[Kernel] Add NVFP4 MoE CUTLASS support for SM120 (#29242)

Signed-off-by: mgoin <mgoin64@gmail.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
This commit is contained in:
Michael Goin
2025-11-25 06:59:07 -08:00
committed by GitHub
parent dbc3d9991a
commit e502098643
8 changed files with 264 additions and 30 deletions
@@ -1132,6 +1132,10 @@ class ModelOptNvFp4FusedMoE(FusedMoEMethodBase):
f"Using FlashInfer {self.flashinfer_moe_backend.value} kernels"
" for ModelOptNvFp4FusedMoE."
)
elif self.use_marlin:
logger.info_once("Using Marlin for ModelOptNvFp4FusedMoE.")
else:
logger.info_once("Using Cutlass for ModelOptNvFp4FusedMoE.")
def maybe_make_prepare_finalize(
self,