Commit History

2 days ago

  1. experts fp8 with 128x128 blocks (grouped kernel), attention 32x32

    ArthurZ HF Staff
    Verified signature
    3a23a8f
  2. every fp8 dim spans at least 2 blocks (grouped kernel)

    ArthurZ HF Staff
    Verified signature
    bc77947
  3. fp8 blocks 32x32 (the triton kernel needs K >= 32)

    ArthurZ HF Staff
    Verified signature
    4ec61d8
  4. tiny Mistral4 with fp8 + nvfp4 layers

    ArthurZ HF Staff
    Verified signature
    ac53c7a