Commit History

  1. Cache causal mask for faster inference

    Verified signature
    89ad9ef

  1. Fix completion_tokens: count tokens not decoded characters

    Verified signature
    80c1b86

  1. Switch to HTTP.serve (non-streaming handler) for proxy compatibility

    Verified signature
    d353df1
  2. Fix: add Content-Length for non-streaming responses, X-Accel-Buffering for SSE

    Verified signature
    6c7eeb4
  3. Upload server.jl with huggingface_hub

    Verified signature
    10ad6b5
  4. Upload folder using huggingface_hub

    Verified signature
    18db312
  5. initial commit

    Verified signature
    4b77973