biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Refactor] Make FP8 Linear Ops use kernel abstraction (#27814)

Browse Source 
Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com>
This commit is contained in:
vllmellm
2026-01-20 14:48:20 +08:00
committed by GitHub
parent e9c83cdc51
commit 148117ea2e
30 changed files with 1467 additions and 1038 deletions
Download Patch File Download Diff File Expand all files Collapse all files

63
vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_fp8.py
View File

@@ -8,9 +8,13 @@ from compressed_tensors.quantization import QuantizationArgs, QuantizationStrate
from torch.nn import Parameter
from vllm._aiter_ops import rocm_aiter_ops
from vllm.logger import init_logger
from vllm.model_executor.layers.quantization.compressed_tensors.schemes import (
CompressedTensorsScheme,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm import (
init_fp8_linear_kernel,
)
from vllm.model_executor.layers.quantization.utils.fp8_utils import (
W8A8BlockFp8LinearOp,
create_fp8_input_scale,
@@ -22,11 +26,14 @@ from vllm.model_executor.layers.quantization.utils.fp8_utils import (
process_fp8_weight_tensor_strategy,
validate_fp8_block_shape,
)
from vllm.model_executor.layers.quantization.utils.quant_utils import GroupShape
from vllm.model_executor.layers.quantization.utils.quant_utils import (
GroupShape,
kFp8DynamicTokenSym,
kFp8StaticTensorSym,
kFp8StaticTokenSym,
)
from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
Fp8LinearOp,
cutlass_block_fp8_supported,
maybe_create_device_identity,
)
from vllm.model_executor.parameter import (
BlockQuantScaleParameter,
@@ -42,6 +49,18 @@ strategy_to_parameter_type = {
QuantizationStrategy.TENSOR: PerTensorScaleParameter,
}
STATIC_QUANT = True
DYNAMIC_QUANT = False
activation_quant_key_mapping = {
STATIC_QUANT: kFp8StaticTensorSym,
DYNAMIC_QUANT: kFp8DynamicTokenSym,
}
weight_quant_key_mapping = {
QuantizationStrategy.CHANNEL: kFp8StaticTokenSym,
QuantizationStrategy.TENSOR: kFp8StaticTensorSym,
}
logger = init_logger(__name__)
class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
def __init__(self, weight_quant: QuantizationArgs, is_static_input_scheme: bool):
@@ -49,22 +68,13 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
self.strategy = weight_quant.strategy
self.out_dtype = torch.get_default_dtype()
self.is_static_input_scheme = is_static_input_scheme
self.weight_block_size = self.weight_quant.block_structure
if self.weight_block_size is not None:
self.act_q_group_shape = GroupShape(1, self.weight_block_size[0])
else:
self.act_q_group_shape = (
GroupShape.PER_TENSOR
if is_static_input_scheme
else GroupShape.PER_TOKEN
)
self.cutlass_block_fp8_supported = cutlass_block_fp8_supported()
self.use_aiter_and_is_supported = rocm_aiter_ops.is_linear_fp8_enabled()
if self.weight_block_size is not None:
self.cutlass_block_fp8_supported = cutlass_block_fp8_supported()
self.use_aiter_and_is_supported = rocm_aiter_ops.is_linear_fp8_enabled()
assert not self.is_static_input_scheme
self.act_q_group_shape = GroupShape(1, self.weight_block_size[0])
self.w8a8_block_fp8_linear = W8A8BlockFp8LinearOp(
weight_group_shape=GroupShape(*self.weight_block_size),
act_quant_group_shape=self.act_q_group_shape,
@@ -72,9 +82,13 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
use_aiter_and_is_supported=self.use_aiter_and_is_supported,
)
else:
self.fp8_linear = Fp8LinearOp(
act_quant_static=self.is_static_input_scheme,
act_quant_group_shape=self.act_q_group_shape,
activation_quant_key = activation_quant_key_mapping[is_static_input_scheme]
weight_quant_key = weight_quant_key_mapping[self.strategy]
self.fp8_linear = init_fp8_linear_kernel(
activation_quant_key=activation_quant_key,
weight_quant_key=weight_quant_key,
out_dtype=self.out_dtype,
module_name=self.__class__.__name__,
)
@classmethod
@@ -93,8 +107,6 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
weight_loader: Callable,
**kwargs,
):
maybe_create_device_identity()
output_size_per_partition = sum(output_partition_sizes)
layer.logical_widths = output_partition_sizes
layer.weight_block_size = None
@@ -143,7 +155,6 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
getattr(layer, "input_scale", None),
)
weight = weight.t()
elif self.strategy == QuantizationStrategy.CHANNEL:
weight, weight_scale, input_scale = process_fp8_weight_channel_strategy(
layer.weight, layer.weight_scale, getattr(layer, "input_scale", None)
@@ -174,7 +185,6 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
layer.input_scale = Parameter(layer.input_scale.max(), requires_grad=False)
else:
layer.input_scale = None
if self.strategy == QuantizationStrategy.BLOCK:
maybe_post_process_fp8_weight_block(layer)
@@ -193,11 +203,4 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
bias=bias,
)
return self.fp8_linear.apply(
input=x,
weight=layer.weight,
weight_scale=layer.weight_scale,
out_dtype=self.out_dtype,
input_scale=layer.input_scale,
bias=bias,
)
return self.fp8_linear.apply_weights(layer, x, bias)

