Pinned Loading

1. open-infra-ai/paged-serving open-infra-ai/paged-serving Public

   Rust LLM Serving 控制面：Paged KV、continuous batching、OpenAI SSE、C ABI 后端与可信压测

   Rust 1

2. open-infra-ai/cuflash open-infra-ai/cuflash Public

   cuflash — FlashAttention from scratch in CUDA C++: scalar → WMMA tensor-core forward, FlashDecoding/Split-KV, Roofline analysis. FP16/BF16/FP32 fwd/bwd, sm_70–sm_90.

   Cuda

3. open-infra-ai/tiny-llm open-infra-ai/tiny-llm Public

   CUDA/C++ LLM 推理运行时：GGUF、W8A16、tokenizer、Paged KV、CUDA Graph 与可复现基准

   C++