[Quantization] Add FlashInfer CuteDSL batched experts backend for NVFP4 MoE (#38251)

Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Roger Wang <hey@rogerw.io>
This commit is contained in:
Yongye Zhu
2026-04-06 11:57:53 -07:00
committed by GitHub
co-authored by Michael Goin Roger Wang
parent 94fbb09894
commit e8ebbdde83
6 changed files with 574 additions and 245 deletions
+1 -1
View File
@@ -17,7 +17,7 @@ from flashinfer import fp4_quantize
from torch.nn import functional as F
from vllm.model_executor.layers.activation import SiluAndMul
from vllm.model_executor.layers.fused_moe.experts.flashinfer_cutedsl_moe import (
from vllm.model_executor.layers.fused_moe.experts.flashinfer_cutedsl_batched_moe import ( # noqa: E501
flashinfer_cutedsl_moe_masked,
)
from vllm.utils.flashinfer import (