[NVIDIA] Add SM100 Flashinfer Cutlass MoE fp8 backend (#22357)

Signed-off-by: Amir Klein <203507526+amirkl94@users.noreply.github.com>
This commit is contained in:
amirkl94
2025-08-19 18:01:53 -04:00
committed by GitHub
parent 21dce80ea9
commit a38b8af4c3
6 changed files with 613 additions and 139 deletions
+137 -84
View File
@@ -9,6 +9,7 @@ from torch.nn import Module
from torch.nn.parameter import Parameter
import vllm.envs as envs
import vllm.model_executor.layers.fused_moe.modular_kernel as mk
from vllm import _custom_ops as ops
from vllm.distributed import get_tensor_model_parallel_world_size
from vllm.logger import init_logger
@@ -23,8 +24,11 @@ from vllm.model_executor.layers.quantization.base_config import (
QuantizationConfig, QuantizeMethodBase)
from vllm.model_executor.layers.quantization.kv_cache import BaseKVCacheMethod
from vllm.model_executor.layers.quantization.utils.flashinfer_utils import (
apply_flashinfer_per_tensor_scale_fp8, register_moe_scaling_factors,
rotate_flashinfer_fp8_moe_weights, swap_w13_to_w31)
FlashinferMoeBackend, apply_flashinfer_per_tensor_scale_fp8,
build_flashinfer_fp8_cutlass_moe_prepare_finalize,
flashinfer_cutlass_moe_fp8, get_flashinfer_moe_backend,
register_moe_scaling_factors, rotate_flashinfer_fp8_moe_weights,
select_cutlass_fp8_gemm_impl, swap_w13_to_w31)
from vllm.model_executor.layers.quantization.utils.fp8_utils import (
get_col_major_tma_aligned_tensor, requant_weight_ue8m0_inplace)
from vllm.model_executor.layers.quantization.utils.marlin_utils_fp8 import (
@@ -145,7 +149,7 @@ class Fp8Config(QuantizationConfig):
return UnquantizedLinearMethod()
return Fp8LinearMethod(self)
elif isinstance(layer, FusedMoE):
return Fp8MoEMethod(self, layer.moe_config)
return Fp8MoEMethod(self, layer)
elif isinstance(layer, Attention):
return Fp8KVCacheMethod(self)
return None
@@ -482,16 +486,20 @@ class Fp8MoEMethod(FusedMoEMethodBase):
quant_config: The quantization config.
"""
def __init__(self, quant_config: Fp8Config, moe: FusedMoEConfig):
super().__init__(moe)
def __init__(self, quant_config: Fp8Config, layer: torch.nn.Module):
super().__init__(layer.moe_config)
self.layer = layer
self.quant_config = quant_config
self.block_quant = self.quant_config.weight_block_size is not None
self.flashinfer_moe_enabled = False
self.flashinfer_moe_backend: Optional[FlashinferMoeBackend] = None
self.fused_experts: Optional[
mk.FusedMoEModularKernel] = None # type: ignore
if envs.VLLM_USE_FLASHINFER_MOE_FP8 and has_flashinfer_moe():
self.flashinfer_moe_backend = get_flashinfer_moe_backend()
logger.info_once(
"Using FlashInfer MoE FP8 kernels for Fp8MoEMethod.")
self.flashinfer_moe_enabled = True
f"Using FlashInfer {self.flashinfer_moe_backend.value} kernels"
)
# For GPUs that lack FP8 hardware support, we can leverage the Marlin
# kernel for fast weight-only FP8 quantization
self.use_marlin = (not current_platform.has_device_capability(89)
@@ -531,6 +539,20 @@ class Fp8MoEMethod(FusedMoEMethodBase):
"CutlassBlockScaledGroupedGemm not supported on the current "
"platform.")
