaries
  • prefill/decode behavior, sampling, tensor parallelism, CUDA Graphs, and benchmark evidence boundaries

Repository: https://github.com/lora-sys/nano-vllm-interactive-guide
Live guide: https://lora-sys.github.io/nano-vllm-interactive-guide/
English discovery page: https://lora-sys.github.io/nano-vllm-interactive-guide/en/

The project explicitly distinguishes concept simulations from real CUDA traces and benchmark claims. Since this list is strongly focused on inference papers and code, would a single entry in a Learning Resources or Tutorials subsection be in scope? If not, no action is needed; I would prefer to follow the list's intended taxonomy.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions