Files
sglang/python/sglang/srt/layers/quantization/modelopt_quant.py
2025-12-18 11:11:42 +08:00

1769 lines
69 KiB
Python
Executable File

# Adapted from https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/layers/quantization/modelopt.py
from __future__ import annotations
import logging
from enum import IntEnum
from typing import TYPE_CHECKING, Any, Dict, List, Optional
import torch
from torch.nn.parameter import Parameter
from sglang.srt.distributed import get_tp_group
from sglang.srt.distributed.device_communicators.pynccl_allocator import (
use_symmetric_memory,
)
from sglang.srt.environ import envs
from sglang.srt.layers.dp_attention import is_allocation_symmetric
from sglang.srt.layers.moe import (
MoeRunner,
MoeRunnerBackend,
MoeRunnerConfig,
get_moe_runner_backend,
)
from sglang.srt.layers.moe.cutlass_moe_params import CutlassMoEParams, CutlassMoEType
from sglang.srt.layers.moe.moe_runner.triton import TritonMoeQuantInfo
from sglang.srt.layers.moe.utils import should_use_flashinfer_cutlass_moe_fp4_allgather
from sglang.srt.layers.parameter import ModelWeightParameter, PerTensorScaleParameter
from sglang.srt.layers.quantization.base_config import (
FusedMoEMethodBase,
LinearMethodBase,
QuantizationConfig,
QuantizeMethodBase,
)
from sglang.srt.layers.quantization.fp8_kernel import scaled_fp8_quant
from sglang.srt.layers.quantization.fp8_utils import (
apply_fp8_linear,
cutlass_fp8_supported,
is_blackwell_supported,
)
from sglang.srt.layers.quantization.kv_cache import BaseKVCacheMethod
from sglang.srt.layers.quantization.unquant import UnquantizedLinearMethod
from sglang.srt.layers.quantization.utils import (
convert_to_channelwise,
is_layer_skipped,
per_tensor_dequantize,
prepare_static_weights_for_trtllm_fp4_moe,
requantize_with_max_scale,
swizzle_blockscale,
)
from sglang.srt.layers.radix_attention import RadixAttention
from sglang.srt.utils.common import (
get_bool_env_var,
is_cuda,
is_sm120_supported,
next_power_of_2,
)
from sglang.srt.utils.patch_torch import register_fake_if_exists
if TYPE_CHECKING:
from sglang.srt.batch_overlap.single_batch_overlap import DownGemmOverlapArgs
from sglang.srt.layers.moe.fused_moe_triton.layer import FusedMoE
from sglang.srt.layers.moe.token_dispatcher import (
CombineInput,
StandardDispatchOutput,
)
try:
if is_sm120_supported():
from flashinfer import fp4_quantize
else:
from sgl_kernel import scaled_fp4_quant as fp4_quantize
except ImportError:
fp4_quantize = None
try:
from flashinfer import mm_fp4 as fp4_gemm
from flashinfer import reorder_rows_for_gated_act_gemm, shuffle_matrix_sf_a
enable_flashinfer_fp4_gemm = True
except ImportError:
if is_cuda():
from sgl_kernel import cutlass_scaled_fp4_mm as fp4_gemm
else:
fp4_gemm = None
enable_flashinfer_fp4_gemm = False
reorder_rows_for_gated_act_gemm = None
shuffle_matrix_a = None
shuffle_matrix_sf_a = None
try:
from flashinfer.fused_moe import cutlass_fused_moe as flashinfer_cutlass_fused_moe
from flashinfer.fused_moe.core import ActivationType
except ImportError:
flashinfer_cutlass_fused_moe = None
# Define a minimal ActivationType enum if flashinfer is not available
class ActivationType(IntEnum):
Swiglu = 3
Relu2 = 6
# Initialize logger for the module
logger = logging.getLogger(__name__)
@torch.library.custom_op("sglang::fp4_gemm", mutates_args=())
def _sglang_fp4_gemm(
input: torch.Tensor,
weight: torch.Tensor,
input_sf: torch.Tensor,
weight_sf: torch.Tensor,
alpha: torch.Tensor,
out_dtype: torch.dtype,
out_features: int,
) -> torch.Tensor:
backend = FLASHINFER_FP4_GEMM_BACKEND if FLASHINFER_FP4_GEMM_BACKEND else "cutlass"
if enable_flashinfer_fp4_gemm:
return fp4_gemm(
input, weight, input_sf, weight_sf, alpha, out_dtype, backend=backend
)
else:
return fp4_gemm(input, weight, input_sf, weight_sf, alpha, out_dtype)
@torch.library.register_fake("sglang::fp4_gemm")
def _sglang_fp4_gemm_fake(
input,
weight,
input_sf,
weight_sf,
alpha,
out_dtype,
out_features: int,
):
M = input.shape[-2]
N = int(out_features)
return input.new_empty((M, N), dtype=out_dtype)
if is_cuda() and (not is_sm120_supported()) and (fp4_quantize is not None):
@register_fake_if_exists("sgl_kernel::scaled_fp4_quant")
def _sgl_kernel_scaled_fp4_quant_fake(
output, input, output_scale, input_global_scale
):
return
CUTEDSL_MOE_SCALAR_INPUT_SCALE = get_bool_env_var(
"SGLANG_CUTEDSL_MOE_SCALAR_INPUT_SCALE", "true"
)
# TODO make it true by default when the DeepEP PR is merged
MOE_NVFP4_DISPATCH = envs.SGLANG_MOE_NVFP4_DISPATCH.get()
FLASHINFER_FP4_GEMM_BACKEND = envs.SGLANG_FLASHINFER_FP4_GEMM_BACKEND.get()
# Supported activation schemes for the current configuration
ACTIVATION_SCHEMES = ["static"]
ACT_STR_TO_TYPE_MAP = {
"silu": ActivationType.Swiglu, # This is the default
"relu2": ActivationType.Relu2,
}
class ModelOptQuantConfig(QuantizationConfig):
def __init__(
self,
kv_cache_quant_algo: Optional[str],
exclude_modules: Optional[List[str]],
packed_modules_mapping: Optional[Dict[str, List[str]]],
):
super().__init__()
self.packed_modules_mapping = packed_modules_mapping
self.exclude_modules = exclude_modules or []
self.kv_cache_quant_algo = kv_cache_quant_algo
def _get_quant_method(
self,
layer: torch.nn.Module,
prefix: str,
*,
Linear: type[LinearMethodBase],
Moe: type[FusedMoEMethodBase],
) -> Optional[QuantizeMethodBase]:
from sglang.srt.layers.linear import LinearBase
from sglang.srt.layers.moe.fused_moe_triton import FusedMoE
if isinstance(layer, LinearBase):
if is_layer_skipped(
prefix, self.exclude_modules, self.packed_modules_mapping
) or self.is_layer_excluded(prefix):
return UnquantizedLinearMethod()
return Linear(self)
elif self.kv_cache_quant_algo and isinstance(layer, RadixAttention):
return ModelOptFp8KVCacheMethod(self)
elif isinstance(layer, FusedMoE):
return Moe(self)
return None
@classmethod
def get_config_filenames(cls) -> List[str]:
return ["hf_quant_config.json"]
def get_scaled_act_names(self) -> List[str]:
return []
class ModelOptFp8Config(ModelOptQuantConfig):
"""Configuration for ModelOpt FP8 quantization, including serialization and compatibility checks."""
def __init__(
self,
is_checkpoint_fp8_serialized: bool = False,
kv_cache_quant_method: Optional[str] = None,
exclude_modules: Optional[List[str]] = None,
packed_modules_mapping: Optional[Dict[str, List[str]]] = None,
) -> None:
"""
Args:
is_checkpoint_fp8_serialized (bool): Indicates if the checkpoint uses serialized FP8 format.
