Hugging Face's logo

kernels-community
/
moe-new-models

Kernels
Model card Files
xet
 Community
moe-new-models / torch-ext

Commit History

  1. Sync with vLLM, add `apply_router_weight_on_input` arg to `fused_moe`⁠

    danieldk HF Staff
    1aea118

  1. Module name has to correspond to repo name

    danieldk HF Staff
    a8ef99b

  2. Fix Llama expert layout

    danieldk HF Staff
    3c64348

  3. Add Llama MoE

    danieldk HF Staff
    6d19dd0

  1. Update for build.toml changes

    danieldk HF Staff
    29fbb6a

  1. Vendor `w8a8_block_fp8_matmul` and `per_token_group_quant_fp8`

    danieldk HF Staff
    b41d28a

  2. Normalize some directory names

    danieldk HF Staff
    3d21f81