def maybe_make_prepare_finalize(
self,
moe: FusedMoEConfig,
) -> Optional[mk.FusedMoEPrepareAndFinalize]:
if self.flashinfer_moe_backend != FlashinferMoeBackend.CUTLASS:
return super().maybe_make_prepare_finalize(moe)
prepare_finalize = build_flashinfer_fp8_cutlass_moe_prepare_finalize(
moe,
layer=self.layer,
)
logger.debug_once("%s", prepare_finalize.__class__.__name__)
return prepare_finalize
def create_weights(self, layer: Module, num_experts: int, hidden_size: int,
intermediate_size_per_partition: int,
params_dtype: torch.dtype, **extra_weight_attrs):
@@ -678,7 +700,7 @@ class Fp8MoEMethod(FusedMoEMethodBase):
normalize_e4m3fn_to_e4m3fnuz(
layer.w2_weight, layer.w2_weight_scale_inv,
layer.w2_input_scale)
elif self.flashinfer_moe_enabled:
elif self.flashinfer_moe_backend is not None:
# NOTE: weights have to be swapped since the activation is
# applied on different half for flashinfer vs vllm
w13_weight = swap_w13_to_w31(layer.w13_weight.data)
@@ -686,9 +708,6 @@ class Fp8MoEMethod(FusedMoEMethodBase):
layer.w13_weight_scale_inv.data)
w2_weight = layer.w2_weight.data
w2_weight_scale_inv = layer.w2_weight_scale_inv.data
if not self.block_quant:
register_moe_scaling_factors(layer)
rotate_flashinfer_fp8_moe_weights(w13_weight, w2_weight)
else:
w13_weight = layer.w13_weight.data
w13_weight_scale_inv = layer.w13_weight_scale_inv.data
@@ -834,6 +853,17 @@ class Fp8MoEMethod(FusedMoEMethodBase):
layer.w13_weight_scale = torch.nn.Parameter(max_w13_scales,
requires_grad=False)
if self.flashinfer_moe_backend is not None:
# NOTE: weights have to be swapped since the activation is
# applied on different half for flashinfer vs vllm
assert not self.block_quant
register_moe_scaling_factors(layer)
w13_weight = swap_w13_to_w31(layer.w13_weight.data)
if self.flashinfer_moe_backend == \
FlashinferMoeBackend.TENSORRT_LLM:
rotate_flashinfer_fp8_moe_weights(w13_weight, w2_weight)
layer.w13_weight.data = w13_weight.data
if self.use_marlin:
prepare_moe_fp8_layer_for_marlin(layer, False)
# Activations not quantized for marlin.
@@ -892,6 +922,13 @@ class Fp8MoEMethod(FusedMoEMethodBase):
per_act_token_quant=False,
allow_deep_gemm=self.allow_deep_gemm,
)
elif self.flashinfer_moe_backend == FlashinferMoeBackend.CUTLASS:
experts = select_cutlass_fp8_gemm_impl(
moe,
self.layer,
)
logger.debug_once("Using %s", experts.__class__.__name__)
return experts
else:
logger.debug(
"TritonOrDeepGemmExperts(%s): block_size=%s, per_act_token=%s",
@@ -930,25 +967,66 @@ class Fp8MoEMethod(FusedMoEMethodBase):
assert logical_to_physical_map is not None
assert logical_replica_count is not None
assert isinstance(layer, FusedMoE)
if not self.flashinfer_moe_enabled:
topk_weights, topk_ids = FusedMoE.select_experts(
hidden_states=x,
router_logits=router_logits,
use_grouped_topk=use_grouped_topk,
top_k=top_k,
renormalize=renormalize,
topk_group=topk_group,
num_expert_group=num_expert_group,
custom_routing_function=custom_routing_function,
scoring_func=scoring_func,
e_score_correction_bias=e_score_correction_bias,
indices_type=self.topk_indices_dtype,
enable_eplb=enable_eplb,
expert_map=expert_map,
expert_load_view=expert_load_view,
logical_to_physical_map=logical_to_physical_map,
logical_replica_count=logical_replica_count,
)
if self.flashinfer_moe_backend == FlashinferMoeBackend.TENSORRT_LLM:
assert activation == 'silu', (
f"Expected 'silu' activation but got {activation}")
assert scoring_func == 'sigmoid', (
f"Expected 'sigmoid' scoring func but got {scoring_func}")
if self.block_quant:
assert (renormalize and use_grouped_topk
and custom_routing_function is None)
return torch.ops.vllm.flashinfer_fused_moe_blockscale_fp8(
routing_logits=router_logits.to(torch.float32),
routing_bias=e_score_correction_bias,
x=x,
w13_weight=layer.w13_weight,
w13_weight_scale_inv=layer.w13_weight_scale_inv,
w2_weight=layer.w2_weight,
w2_weight_scale_inv=layer.w2_weight_scale_inv,
global_num_experts=global_num_experts,
top_k=top_k,
num_expert_group=num_expert_group,
topk_group=topk_group,
intermediate_size=layer.intermediate_size_per_partition,