"""
super().__init__(kv_cache_quant_method, exclude_modules, packed_modules_mapping)
self.is_checkpoint_fp8_serialized = is_checkpoint_fp8_serialized
if is_checkpoint_fp8_serialized:
logger.warning(
"Detected ModelOpt FP8 checkpoint. The format is experimental and subject to change."
)
@classmethod
def override_quantization_method(cls, hf_quant_config, user_quant):
"""Override quantization method based on the model's config."""
return cls._modelopt_override_quantization_method(hf_quant_config, user_quant)
@classmethod
def get_name(cls) -> str:
return "modelopt_fp8"
@classmethod
def get_supported_act_dtypes(cls) -> List[torch.dtype]:
return [torch.bfloat16, torch.half]
@classmethod
def get_min_capability(cls) -> int:
return 89 # Minimum hardware capability (e.g., Hopper GPUs).
@classmethod
def from_config(cls, config: Dict[str, Any]) -> ModelOptFp8Config:
# Handle two different config formats:
# 1. hf_quant_config.json format: {"quantization": {"quant_algo": "FP8", ...}}
# 2. config.json quantization_config format: {"quant_algo": "FP8", ...}
# In future modelopt will deprecate hf_quant_config.json, and only keep config.json.
# For legacy reasons, we keep hf_quant_config.json for now.
# Initialize variables
kv_cache_quant_method = None
exclude_modules = None
# Try flat format first (config.json quantization_config - preferred format)
quant_method = config.get("quant_algo")
if quant_method is not None:
# Flat format (config.json quantization_config)
# For kv_cache, check if kv_cache_scheme exists and extract algo
kv_cache_scheme = config.get("kv_cache_scheme")
if (
kv_cache_scheme
and kv_cache_scheme.get("type") == "float"
and kv_cache_scheme.get("num_bits") == 8
):
kv_cache_quant_method = "FP8"
# Map 'ignore' field to 'exclude_modules'
exclude_modules = config.get("ignore")
else:
# Fall back to nested format (hf_quant_config.json - legacy format)
try:
quantization_section = cls.get_from_keys(config, ["quantization"])
quant_method = quantization_section.get("quant_algo")
kv_cache_quant_method = quantization_section.get("kv_cache_quant_algo")
exclude_modules = quantization_section.get("exclude_modules")
except ValueError:
raise ValueError(
"Cannot find 'quant_algo' in the model's quantization config. "
"Expected either flat format (config.json) or nested format (hf_quant_config.json)."
)
if quant_method is None:
raise ValueError(
"Cannot find 'quant_algo' in the model's quantization config. "
)
if "FP8" not in quant_method:
raise ValueError(
"ModelOptFp8Config only supports static FP8 quantization in SGLang. "
"For FP4 quantization, use ModelOptFp4Config. "
"Check the quantization config for your model's configuration."
)
return cls(
is_checkpoint_fp8_serialized=True,
kv_cache_quant_method=kv_cache_quant_method,
exclude_modules=exclude_modules,
packed_modules_mapping=config.get("packed_modules_mapping"),
)
def is_layer_excluded(self, prefix: str) -> bool:
if len(self.exclude_modules) == 0:
return False
return any(
module in prefix
or (
prefix.startswith("language_model.")
and module in prefix.removeprefix("language_model.")
)
for module in self.exclude_modules
)
def get_quant_method(
self, layer: torch.nn.Module, prefix: str
) -> Optional[QuantizeMethodBase]:
return self._get_quant_method(
layer, prefix, Linear=ModelOptFp8LinearMethod, Moe=ModelOptFp8MoEMethod
)
class ModelOptFp8LinearMethod(LinearMethodBase):
"""Linear method for ModelOpt static FP8 quantization.
Supports loading FP8 checkpoints with static weight and activation scales.
Future support may include dynamic scales.
**Limitations**:
1. Only supports per-tensor quantization due to `torch._scaled_mm` limitations.
2. Only supports the `float8_e4m3fn` data type.
Args:
quant_config (ModelOptFp8Config): The ModelOpt quantization configuration.
"""
def __init__(self, quant_config: ModelOptFp8Config):
super().__init__()
self.quant_config = quant_config
self.cutlass_fp8_supported = cutlass_fp8_supported()
def create_weights(
self,
layer: torch.nn.Module,
input_size_per_partition: int,
output_partition_sizes: List[int],
params_dtype: torch.dtype,
**extra_weight_attrs,
) -> None:
"""Creates and registers weights, weight scales, and input scales for FP8 quantization."""
output_size_per_partition = sum(output_partition_sizes)
weight_loader = extra_weight_attrs.get("weight_loader")
weight_dtype = (
torch.float8_e4m3fn
if self.quant_config.is_checkpoint_fp8_serialized
else params_dtype
)
# Set layer attributes
layer.logical_widths = output_partition_sizes
layer.input_size_per_partition = input_size_per_partition
layer.output_size_per_partition = output_size_per_partition
# Register weight
layer.register_parameter(
"weight",
ModelWeightParameter(
data=torch.empty(
output_size_per_partition,
input_size_per_partition,
dtype=weight_dtype,
),
input_dim=1,
output_dim=0,
weight_loader=weight_loader,
),
)
if self.quant_config.is_checkpoint_fp8_serialized:
# Register weight and input scales
for scale_name in ["weight_scale", "input_scale"]:
layer.register_parameter(
scale_name,
PerTensorScaleParameter(
data=torch.full(
(len(output_partition_sizes),),
torch.finfo(torch.float32).min,
dtype=torch.float32,
),
weight_loader=weight_loader,
),
)
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
"""Requantizes weights after loading using the maximum scale."""
max_w_scale, quantized_weight = requantize_with_max_scale(
layer.weight, layer.weight_scale, layer.logical_widths
)
layer.weight = Parameter(quantized_weight.t(), requires_grad=False)
# cutlass sgl-kernel only supports per-channel scale
if self.cutlass_fp8_supported:
max_w_scale = convert_to_channelwise(max_w_scale, layer.logical_widths)
layer.weight_scale = Parameter(max_w_scale, requires_grad=False)
layer.input_scale = Parameter(layer.input_scale.max(), requires_grad=False)
def apply(
self,
layer: torch.nn.Module,
x: torch.Tensor,
bias: Optional[torch.Tensor] = None,
) -> torch.Tensor:
"""Applies FP8 linear transformation."""
return apply_fp8_linear(
input=x,
weight=layer.weight,
weight_scale=layer.weight_scale,
input_scale=layer.input_scale,
bias=bias,
cutlass_fp8_supported=self.cutlass_fp8_supported,
)
class ModelOptFp8KVCacheMethod(BaseKVCacheMethod):
"""
Handles loading FP8 kv-cache scaling factors from modelopt quantized checkpoints.
"""
def __init__(self, quant_config: ModelOptFp8Config):
super().__init__(quant_config)
class ModelOptFp8MoEMethod(FusedMoEMethodBase):
"""MoE method for ModelOpt FP8.
Supports loading FP8 checkpoints with static weight scale and activation scale.
Args:
quant_config: The ModelOpt quantization config.