31
vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_int8.py
View File

@@ -11,8 +11,7 @@ from vllm.model_executor.layers.quantization.compressed_tensors.schemes import (
CompressedTensorsScheme,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm import (
ScaledMMLinearLayerConfig,
choose_scaled_mm_linear_kernel,
init_int8_linear_kernel,
)
from vllm.model_executor.parameter import (
BasevLLMParameter,
@@ -25,8 +24,6 @@ logger = init_logger(__name__)
class CompressedTensorsW8A8Int8(CompressedTensorsScheme):
_kernel_backends_being_used: set[str] = set()
def __init__(
self, strategy: str, is_static_input_scheme: bool, input_symmetric: bool
):
@@ -50,18 +47,13 @@ class CompressedTensorsW8A8Int8(CompressedTensorsScheme):
):
layer.logical_widths = output_partition_sizes
scaled_mm_linear_kernel_config = ScaledMMLinearLayerConfig(
self.kernel = init_int8_linear_kernel(
is_channelwise=(self.strategy == QuantizationStrategy.CHANNEL),
is_static_input_scheme=self.is_static_input_scheme,
input_symmetric=self.input_symmetric,
module_name=self.__class__.__name__,
)
kernel_type = choose_scaled_mm_linear_kernel(scaled_mm_linear_kernel_config)
if kernel_type.__name__ not in self._kernel_backends_being_used:
logger.info("Using %s for CompressedTensorsW8A8Int8", kernel_type.__name__)
self._kernel_backends_being_used.add(kernel_type.__name__)
# WEIGHT
weight = ModelWeightParameter(
data=torch.empty(
@@ -90,12 +82,12 @@ class CompressedTensorsW8A8Int8(CompressedTensorsScheme):
layer.register_parameter("weight_scale", weight_scale)
# INPUT SCALE
input_zero_point = None
input_scale = None
if self.is_static_input_scheme:
input_scale = BasevLLMParameter(
data=torch.empty(1, dtype=torch.float32), weight_loader=weight_loader
)
layer.register_parameter("input_scale", input_scale)
if not self.input_symmetric:
# Note: compressed-tensors stores the zp using the same dtype
# as the weights
@@ -103,16 +95,11 @@ class CompressedTensorsW8A8Int8(CompressedTensorsScheme):
input_zero_point = BasevLLMParameter(
data=torch.empty(1, dtype=torch.int8), weight_loader=weight_loader
)
layer.register_parameter("input_zero_point", input_zero_point)
self.kernel = kernel_type(
c=scaled_mm_linear_kernel_config,
w_q_param_name="weight",
w_s_param_name="weight_scale",
i_s_param_name="input_scale",
i_zp_param_name="input_zero_point",
azp_adj_param_name="azp_adj",
)
layer.register_parameter("input_zero_point", input_zero_point)
layer.register_parameter("input_scale", input_scale)
if not hasattr(layer, "azp_adj"):
layer.register_parameter("azp_adj", None)
# Checkpoints are serialized in compressed-tensors format, which is
# different from the format the kernel may want. Handle repacking here.

