Support mnnvl all2allv from Flashinfer (#21003)

Signed-off-by: Shu Wang <shuw@nvidia.com>
Signed-off-by: Shu Wang. <shuw@nvidia.com>
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
This commit is contained in:
Shu Wang
2025-09-24 14:38:16 -04:00
committed by GitHub
co-authored by Tyler Michael Smith Tyler Michael Smith
parent 2dda3e35d0
commit 54e42b72db
10 changed files with 410 additions and 40 deletions
@@ -12,7 +12,7 @@ from vllm.model_executor.layers.fused_moe.config import (FusedMoEConfig,
from vllm.model_executor.layers.fused_moe.flashinfer_cutlass_moe import (
FlashInferExperts)
from vllm.model_executor.layers.fused_moe.flashinfer_cutlass_prepare_finalize import ( # noqa: E501
FlashInferCutlassMoEPrepareAndFinalize)
create_flashinfer_prepare_finalize)
from vllm.platforms import current_platform
from vllm.utils.flashinfer import has_flashinfer_cutlass_fused_moe
@@ -51,7 +51,9 @@ def build_flashinfer_fp4_cutlass_moe_prepare_finalize(
moe: FusedMoEConfig) -> mk.FusedMoEPrepareAndFinalize:
"""Create a FlashInfer CUTLASS fused-MoE prepare finalize kernel"""
use_dp = moe.moe_parallel_config.dp_size > 1
return FlashInferCutlassMoEPrepareAndFinalize(use_dp)
enable_alltoallv = envs.VLLM_ALL2ALL_BACKEND == "flashinfer_all2allv"
return create_flashinfer_prepare_finalize(
use_dp=use_dp, use_nvfp4=True, enable_alltoallv=enable_alltoallv)
def select_nvfp4_gemm_impl(
@@ -13,7 +13,7 @@ from vllm.model_executor.layers.fused_moe.config import (FusedMoEConfig,
from vllm.model_executor.layers.fused_moe.flashinfer_cutlass_moe import (
FlashInferExperts)
from vllm.model_executor.layers.fused_moe.flashinfer_cutlass_prepare_finalize import ( # noqa: E501
FlashInferCutlassMoEPrepareAndFinalize)
create_flashinfer_prepare_finalize)
logger = init_logger(__name__)
@@ -173,7 +173,7 @@ def build_flashinfer_fp8_cutlass_moe_prepare_finalize(
moe: Optional[FusedMoEConfig], ) -> mk.FusedMoEPrepareAndFinalize:
"""Create a FlashInfer CUTLASS fused-MoE prepare finalize kernel"""
use_dp = moe.moe_parallel_config.dp_size > 1 if moe is not None else False
return FlashInferCutlassMoEPrepareAndFinalize(use_dp)
return create_flashinfer_prepare_finalize(use_dp)
def select_cutlass_fp8_gemm_impl(