"""
def __init__(self, quant_config: ModelOptFp8Config):
self.quant_config = quant_config
self.cutlass_fp8_supported = cutlass_fp8_supported()
def create_weights(
self,
layer: torch.nn.Module,
num_experts: int,
hidden_size: int,
intermediate_size_per_partition: int,
params_dtype: torch.dtype,
**extra_weight_attrs,
):
from sglang.srt.layers.moe.fused_moe_triton import FusedMoeWeightScaleSupported
# Use FP8 dtype if checkpoint is serialized, otherwise use the default dtype
weight_dtype = (
torch.float8_e4m3fn
if self.quant_config.is_checkpoint_fp8_serialized
else params_dtype
)
weight_loader = extra_weight_attrs.get("weight_loader")
num_shards = 2 if layer.moe_runner_config.is_gated else 1
intermediate_size = num_shards * intermediate_size_per_partition
w13_weight = ModelWeightParameter(
data=torch.empty(
num_experts,
intermediate_size,
hidden_size,
dtype=weight_dtype,
),
input_dim=2,
output_dim=1,
weight_loader=weight_loader,
)
layer.register_parameter("w13_weight", w13_weight)
w2_weight = ModelWeightParameter(
data=torch.empty(
num_experts,
hidden_size,
intermediate_size_per_partition,
dtype=weight_dtype,
),
input_dim=2,
output_dim=1,
weight_loader=weight_loader,
)
layer.register_parameter("w2_weight", w2_weight)
if self.quant_config.is_checkpoint_fp8_serialized:
# WEIGHT SCALES - Per-tensor scaling for ModelOpts
# Allocate 2 scales for w1 and w3 respectively.
# They will be combined to a single scale after weight loading.
w13_scale_shape = (num_experts, num_shards)
w13_weight_scale = PerTensorScaleParameter(
data=torch.full(
w13_scale_shape,
torch.finfo(torch.float32).min,
dtype=torch.float32,
),
weight_loader=weight_loader,
)
w2_weight_scale = PerTensorScaleParameter(
data=torch.full(
(num_experts,), torch.finfo(torch.float32).min, dtype=torch.float32
),
weight_loader=weight_loader,
)
layer.register_parameter("w13_weight_scale", w13_weight_scale)
layer.register_parameter("w2_weight_scale", w2_weight_scale)
# Set weight loader attributes for scales
extra_weight_attrs.update(
{"quant_method": FusedMoeWeightScaleSupported.TENSOR.value}
)
# INPUT SCALES - Per-tensor scaling for ModelOpt
w13_input_scale = PerTensorScaleParameter(
data=torch.full((num_experts,), 1.0, dtype=torch.float32),
weight_loader=weight_loader,
)
w2_input_scale = PerTensorScaleParameter(
data=torch.full((num_experts,), 1.0, dtype=torch.float32),
weight_loader=weight_loader,
)
layer.register_parameter("w13_input_scale", w13_input_scale)
layer.register_parameter("w2_input_scale", w2_input_scale)
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
"""Process FP8 MoE weights after loading from serialized checkpoint.
Only supports pre-quantized checkpoints with FP8 weights and scales.
"""
layer.w13_weight = Parameter(layer.w13_weight.data, requires_grad=False)
layer.w2_weight = Parameter(layer.w2_weight.data, requires_grad=False)
# Handle scale parameters
if hasattr(layer, "w13_weight_scale") and layer.w13_weight_scale is not None:
# Fp8 moe kernel needs single weight scale for w13 per expert.
# We take the max of the w1 and w3 scales then dequant and requant each expert.
if layer.w13_weight_scale.dim() == 2: # Shape: (num_experts, 2)
# Get the maximum scale across w1 and w3 for each expert
max_w13_scales = layer.w13_weight_scale.max(dim=1).values
# Requantize each expert's weights using the combined scale
# w13_weight has shape (num_experts, 2 * intermediate_size_per_partition, hidden_size)
# where the first intermediate_size_per_partition rows are w1, the next are w3
num_shards = 2 if layer.moe_runner_config.is_gated else 1
intermediate_size_per_partition = (
layer.w13_weight.shape[1] // num_shards
)
for expert_id in range(layer.w13_weight.shape[0]):
start = 0
for shard_id in range(num_shards): # (w1 and w3) or w13
# Dequantize using the original scale for this shard
dq_weight = per_tensor_dequantize(
layer.w13_weight[expert_id][
start : start + intermediate_size_per_partition, :
],
layer.w13_weight_scale[expert_id][shard_id],
)
# Requantize using the combined max scale
(
layer.w13_weight[expert_id][
start : start + intermediate_size_per_partition, :
],
_,
) = scaled_fp8_quant(dq_weight, max_w13_scales[expert_id])
start += intermediate_size_per_partition
# Update the scale parameter to be per-expert instead of per-shard
layer.w13_weight_scale = Parameter(max_w13_scales, requires_grad=False)
else:
layer.w13_weight_scale = Parameter(
layer.w13_weight_scale.data, requires_grad=False
)
if hasattr(layer, "w2_weight_scale") and layer.w2_weight_scale is not None:
layer.w2_weight_scale = Parameter(
layer.w2_weight_scale.data, requires_grad=False
)
if hasattr(layer, "w13_input_scale") and layer.w13_input_scale is not None:
layer.w13_input_scale = Parameter(
layer.w13_input_scale.max(), requires_grad=False
)
if hasattr(layer, "w2_input_scale") and layer.w2_input_scale is not None:
layer.w2_input_scale = Parameter(
layer.w2_input_scale.max(), requires_grad=False
)
# Align FP8 weights to FlashInfer per-tensor kernel layout if enabled
if get_moe_runner_backend().is_flashinfer_trtllm():
from flashinfer import reorder_rows_for_gated_act_gemm, shuffle_matrix_a
# 1) Swap W13 halves: [Up, Gate] -> [Gate, Up] expected by FI
num_experts, two_n, hidden = layer.w13_weight.shape
inter = two_n // 2
w13_swapped = (
layer.w13_weight.reshape(num_experts, 2, inter, hidden)
.flip(dims=[1])
.reshape(num_experts, two_n, hidden)
)
# 2) Reorder rows for fused gated activation (W13)
w13_interleaved = [
reorder_rows_for_gated_act_gemm(w13_swapped[i])
for i in range(num_experts)
]
w13_interleaved = torch.stack(w13_interleaved).reshape(
num_experts, two_n, hidden
)
# 3) Shuffle weights for transposed MMA output (both W13, W2)
epilogue_tile_m = 128
w13_shuffled = [
shuffle_matrix_a(w13_interleaved[i].view(torch.uint8), epilogue_tile_m)
for i in range(num_experts)
]
w2_shuffled = [
shuffle_matrix_a(layer.w2_weight[i].view(torch.uint8), epilogue_tile_m)
for i in range(num_experts)
]
layer.w13_weight = Parameter(
torch.stack(w13_shuffled).view(torch.float8_e4m3fn),
requires_grad=False,
)
layer.w2_weight = Parameter(
torch.stack(w2_shuffled).view(torch.float8_e4m3fn),
requires_grad=False,
)
# Precompute and register per-expert output scaling factors for FI MoE
if get_moe_runner_backend().is_flashinfer_trtllm():
# Note: w13_input_scale and w2_input_scale are scalar Parameters post-reduction
assert (
hasattr(layer, "w13_input_scale") and layer.w13_input_scale is not None
)
assert hasattr(layer, "w2_input_scale") and layer.w2_input_scale is not None
assert (
hasattr(layer, "w13_weight_scale")
and layer.w13_weight_scale is not None
)
assert (
hasattr(layer, "w2_weight_scale") and layer.w2_weight_scale is not None
)
input_scale = layer.w13_input_scale.to(torch.float32)
activation_scale = layer.w2_input_scale.to(torch.float32)
w13_weight_scale = layer.w13_weight_scale.to(torch.float32)
w2_weight_scale = layer.w2_weight_scale.to(torch.float32)
output1_scales_scalar = (
w13_weight_scale * input_scale * (1.0 / activation_scale)
)