28
vllm/model_executor/layers/quantization/fbgemm_fp8.py
View File

@@ -18,17 +18,19 @@ from vllm.model_executor.layers.quantization.base_config import (
QuantizationConfig,
QuantizeMethodBase,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm import (
init_fp8_linear_kernel,
)
from vllm.model_executor.layers.quantization.utils.marlin_utils_fp8 import (
apply_fp8_marlin_linear,
prepare_fp8_layer_for_marlin,
)
from vllm.model_executor.layers.quantization.utils.quant_utils import (
GroupShape,
is_layer_skipped,
kFp8DynamicTokenSym,
kFp8StaticTokenSym,
)
from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
Fp8LinearOp,
maybe_create_device_identity,
normalize_e4m3fn_to_e4m3fnuz,
)
from vllm.model_executor.parameter import (
@@ -91,10 +93,13 @@ class FBGEMMFp8Config(QuantizationConfig):
class FBGEMMFp8LinearMethod(LinearMethodBase):
def __init__(self, quant_config: FBGEMMFp8Config):
self.quant_config = quant_config
self.fp8_linear = Fp8LinearOp(
act_quant_static=False, act_quant_group_shape=GroupShape.PER_TOKEN
)
self.out_dtype = torch.get_default_dtype()
self.fp8_linear = init_fp8_linear_kernel(
activation_quant_key=kFp8DynamicTokenSym,
weight_quant_key=kFp8StaticTokenSym,
out_dtype=torch.get_default_dtype(),
module_name=self.__class__.__name__,
)
def create_weights(
self,
@@ -106,7 +111,6 @@ class FBGEMMFp8LinearMethod(LinearMethodBase):
params_dtype: torch.dtype,
**extra_weight_attrs,
):
maybe_create_device_identity()
weight_loader = extra_weight_attrs.get("weight_loader")
del input_size, output_size
output_size_per_partition = sum(output_partition_sizes)
@@ -184,12 +188,4 @@ class FBGEMMFp8LinearMethod(LinearMethodBase):
bias=bias,
)
return self.fp8_linear.apply(
input=x,
weight=layer.weight,
weight_scale=layer.weight_scale,
out_dtype=self.out_dtype,
input_scale=None,
input_scale_ub=layer.input_scale_ub,
bias=bias,
)
return self.fp8_linear.apply_weights(layer, x, bias)