expert_offset=layer.ep_rank * layer.local_num_experts,
local_num_experts=layer.local_num_experts,
block_shape=self.quant_config.weight_block_size,
routed_scaling=1.0,
)
else:
assert (not renormalize
and custom_routing_function is not None)
return apply_flashinfer_per_tensor_scale_fp8(
layer=layer,
hidden_states=x,
router_logits=router_logits,
routing_bias=e_score_correction_bias,
global_num_experts=global_num_experts,
top_k=top_k,
num_expert_group=num_expert_group,
topk_group=topk_group,
apply_router_weight_on_input=apply_router_weight_on_input)
topk_weights, topk_ids = FusedMoE.select_experts(
hidden_states=x,
router_logits=router_logits,
use_grouped_topk=use_grouped_topk,
top_k=top_k,
renormalize=renormalize,
topk_group=topk_group,
num_expert_group=num_expert_group,
custom_routing_function=custom_routing_function,
scoring_func=scoring_func,
e_score_correction_bias=e_score_correction_bias,
indices_type=self.topk_indices_dtype,
enable_eplb=enable_eplb,
expert_map=expert_map,
expert_load_view=expert_load_view,
logical_to_physical_map=logical_to_physical_map,
logical_replica_count=logical_replica_count,
)
if self.rocm_aiter_moe_enabled:
from vllm.model_executor.layers.fused_moe.rocm_aiter_fused_moe import ( # noqa: E501
@@ -988,63 +1066,38 @@ class Fp8MoEMethod(FusedMoEMethodBase):
apply_router_weight_on_input=apply_router_weight_on_input,
global_num_experts=global_num_experts,
expert_map=expert_map)
elif self.flashinfer_moe_enabled:
assert activation == 'silu'
assert scoring_func == 'sigmoid'
if self.block_quant:
assert (renormalize and use_grouped_topk
and custom_routing_function is None)
return torch.ops.vllm.flashinfer_fused_moe_blockscale_fp8(
routing_logits=router_logits.to(torch.float32),
routing_bias=e_score_correction_bias,
x=x,
w13_weight=layer.w13_weight,
w13_weight_scale_inv=layer.w13_weight_scale_inv,
w2_weight=layer.w2_weight,
w2_weight_scale_inv=layer.w2_weight_scale_inv,
elif self.flashinfer_moe_backend == FlashinferMoeBackend.CUTLASS:
assert self.block_quant is None
assert (not renormalize and custom_routing_function is not None)
assert activation == 'silu', (
f"Expected 'silu' activation but got {activation}")
assert scoring_func == 'sigmoid', (
f"Expected 'sigmoid' scoring func but got {scoring_func}")
if self.fused_experts is not None:
return self.fused_experts(
x,
layer.w13_weight,
layer.w2_weight,
topk_weights,
topk_ids,
inplace=False,
activation=activation,
global_num_experts=global_num_experts,
top_k=top_k,
num_expert_group=num_expert_group,
topk_group=topk_group,
intermediate_size=layer.intermediate_size_per_partition,
expert_offset=layer.ep_rank * layer.local_num_experts,
local_num_experts=layer.local_num_experts,
block_shape=self.quant_config.weight_block_size,
routed_scaling=1.0,
expert_map=expert_map,
apply_router_weight_on_input=apply_router_weight_on_input,
)
else:
assert (not renormalize
and custom_routing_function is not None)
return apply_flashinfer_per_tensor_scale_fp8(
layer=layer,
hidden_states=x,
router_logits=router_logits,
routing_bias=e_score_correction_bias,
return flashinfer_cutlass_moe_fp8(
x,
layer,
topk_weights,
topk_ids,
inplace=False,
activation=activation,
global_num_experts=global_num_experts,
top_k=top_k,
num_expert_group=num_expert_group,
topk_group=topk_group,
apply_router_weight_on_input=apply_router_weight_on_input)
elif self.fused_experts is not None:
return self.fused_experts(
hidden_states=x,
w1=layer.w13_weight,
w2=layer.w2_weight,
topk_weights=topk_weights,
topk_ids=topk_ids,
inplace=True,
activation=activation,
global_num_experts=global_num_experts,
apply_router_weight_on_input=apply_router_weight_on_input,
expert_map=expert_map,
w1_scale=(layer.w13_weight_scale_inv
if self.block_quant else layer.w13_weight_scale),
w2_scale=(layer.w2_weight_scale_inv
if self.block_quant else layer.w2_weight_scale),
a1_scale=layer.w13_input_scale,
a2_scale=layer.w2_input_scale,
)
expert_map=expert_map,
apply_router_weight_on_input=apply_router_weight_on_input,
)
else:
from vllm.model_executor.layers.fused_moe import fused_experts
return fused_experts(