output1_scales_gate_scalar = w13_weight_scale * input_scale
output2_scales_scalar = activation_scale * w2_weight_scale
layer.output1_scales_scalar = Parameter(
output1_scales_scalar, requires_grad=False
)
layer.output1_scales_gate_scalar = Parameter(
output1_scales_gate_scalar, requires_grad=False
)
layer.output2_scales_scalar = Parameter(
output2_scales_scalar, requires_grad=False
)
elif get_moe_runner_backend().is_flashinfer_cutlass():
assert (
hasattr(layer, "w13_input_scale") and layer.w13_input_scale is not None
)
assert hasattr(layer, "w2_input_scale") and layer.w2_input_scale is not None
assert (
hasattr(layer, "w13_weight_scale")
and layer.w13_weight_scale is not None
)
assert (
hasattr(layer, "w2_weight_scale") and layer.w2_weight_scale is not None
)
input_scale = layer.w13_input_scale.to(torch.float32)
activation_scale = layer.w2_input_scale.to(torch.float32)
w13_weight_scale = layer.w13_weight_scale.to(torch.float32)
w2_weight_scale = layer.w2_weight_scale.to(torch.float32)
layer.fc1_dequant = Parameter(
w13_weight_scale * input_scale, requires_grad=False
)
layer.fc2_quant = Parameter(
activation_scale.reciprocal(), requires_grad=False
)
layer.fc2_dequant = Parameter(
activation_scale * w2_weight_scale, requires_grad=False
)
layer.fc1_input_dequant = Parameter(input_scale, requires_grad=False)
def create_moe_runner(
self, layer: torch.nn.Module, moe_runner_config: MoeRunnerConfig
):
self.moe_runner_config = moe_runner_config
self.runner = MoeRunner(MoeRunnerBackend.TRITON, moe_runner_config)
def apply(
self,
layer: torch.nn.Module,
dispatch_output: StandardDispatchOutput,
) -> CombineInput:
x = dispatch_output.hidden_states
topk_output = dispatch_output.topk_output
# Fast path: TRT-LLM FP8 per-tensor MoE using BYPASSED TopK routing
from sglang.srt.layers.moe.topk import TopKOutputChecker
if (
get_moe_runner_backend().is_flashinfer_trtllm()
and TopKOutputChecker.format_is_bypassed(topk_output)
):
router_logits = topk_output.router_logits
topk_config = topk_output.topk_config
# Constraints
assert (
self.moe_runner_config.activation == "silu"
), "Only silu is supported for flashinfer fp8 moe"
from flashinfer import RoutingMethodType
from flashinfer.fused_moe import trtllm_fp8_per_tensor_scale_moe
correction_bias = (
None
if topk_config.correction_bias is None
else topk_config.correction_bias
)
# Pre-quantize activations to FP8 per-tensor using provided input scale
x_fp8, _ = scaled_fp8_quant(x, layer.w13_input_scale)
use_routing_scales_on_input = True
routed_scaling_factor = self.moe_runner_config.routed_scaling_factor
# Enforce Llama4 routing for ModelOpt FP8 MoE for now.
# TODO(brayden): support other routing methods
assert topk_config.top_k == 1, "ModelOpt FP8 MoE requires top_k==1"
assert (
not topk_config.num_expert_group
), "ModelOpt FP8 MoE does not support expert grouping"
assert (
not topk_config.topk_group
), "ModelOpt FP8 MoE does not support grouped top-k"
routing_method_type = RoutingMethodType.Llama4
# FlashInfer TRTLLM requires routing_logits (and bias) to be bfloat16
routing_logits_cast = router_logits.to(torch.bfloat16)
routing_bias_cast = (
None if correction_bias is None else correction_bias.to(torch.bfloat16)
)
with use_symmetric_memory(
get_tp_group(), disabled=not is_allocation_symmetric()
):
# FIXME: there is a bug in the trtllm_fp8_block_scale_moe.
# It ignored the `output`` argument. https://github.com/flashinfer-ai/flashinfer/blob/da01b1bd8f9f22aec8c0eea189ad54860b034947/flashinfer/fused_moe/core.py#L1323-L1325
# so we put the whole function under the ``use_symmetric_memory`` context manager.
# If the bug is fixed, we can only put the output tensor allocation under the context manager.
output = trtllm_fp8_per_tensor_scale_moe(
routing_logits=routing_logits_cast,
routing_bias=routing_bias_cast,
hidden_states=x_fp8,
gemm1_weights=layer.w13_weight,
output1_scales_scalar=layer.output1_scales_scalar,
output1_scales_gate_scalar=layer.output1_scales_gate_scalar,
gemm2_weights=layer.w2_weight,
output2_scales_scalar=layer.output2_scales_scalar,
num_experts=layer.num_experts,
top_k=topk_config.top_k,
n_group=0,
topk_group=0,
intermediate_size=layer.w2_weight.shape[2],
local_expert_offset=layer.moe_ep_rank * layer.num_local_experts,
local_num_experts=layer.num_local_experts,
routed_scaling_factor=(
routed_scaling_factor
if routed_scaling_factor is not None
else 1.0
),
use_routing_scales_on_input=use_routing_scales_on_input,
tile_tokens_dim=None,
routing_method_type=routing_method_type,
tune_max_num_tokens=next_power_of_2(x.shape[0]),
)
from sglang.srt.layers.moe.token_dispatcher import StandardCombineInput
return StandardCombineInput(hidden_states=output)
if get_moe_runner_backend().is_flashinfer_cutlass():
activation = ACT_STR_TO_TYPE_MAP[self.moe_runner_config.activation]
assert (
(
activation is ActivationType.Relu2
and not self.moe_runner_config.is_gated
)
or activation is ActivationType.Swiglu
and self.moe_runner_config.is_gated
), "Only Relu2 non-gated or Swiglu gated are supported for flashinfer cutlass fp8 moe"
topk_weights, topk_ids = topk_output.topk_weights, topk_output.topk_ids
x_fp8, _ = scaled_fp8_quant(x, layer.w13_input_scale)
output_dtype = x.dtype
original_col = x.shape[1]
x_sf = None
with use_symmetric_memory(
get_tp_group(), disabled=not is_allocation_symmetric()
):
symm_output = torch.empty(
x.shape[0], original_col, dtype=output_dtype, device=x.device
)
output = flashinfer_cutlass_fused_moe(
output=symm_output,
input=x_fp8,
token_selected_experts=topk_ids.to(torch.int),
token_final_scales=topk_weights,
fc1_expert_weights=layer.w13_weight,
fc2_expert_weights=layer.w2_weight,
output_dtype=output_dtype,
input_sf=x_sf,
quant_scales=[
layer.fc1_dequant,
layer.fc2_quant,
layer.fc2_dequant,
layer.fc1_input_dequant,
],
ep_size=layer.moe_ep_size,
ep_rank=layer.moe_ep_rank,
tp_size=layer.moe_tp_size,
tp_rank=layer.moe_tp_rank,
tune_max_num_tokens=next_power_of_2(x.shape[0]),
activation_type=activation,
)[0]
from sglang.srt.layers.moe.token_dispatcher import StandardCombineInput
return StandardCombineInput(hidden_states=output)
quant_info = TritonMoeQuantInfo(
w13_weight=layer.w13_weight,
w2_weight=layer.w2_weight,
use_fp8_w8a8=True,
per_channel_quant=False,
w13_scale=layer.w13_weight_scale,
w2_scale=layer.w2_weight_scale,
a13_scale=layer.w13_input_scale,
a2_scale=layer.w2_input_scale,
)
return self.runner.run(dispatch_output, quant_info)
class ModelOptFp4Config(ModelOptQuantConfig):
"""Config class for FP4."""
def __init__(
self,
is_checkpoint_nvfp4_serialized: bool = False,
kv_cache_quant_algo: str = None,
group_size: int = None,
exclude_modules: List[str] = None,
packed_modules_mapping: Optional[Dict[str, List[str]]] = None,
) -> None:
super().__init__(kv_cache_quant_algo, exclude_modules, packed_modules_mapping)
self.is_checkpoint_nvfp4_serialized = is_checkpoint_nvfp4_serialized
if is_checkpoint_nvfp4_serialized:
logger.warning(
"Detected nvfp4 checkpoint. Please note that the "
"format is experimental and subject to change."