47
vllm/model_executor/layers/quantization/fp8.py
View File

@@ -48,6 +48,9 @@ from vllm.model_executor.layers.quantization.base_config import (
QuantizationConfig,
QuantizeMethodBase,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm import (
init_fp8_linear_kernel,
)
from vllm.model_executor.layers.quantization.kv_cache import BaseKVCacheMethod
from vllm.model_executor.layers.quantization.utils.flashinfer_utils import (
apply_fi_trtllm_fp8_per_tensor_moe,
@@ -76,12 +79,13 @@ from vllm.model_executor.layers.quantization.utils.marlin_utils_fp8 import (
from vllm.model_executor.layers.quantization.utils.quant_utils import (
GroupShape,
is_layer_skipped,
kFp8DynamicTensorSym,
kFp8DynamicTokenSym,
kFp8StaticTensorSym,
)
from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
Fp8LinearOp,
cutlass_block_fp8_supported,
cutlass_fp8_supported,
maybe_create_device_identity,
normalize_e4m3fn_to_e4m3fnuz,
)
from vllm.model_executor.parameter import (
@@ -328,28 +332,30 @@ class Fp8LinearMethod(LinearMethodBase):
self.weight_block_size = self.quant_config.weight_block_size
self.block_quant = self.weight_block_size is not None
self.act_q_static = self.quant_config.activation_scheme == "static"
if self.weight_block_size:
self.act_q_group_shape = GroupShape(1, self.weight_block_size[0])
else:
# Use per-token quantization for better perf if dynamic and cutlass
if not self.act_q_static and cutlass_fp8_supported():
self.act_q_group_shape = GroupShape.PER_TOKEN
else:
self.act_q_group_shape = GroupShape.PER_TENSOR
if self.block_quant:
assert not self.act_q_static
assert self.weight_block_size is not None
self.w8a8_block_fp8_linear = W8A8BlockFp8LinearOp(
weight_group_shape=GroupShape(*self.weight_block_size),
act_quant_group_shape=self.act_q_group_shape,
act_quant_group_shape=GroupShape(1, self.weight_block_size[0]),
cutlass_block_fp8_supported=self.cutlass_block_fp8_supported,
use_aiter_and_is_supported=self.use_aiter_and_is_supported,
)
else:
self.fp8_linear = Fp8LinearOp(
act_quant_static=self.act_q_static,
act_quant_group_shape=self.act_q_group_shape,
# Use per-token quantization for better perf if dynamic and cutlass
if self.act_q_static:
activation_quant_key = kFp8StaticTensorSym
elif cutlass_fp8_supported():
activation_quant_key = kFp8DynamicTokenSym
else:
activation_quant_key = kFp8DynamicTensorSym
self.fp8_linear = init_fp8_linear_kernel(
activation_quant_key=activation_quant_key,
weight_quant_key=kFp8StaticTensorSym,
out_dtype=torch.get_default_dtype(),
module_name=self.__class__.__name__,
)
def create_weights(
@@ -362,8 +368,6 @@ class Fp8LinearMethod(LinearMethodBase):
params_dtype: torch.dtype,
**extra_weight_attrs,
):
maybe_create_device_identity()
output_size_per_partition = sum(output_partition_sizes)
weight_loader = extra_weight_attrs.get("weight_loader")
layer.logical_widths = output_partition_sizes
@@ -462,8 +466,6 @@ class Fp8LinearMethod(LinearMethodBase):
scale = create_fp8_input_scale(output_partition_sizes, weight_loader)
set_weight_attrs(scale, {"scale_type": "input_scale"})
layer.register_parameter("input_scale", scale)
else:
layer.register_parameter("input_scale", None)
def process_weights_after_loading(self, layer: Module) -> None:
if getattr(layer, "_already_called_process_weights_after_loading", False):
@@ -602,14 +604,7 @@ class Fp8LinearMethod(LinearMethodBase):
bias=bias,
)
return self.fp8_linear.apply(
input=x,
weight=layer.weight,
weight_scale=layer.weight_scale,
out_dtype=self.out_dtype,
input_scale=layer.input_scale,
bias=bias,
)
return self.fp8_linear.apply_weights(layer, x, bias)
class Fp8MoEMethod(FusedMoEMethodBase):

179
vllm/model_executor/layers/quantization/kernels/scaled_mm/ScaledMMLinearKernel.py
View File

