Alexander Korolev
57e6336047
mxfp4 recipes: --linear-backend b12x (native SM120 GEMM) - the flashinfer mxfp4 linear kernel wrongly claims sm_12x support and crashes in cute-dsl mm_fp4; --moe-backend is inert for the compressed-tensors MoE scheme (it picks MarlinExperts itself)
|
1 місяць тому | |
|---|---|---|
| .. | ||
| recipe.toml | 1 місяць тому | |