)
self.group_size = group_size
@classmethod
def override_quantization_method(cls, hf_quant_config, user_quant):
"""Override quantization method based on the model's config."""
return cls._modelopt_override_quantization_method(hf_quant_config, user_quant)
@classmethod
def get_name(cls) -> str:
return "modelopt_fp4"
@classmethod
def get_supported_act_dtypes(cls) -> List[torch.dtype]:
return [torch.bfloat16, torch.half, torch.float8_e4m3fn]
@classmethod
def get_min_capability(cls) -> int:
return 100
@staticmethod
def common_group_size(cfg: dict) -> int:
"""Return the unique group_size across the config; raise if missing/mismatched."""
sizes = set()
# Top-level and 'quantization' block
v = cfg.get("group_size")
if isinstance(v, int):
sizes.add(v)
q = cfg.get("quantization")
if isinstance(q, dict):
v = q.get("group_size")
if isinstance(v, int):
sizes.add(v)
# config_groups: accept group-level or nested dicts (e.g., weights/input_activations)
for g in (cfg.get("config_groups") or {}).values():
if isinstance(g, dict):
v = g.get("group_size")
if isinstance(v, int):
sizes.add(v)
for sub in g.values():
if isinstance(sub, dict):
v = sub.get("group_size")
if isinstance(v, int):
sizes.add(v)
if not sizes:
raise ValueError("No group_size found in config.")
if len(sizes) > 1:
raise ValueError(f"Inconsistent group_size values: {sorted(sizes)}")
return next(iter(sizes))
@classmethod
def from_config(cls, config: Dict[str, Any]) -> ModelOptFp4Config:
# Handle two different config formats:
# 1. hf_quant_config.json format: {"quantization": {"quant_algo": "NVFP4", ...}}
# 2. config.json quantization_config format: {"quant_algo": "NVFP4", ...}
# In future modelopt will deprecate hf_quant_config.json, and only keep config.json.
# For legacy reasons, we keep hf_quant_config.json for now.
# Initialize variables
kv_cache_quant_algo = None
group_size = None
exclude_modules = []
# Try flat format first (config.json quantization_config - preferred format)
quant_method = config.get("quant_algo")
if quant_method is not None:
# Flat format (config.json quantization_config)
# Note: FP4 models in config.json format may not have all the detailed fields
# that are present in hf_quant_config.json, so we need to handle defaults
kv_cache_quant_algo = config.get("kv_cache_quant_algo")
if not kv_cache_quant_algo:
# For config.json format, derive from kv_cache_scheme if available
kv_cache_scheme = config.get("kv_cache_scheme")
if (
kv_cache_scheme
and kv_cache_scheme.get("type") == "float"
and kv_cache_scheme.get("num_bits") == 8
):
kv_cache_quant_algo = "FP8"
else:
kv_cache_quant_algo = "auto"
group_size = config.get("group_size")
# If group_size is not at top level, try to extract from config_groups
if group_size is None:
config_groups = config.get("config_groups", {})
if config_groups:
# Get group_size from the first group's weights config
first_group = next(iter(config_groups.values()), {})
weights_config = first_group.get("weights", {})
group_size = weights_config.get("group_size")
exclude_modules = config.get("ignore", [])
else:
# Fall back to nested format (hf_quant_config.json - legacy format)
try:
quant_config = cls.get_from_keys(config, ["quantization"])
quant_method = quant_config["quant_algo"]
kv_cache_quant_algo = quant_config.get("kv_cache_quant_algo")
if not kv_cache_quant_algo:
kv_cache_quant_algo = "auto"
group_size = ModelOptFp4Config.common_group_size(config)
exclude_modules = quant_config.get("exclude_modules", [])
except (ValueError, KeyError):
raise ValueError(
"Cannot find 'quant_algo' in the model's quantization config. "
"Expected either flat format (config.json) or nested format (hf_quant_config.json)."
)
if not quant_method in ["FP8", "NVFP4"]:
raise ValueError(
f"ModelOpt currently only supports: FP8, NVFP4"
" quantizations in sglang. Please check the "
"quantization config for your model's configuration."
)
is_checkpoint_nvfp4_serialized = "NVFP4" in quant_method
if group_size is None or exclude_modules is None:
logger.warning(
f"group_size: {group_size},"
f"kv_cache_quant_algo: {kv_cache_quant_algo},"
f"exclude_modules: {exclude_modules}"
)
raise ValueError(
"NVFP4 quantization requires group_size and exclude_modules "
"specified in the quantization config"
)
return cls(
is_checkpoint_nvfp4_serialized,
kv_cache_quant_algo,
group_size,
exclude_modules,
config.get("packed_modules_mapping"),
)
def is_layer_excluded(self, prefix: str):
import regex as re
fused_patterns = ["q_a_proj", "q_b_proj", "kv_a_proj_with_mqa", "kv_b_proj"]
prefix_split = prefix.split(".")
for pattern in self.exclude_modules:
regex_str = pattern.replace(".", r"\.").replace("*", r".*")
pattern_split = pattern.split(".")
if re.fullmatch(regex_str, prefix):
return True
elif (
pattern_split[-1] in fused_patterns
and pattern_split[-1] in prefix_split[-1]
):
# Check if the last part of the excluded pattern is contained in the last part of the prefix
# This handles fused modules like fused_qkv_a_proj_with_mqa that contain q_a_proj and kv_a_proj_with_mqa
# e.g., model.layers.{i}.self_attn.{fused_weight_name}
assert len(prefix_split) == 5 and len(pattern_split) == 5
return True
return False
def get_quant_method(self, layer: torch.nn.Module, prefix: str):
return self._get_quant_method(
layer,
prefix,
Linear=ModelOptFp4LinearMethod,
Moe=ModelOptNvFp4FusedMoEMethod, # FlashInferFP4MoE needs the same quantization method but with compatible attribute handling
)
class ModelOptFp4LinearMethod(LinearMethodBase):
"""Linear method for NVFP4.
Supports loading NVFP4 checkpoints with the following structure:
|Tensor Name | datatype | shape |
|----------------------------------------------------|
|input_scale | torch.float32 | scalar |
|weight | NVFP4(SE2M1) | [1, X, y/2] |
|weight_scale | FP8-E4M3 | [X, Y] |
|weight_scale_2 | torch.float32 | scalar |
The weights are quantized per block of 16 elements.
Args: quant_config: The ModelOpt quantization config.
"""
def __init__(self, quant_config: ModelOptFp4Config):
self.quant_config = quant_config
def create_weights(
self,
layer: torch.nn.Module,
input_size_per_partition: int,
output_partition_sizes: List[int],
input_size: int,
output_size: int,
params_dtype: torch.dtype,
**extra_weight_attrs,
):
del input_size, output_size
if not self.quant_config.is_checkpoint_nvfp4_serialized:
raise ValueError(
"NVFP4 quantization was selected, "
" dynamic quantization is not supported."
)
output_size_per_partition = sum(output_partition_sizes)
weight_loader = extra_weight_attrs.get("weight_loader")
layer.logical_widths = output_partition_sizes
layer.input_size_per_partition = input_size_per_partition
layer.output_size_per_partition = output_size_per_partition
if input_size_per_partition % 16 != 0:
raise ValueError(
"Unsupported model when in features size is " "not multiple of 16"
)
weight_dtype = (
torch.float8_e4m3fn
if self.quant_config.is_checkpoint_nvfp4_serialized
else params_dtype
)
weight = ModelWeightParameter(
data=torch.empty(
# 2 fp4 data is packed in one uint8 in the input dimension
output_size_per_partition,
input_size_per_partition // 2,
dtype=torch.uint8,
),
input_dim=1,
output_dim=0,
weight_loader=weight_loader,
)
layer.register_parameter("weight", weight)
input_scale = PerTensorScaleParameter(
data=torch.empty(len(output_partition_sizes), dtype=torch.float32),
weight_loader=weight_loader,
)
layer.register_parameter("input_scale", input_scale)
weight_scale_2 = PerTensorScaleParameter(
data=torch.empty(len(output_partition_sizes), dtype=torch.float32),
weight_loader=weight_loader,
)
layer.register_parameter("weight_scale_2", weight_scale_2)
weight_scale = ModelWeightParameter(
data=torch.empty(
output_size_per_partition,
input_size_per_partition // self.quant_config.group_size,
dtype=weight_dtype,
),
input_dim=1,
output_dim=0,
weight_loader=weight_loader,
)
layer.register_parameter("weight_scale", weight_scale)
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
input_scale_2 = layer.input_scale.max().to(torch.float32)
weight_scale_2 = layer.weight_scale_2.max().to(torch.float32)
layer.input_scale = Parameter(input_scale_2, requires_grad=False)
layer.weight_scale_2 = Parameter(weight_scale_2, requires_grad=False)
layer.alpha = Parameter(
layer.input_scale * layer.weight_scale_2, requires_grad=False
)
layer.input_scale_inv = Parameter(
(1 / input_scale_2).to(torch.float32), requires_grad=False
)
if FLASHINFER_FP4_GEMM_BACKEND == "trtllm":
# FlashInfer TRTLLM FP4 GEMM requires a different weight layout.
# FlashInfer provides nvfp4_quantize to quantize + shuffle the
# layout but we use our own quantization so we have to call
# shuffles ourselves.
from flashinfer import shuffle_matrix_a, shuffle_matrix_sf_a
weight = layer.weight
scale = layer.weight_scale
epilogue_tile_m = 128
weight = shuffle_matrix_a(weight.view(torch.uint8), epilogue_tile_m)
scale = (
shuffle_matrix_sf_a(scale.view(torch.uint8), epilogue_tile_m)
.reshape(scale.shape)
.view(torch.float8_e4m3fn)
)
layer.weight_scale_interleaved = Parameter(scale, requires_grad=False)
layer.weight = Parameter(weight, requires_grad=False)
return
# Pad and blockwise interleave weight_scale
scales = layer.weight_scale
scale_ndim = scales.ndim
if scale_ndim == 2:
scales = scales.unsqueeze(0)
assert scales.ndim == 3
B, M, K = scales.shape
round_up_multiple = lambda x, m: (x + m - 1) // m * m
M_padded = round_up_multiple(M, 128)
K_padded = round_up_multiple(K, 4)
padded_scales = torch.zeros((B, M_padded, K_padded), dtype=scales.dtype)
padded_scales[:B, :M, :K] = scales
batches, rows, cols = padded_scales.shape
assert rows % 128 == 0
assert cols % 4 == 0
padded_scales = padded_scales.reshape(batches, rows // 128, 4, 32, cols // 4, 4)
padded_scales = padded_scales.permute((0, 1, 4, 3, 2, 5))
padded_scales = padded_scales.contiguous().cuda()
padded_scales = (
padded_scales.reshape(M_padded, K_padded)
if scale_ndim == 2
else padded_scales.reshape(B, M_padded, K_padded)
)
layer.weight_scale_interleaved = Parameter(padded_scales, requires_grad=False)
def apply(
self,
layer: torch.nn.Module,
x: torch.Tensor,
bias: Optional[torch.Tensor] = None,
) -> torch.Tensor:
output_dtype = x.dtype
x_m, _ = x.shape
w_n, _ = layer.weight.shape
output_shape = [x_m, w_n]
# Quantize BF16 or FP16 to (FP4 and interleaved block scale)
x_fp4, x_scale_interleaved = fp4_quantize(x, layer.input_scale_inv)
assert x_fp4.dtype == torch.uint8
assert layer.weight.dtype == torch.uint8
assert layer.weight_scale_interleaved.dtype == torch.float8_e4m3fn
assert layer.alpha.dtype == torch.float32
w = layer.weight
w_scale_interleaved = layer.weight_scale_interleaved
if enable_flashinfer_fp4_gemm:
w = layer.weight.T
w_scale_interleaved = layer.weight_scale_interleaved.T
# TODO(shuw@nvidia.com)
# Remove the default after flashinfer bumped to 0.5.1
backend = (
FLASHINFER_FP4_GEMM_BACKEND if FLASHINFER_FP4_GEMM_BACKEND else "cutlass"
)
out = _sglang_fp4_gemm(
x_fp4,
w,
x_scale_interleaved,
w_scale_interleaved,
layer.alpha,
output_dtype,
w_n,
)
if bias is not None:
out = out + bias
return out.view(*output_shape)
class ModelOptNvFp4FusedMoEMethod(FusedMoEMethodBase):
"""
MoE Method for FP4 Quantization with Blockscales and PerTensorScales
Args:
quant_config: NVFP4 Quant Config
"""
def __init__(self, quant_config: ModelOptFp4Config):
self.quant_config = quant_config
if not is_blackwell_supported():
raise ValueError(
"Current platform does not support NVFP4"
" quantization. Please use Blackwell and"
" above."
)
self.enable_flashinfer_trtllm_moe = (
get_moe_runner_backend().is_flashinfer_trtllm()
)
self._cache_permute_indices = {}
@property
def enable_flashinfer_cutlass_moe(self) -> bool:
from sglang.srt.layers.moe import get_moe_runner_backend
"""Access the global enable_flashinfer_cutlass_moe setting."""
return get_moe_runner_backend().is_flashinfer_cutlass()
@property
def enable_flashinfer_cutedsl_moe(self) -> bool:
from sglang.srt.layers.moe import get_moe_runner_backend
"""Access the global enable_flashinfer_cutedsl_moe setting."""
return get_moe_runner_backend().is_flashinfer_cutedsl()
def create_weights(
self,
layer: torch.nn.Module,
num_experts: int,
hidden_size: int,
intermediate_size_per_partition: int,
params_dtype: torch.dtype,
**extra_weight_attrs,
):
if not self.quant_config.is_checkpoint_nvfp4_serialized:
raise ValueError(
"NVFP4 quantization was selected, "
" dynamic quantization is not supported."
)
# TODO(ch-wan): check if this is needed
layer.intermediate_size_per_partition = intermediate_size_per_partition
layer.params_dtype = params_dtype
layer.quant_config = self.quant_config
weight_dtype = torch.uint8
weight_scale_dtype = torch.float8_e4m3fn
weight_loader = extra_weight_attrs.get("weight_loader")
# GEMM 1
num_shards = 2 if layer.moe_runner_config.is_gated else 1
w13_weight = ModelWeightParameter(
data=torch.empty(
layer.num_local_experts,
num_shards * intermediate_size_per_partition,
# 2 fp4 items are packed in the input dimension
hidden_size // 2,
dtype=weight_dtype,
),
input_dim=1,
output_dim=2,
weight_loader=weight_loader,
)
layer.register_parameter("w13_weight", w13_weight)
# GEMM 2
w2_weight = ModelWeightParameter(
data=torch.empty(
layer.num_local_experts,
hidden_size,
# 2 fp4 items are packed in the input dimension
intermediate_size_per_partition // 2,
dtype=weight_dtype,
),
input_dim=1,
output_dim=2,
weight_loader=weight_loader,
)
layer.register_parameter("w2_weight", w2_weight)
w13_weight_scale = ModelWeightParameter(
data=torch.empty(
layer.num_local_experts,
num_shards * intermediate_size_per_partition,
hidden_size // self.quant_config.group_size,
dtype=weight_scale_dtype,
),
input_dim=1,
output_dim=2,
weight_loader=weight_loader,
)
layer.register_parameter("w13_weight_scale", w13_weight_scale)
# Only use `swizzle_blockscale` for shapes, not for real content
layer.w13_blockscale_swizzled = Parameter(
swizzle_blockscale(layer.w13_weight_scale), requires_grad=False
)
w2_weight_scale = ModelWeightParameter(
data=torch.empty(
layer.num_local_experts,
hidden_size,
intermediate_size_per_partition // self.quant_config.group_size,
dtype=weight_scale_dtype,
),
input_dim=1,
output_dim=2,
weight_loader=weight_loader,
)
layer.register_parameter("w2_weight_scale", w2_weight_scale)
layer.w2_blockscale_swizzled = Parameter(
swizzle_blockscale(layer.w2_weight_scale), requires_grad=False
)
from sglang.srt.layers.moe.fused_moe_triton import FusedMoeWeightScaleSupported
extra_weight_attrs.update(
{"quant_method": FusedMoeWeightScaleSupported.BLOCK.value}
)
w13_weight_scale_shape = (
(layer.num_local_experts, 2)
if layer.moe_runner_config.is_gated
else (layer.num_local_experts,)
)
w13_weight_scale_2 = PerTensorScaleParameter(
data=torch.empty(w13_weight_scale_shape, dtype=torch.float32),
weight_loader=weight_loader,
)
layer.register_parameter("w13_weight_scale_2", w13_weight_scale_2)
w2_weight_scale_2 = PerTensorScaleParameter(
data=torch.empty(layer.num_local_experts, dtype=torch.float32),
weight_loader=weight_loader,
)
layer.register_parameter("w2_weight_scale_2", w2_weight_scale_2)
extra_weight_attrs.update(
{"quant_method": FusedMoeWeightScaleSupported.TENSOR.value}
)
w13_input_scale_shape = (layer.num_experts, num_shards)
w13_input_scale = PerTensorScaleParameter(
data=torch.empty(w13_input_scale_shape, dtype=torch.float32),
weight_loader=weight_loader,
)
w13_input_scale._sglang_require_global_experts = True
layer.register_parameter("w13_input_scale", w13_input_scale)
w2_input_scale = PerTensorScaleParameter(
data=torch.empty(layer.num_experts, dtype=torch.float32),
weight_loader=weight_loader,
)
w2_input_scale._sglang_require_global_experts = True
layer.register_parameter("w2_input_scale", w2_input_scale)
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
"""Process FP4 MoE weights after loading from serialized checkpoint.
Only supports pre-quantized checkpoints with FP8 weights and scales.
"""
# GEMM 1 scale processing
if layer.moe_runner_config.is_gated:
if not torch.allclose(
layer.w13_weight_scale_2[:, 0], layer.w13_weight_scale_2[:, 1]
):
logger.warning_once(
"w1_weight_scale_2 must match w3_weight_scale_2. "
"Accuracy may be affected."
)
w13_weight_scale_2 = layer.w13_weight_scale_2[:, 0]
else:
w13_weight_scale_2 = layer.w13_weight_scale_2[:]
layer.w13_weight_scale_2 = Parameter(w13_weight_scale_2, requires_grad=False)
# Calculate input scales based on strategy
if self.enable_flashinfer_cutlass_moe or self.enable_flashinfer_trtllm_moe:
w13_input_scale = layer.w13_input_scale.max().to(torch.float32)
w2_input_scale = layer.w2_input_scale.max().to(torch.float32)
elif self.enable_flashinfer_cutedsl_moe:
# All-expert-one-input-scale is mathematically different from default per-expert-input-scale
# Thus we allow users to switch the flag to do thorough testing
if CUTEDSL_MOE_SCALAR_INPUT_SCALE:
w13_input_scale = (
layer.w13_input_scale.max()
.to(torch.float32)
.repeat(layer.w13_input_scale.shape[0])
)
else:
w13_input_scale = layer.w13_input_scale.max(dim=1).values.to(
torch.float32
)
w2_input_scale = layer.w2_input_scale
def _slice_scale(w):
assert w.shape == (layer.num_experts,)
assert layer.moe_ep_size * layer.num_local_experts == layer.num_experts
return w[
layer.moe_ep_rank
* layer.num_local_experts : (layer.moe_ep_rank + 1)
* layer.num_local_experts
]
w13_input_scale = _slice_scale(w13_input_scale)
w2_input_scale = _slice_scale(w2_input_scale)
if MOE_NVFP4_DISPATCH:
assert torch.all(w13_input_scale == w13_input_scale[0])
w13_input_scale = w13_input_scale[0]
else:
w13_input_scale = layer.w13_input_scale.max(dim=-1).values.to(torch.float32)
w2_input_scale = layer.w2_input_scale
# Create shared parameters
layer.g1_alphas = Parameter(
(w13_input_scale * w13_weight_scale_2).to(torch.float32),
requires_grad=False,
)
layer.g2_alphas = Parameter(
(w2_input_scale * layer.w2_weight_scale_2).to(torch.float32),
requires_grad=False,
)
layer.w13_input_scale_quant = Parameter(
(1 / w13_input_scale).to(torch.float32), requires_grad=False
)
layer.w2_input_scale_quant = Parameter(
(1 / w2_input_scale).to(torch.float32), requires_grad=False
)
layer.dispatcher.set_quant_config(
{
"input_global_scale": (
layer.w13_input_scale_quant
if MOE_NVFP4_DISPATCH
or should_use_flashinfer_cutlass_moe_fp4_allgather()
else None
)
}
)
# Validate weight scales
assert_dim = 2 if layer.moe_runner_config.is_gated else 1
for name, weight_scale in [
("w13", layer.w13_weight_scale),
("w2", layer.w2_weight_scale),
]:
assert (
weight_scale.shape[assert_dim] % 16 == 0
), f"Expected {name}_weight_scale.dim({assert_dim}) to be divisible by 16"
assert (
weight_scale.dtype == torch.float8_e4m3fn
), f"{name} Weight Blockscale must be represented as FP8-E4M3"
# Weight processing based on strategy
if (
self.enable_flashinfer_trtllm_moe
and reorder_rows_for_gated_act_gemm is not None
and shuffle_matrix_sf_a is not None
):
# FlashInfer TRTLLM processing - handles both w13 and w2
(
gemm1_weights_fp4_shuffled,
gemm1_scales_fp4_shuffled,
gemm2_weights_fp4_shuffled,
gemm2_scales_fp4_shuffled,
) = prepare_static_weights_for_trtllm_fp4_moe(
layer.w13_weight,
layer.w2_weight,
layer.w13_weight_scale,
layer.w2_weight_scale,
layer.w2_weight.size(-2), # hidden_size
layer.w13_weight.size(-2) // 2, # intermediate_size
layer.w13_weight.size(0), # num_experts
)
# Set flashinfer parameters
layer.gemm1_weights_fp4_shuffled = Parameter(
gemm1_weights_fp4_shuffled, requires_grad=False
)
layer.gemm2_weights_fp4_shuffled = Parameter(
gemm2_weights_fp4_shuffled, requires_grad=False
)
layer.gemm1_scales_fp4_shuffled = Parameter(
gemm1_scales_fp4_shuffled, requires_grad=False
)
layer.gemm2_scales_fp4_shuffled = Parameter(
gemm2_scales_fp4_shuffled, requires_grad=False
)
# Additional parameter needed for TRT-LLM
layer.g1_scale_c = Parameter(
(layer.w2_input_scale_quant * layer.g1_alphas).to(torch.float32),
requires_grad=False,
)
# Clean up weights that won't be used by TRT-LLM
del (
layer.w2_weight,
layer.w2_weight_scale,
layer.w13_weight,
layer.w13_weight_scale,
)
else:
# CUTLASS processing - handle w13 and w2 separately
# Process w13 weights
w13_blockscale_swizzled = swizzle_blockscale(layer.w13_weight_scale)
del layer.w13_weight_scale
layer.w13_blockscale_swizzled.data.copy_(w13_blockscale_swizzled)
w13_weight = layer.w13_weight
intermediate_size_pad = w13_blockscale_swizzled.size(1) - w13_weight.size(1)
if intermediate_size_pad:
# padding gated activations will require to split w1 and w3
# and pad them individually
assert not layer.moe_runner_config.is_gated, (
"The intermediate size required padding, "
"but padding is also implemented for gated activations"
)
layer.w13_weight = Parameter(
torch.nn.functional.pad(
w13_weight, (0, 0, 0, intermediate_size_pad)
),
requires_grad=False,
)
layer.w2_weight = Parameter(
torch.nn.functional.pad(
layer.w2_weight, (0, intermediate_size_pad // 2, 0, 0)
),
requires_grad=False,
)
layer.w2_weight_scale = Parameter(
torch.nn.functional.pad(
layer.w2_weight_scale, (0, intermediate_size_pad // 16)
),
requires_grad=False,
)
layer.w2_blockscale_swizzled = Parameter(
swizzle_blockscale(layer.w2_weight_scale), requires_grad=False
)
layer.w13_weight = Parameter(layer.w13_weight.data, requires_grad=False)
# Process w2 weights
w2_blockscale_swizzled = swizzle_blockscale(layer.w2_weight_scale)
del layer.w2_weight_scale
layer.w2_blockscale_swizzled.data.copy_(w2_blockscale_swizzled)
# Both flashinfer cutlass and regular cutlass use same processing for w2
# Set up CUTLASS MoE parameters
device = layer.w13_weight.device
layer.cutlass_moe_params = CutlassMoEParams(
CutlassMoEType.BlockscaledFP4,
device,
num_experts=layer.num_experts, # global num experts
intermediate_size_per_partition=layer.w2_weight.shape[2] * 2, # n
hidden_size=layer.w13_weight.shape[2] * 2,
) # k
@property
def load_up_proj_weight_first(self) -> bool:
# FlashInfer CUTLASS kernel assumes [Up, Gate] Proj as W13
return self.enable_flashinfer_cutlass_moe and self.moe_runner_config.is_gated
def create_moe_runner(
self, layer: torch.nn.Module, moe_runner_config: MoeRunnerConfig
):
self.moe_runner_config = moe_runner_config
def apply(
self,
layer: FusedMoE,
dispatch_output: StandardDispatchOutput,
) -> CombineInput:
x = dispatch_output.hidden_states
x_sf = dispatch_output.hidden_states_scale
topk_output = dispatch_output.topk_output
activation = self.moe_runner_config.activation
assert (
activation in ACT_STR_TO_TYPE_MAP
), f"{activation=} missing from {ACT_STR_TO_TYPE_MAP.keys()=}"
moe_runner_config = self.moe_runner_config
# Check if this is a FlashInferFP4MoE layer that should handle its own forward
if hasattr(layer, "gemm1_weights_fp4_shuffled"):
# This layer was processed with flashinfer TRTLLM - delegate to its own forward
from sglang.srt.layers.moe.token_dispatcher import StandardCombineInput
return StandardCombineInput(hidden_states=layer.forward(x, topk_output))
if self.enable_flashinfer_cutlass_moe:
assert (
not moe_runner_config.apply_router_weight_on_input
), "apply_router_weight_on_input is not supported for Flashinfer"
# TRTLLM Cutlass moe takes in activations in BF16/Half/nvfp4 precision
# and fp4 quantized weights loaded from the checkpoint
topk_weights, topk_ids = topk_output.topk_weights, topk_output.topk_ids
output_dtype = torch.bfloat16
# If x_sf is not None, x is FP4 packed (half size), so we need * 2
# If x_sf is None, x is not packed, so output_col = x.shape[1]
output_col = x.shape[1]
if x_sf is not None and layer.moe_runner_config.is_gated:
output_col *= 2
with use_symmetric_memory(
get_tp_group(), disabled=not is_allocation_symmetric()
):
symm_output = torch.empty(
x.shape[0],
output_col,
dtype=output_dtype,
device=x.device,
)
output = flashinfer_cutlass_fused_moe(
output=symm_output,
input=x,
token_selected_experts=topk_ids.to(torch.int),
token_final_scales=topk_weights,
fc1_expert_weights=layer.w13_weight.view(torch.long),
fc2_expert_weights=layer.w2_weight.view(torch.long),
output_dtype=output_dtype,
input_sf=x_sf,
quant_scales=[
layer.w13_input_scale_quant,
layer.w13_blockscale_swizzled.view(torch.int32),
layer.g1_alphas,
layer.w2_input_scale_quant,
layer.w2_blockscale_swizzled.view(torch.int32),
layer.g2_alphas,
],
ep_size=layer.moe_ep_size,
ep_rank=layer.moe_ep_rank,
tp_size=layer.moe_tp_size,
tp_rank=layer.moe_tp_rank,
tune_max_num_tokens=next_power_of_2(x.shape[0]),
activation_type=ACT_STR_TO_TYPE_MAP[activation],
)[0]
from sglang.srt.layers.moe.token_dispatcher import StandardCombineInput
return StandardCombineInput(hidden_states=output)
from sglang.srt.layers.moe.cutlass_moe import cutlass_moe_fp4
topk_weights, topk_ids = topk_output.topk_weights, topk_output.topk_ids
output = cutlass_moe_fp4(
a=x,
a1_gscale=layer.w13_input_scale_quant,
w1_fp4=layer.w13_weight,
w1_blockscale=layer.w13_blockscale_swizzled,
w1_alphas=layer.g1_alphas,
a2_gscale=layer.w2_input_scale_quant,
w2_fp4=layer.w2_weight,
w2_blockscale=layer.w2_blockscale_swizzled,
w2_alphas=layer.g2_alphas,
topk_weights=topk_weights,
topk_ids=topk_ids,
params=layer.cutlass_moe_params,
apply_router_weight_on_input=moe_runner_config.apply_router_weight_on_input,
).to(x.dtype)
# Scale by routed_scaling_factor is fused into select_experts.
from sglang.srt.layers.moe.token_dispatcher import StandardCombineInput
return StandardCombineInput(hidden_states=output)
def apply_without_routing_weights(
self,
layer: FusedMoE,
x: tuple[torch.Tensor, Optional[torch.Tensor]],
masked_m: torch.Tensor,
moe_runner_config: MoeRunnerConfig,
) -> torch.Tensor:
assert (
moe_runner_config.activation == "silu"
), "Only SiLU activation is supported."
assert self.enable_flashinfer_cutedsl_moe, "only support flashinfer cutedsl moe"
assert (
not moe_runner_config.apply_router_weight_on_input
), "apply_router_weight_on_input is not supported for Flashinfer"
from sglang.srt.layers.moe.flashinfer_cutedsl_moe import (
flashinfer_cutedsl_moe_masked,
)
down_gemm_overlap_args: Optional[DownGemmOverlapArgs] = getattr(
layer, "down_gemm_overlap_args", None
)
out = flashinfer_cutedsl_moe_masked(
hidden_states=x,
input_global_scale=(
None if MOE_NVFP4_DISPATCH else layer.w13_input_scale_quant
),
w1=layer.w13_weight,
w1_blockscale=layer.w13_blockscale_swizzled,
w1_alpha=layer.g1_alphas,
w2=layer.w2_weight,
a2_global_scale=layer.w2_input_scale_quant,
w2_blockscale=layer.w2_blockscale_swizzled,
w2_alpha=layer.g2_alphas,
masked_m=masked_m,
**(
dict(
down_sm_count=down_gemm_overlap_args.num_sms,
down_signals=down_gemm_overlap_args.signal,
down_start_event=down_gemm_overlap_args.start_event,
)
if down_gemm_overlap_args is not None
else {}
),
)
return out