@@ -2,19 +2,58 @@
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from abc import ABC, abstractmethod
from collections.abc import Sequence
from dataclasses import dataclass
from typing import Generic, TypeVar
import torch
from vllm.model_executor.layers.quantization.input_quant_fp8 import QuantFP8
from vllm.model_executor.layers.quantization.utils.quant_utils import (
QuantKey,
)
from vllm.platforms import current_platform
@dataclass
class ScaledMMLinearLayerConfig:
is_channelwise: bool
pass
@dataclass
class Int8ScaledMMLinearLayerConfig(ScaledMMLinearLayerConfig):
# TODO: Chnage to QuantKey like FP8ScaledMMLinearLayerConfig
is_static_input_scheme: bool
is_channelwise: bool
input_symmetric: bool
class ScaledMMLinearKernel(ABC):
@dataclass
class FP8ScaledMMLinearLayerConfig(ScaledMMLinearLayerConfig):
weight_quant_key: QuantKey
activation_quant_key: QuantKey
out_dtype: torch.dtype | None
_FP8ParamsT = tuple[
torch.Tensor, # weight
torch.Tensor, # weight_scale
torch.Tensor | None, # input_scale,
torch.Tensor | None, # input_scale_ub,
]
_Int8ParamsT = tuple[
torch.Tensor, # weight
torch.Tensor, # weight_scale
torch.Tensor | None, # input_scale,
torch.Tensor | None, # input_zp
torch.Tensor | None, # azp_adj
]
_ParamsT = TypeVar("_ParamsT", _Int8ParamsT, _FP8ParamsT)
_ConfigT = TypeVar("_ConfigT", bound=ScaledMMLinearLayerConfig)
class ScaledMMLinearKernel(Generic[_ConfigT, _ParamsT], ABC):
@classmethod
@abstractmethod
def is_supported(
@@ -24,26 +63,14 @@ class ScaledMMLinearKernel(ABC):
@classmethod
@abstractmethod
def can_implement(cls, c: ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
def can_implement(cls, c: _ConfigT) -> tuple[bool, str | None]:
raise NotImplementedError
def __init__(
self,
c: ScaledMMLinearLayerConfig,
w_q_param_name: str,
w_s_param_name: str,
i_s_param_name: str,
i_zp_param_name: str,
azp_adj_param_name: str,
) -> None:
assert self.can_implement(c)
assert self.is_supported()
def __init__(self, c: _ConfigT, layer_param_names: Sequence[str]) -> None:
assert self.can_implement(c)[0]
assert self.is_supported()[0]
self.config = c
self.w_q_name = w_q_param_name
self.w_s_name = w_s_param_name
self.i_s_name = i_s_param_name
self.i_zp_name = i_zp_param_name
self.azp_adj_name = azp_adj_param_name
self.layer_param_names = layer_param_names
@abstractmethod
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
@@ -58,19 +85,103 @@ class ScaledMMLinearKernel(ABC):
) -> torch.Tensor:
raise NotImplementedError
def _get_weight_params(
self, layer: torch.nn.Module
) -> tuple[
torch.Tensor, # weight
torch.Tensor, # weight_scale
torch.Tensor | None, # input_scale,
torch.Tensor | None, # input_zp
torch.Tensor | None, # azp_adj
]:
return (
getattr(layer, self.w_q_name),
getattr(layer, self.w_s_name),
getattr(layer, self.i_s_name),
getattr(layer, self.i_zp_name),
getattr(layer, self.azp_adj_name),
# return a covariant type in the subclass
@abstractmethod
def _get_layer_params(self, layer) -> _ParamsT:
raise NotImplementedError
class FP8ScaledMMLinearKernel(
ScaledMMLinearKernel[FP8ScaledMMLinearLayerConfig, _FP8ParamsT], ABC
):
def __init__(
self, c: FP8ScaledMMLinearLayerConfig, layer_param_names: Sequence[str]
) -> None:
act_scale_descriptor = c.activation_quant_key.scale
self.quant_fp8 = QuantFP8(
static=act_scale_descriptor.static,
group_shape=act_scale_descriptor.group_shape,
num_token_padding=self.get_output_padding(),
)
self.fp8_dtype = current_platform.fp8_dtype()
super().__init__(c, layer_param_names)
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
pass
def _get_layer_params(self, layer) -> _FP8ParamsT:
w, w_s, x_s, x_s_ub = self.layer_param_names
return (
getattr(layer, w),
getattr(layer, w_s),
getattr(layer, x_s, None),
getattr(layer, x_s_ub, None),
)
def apply_weights(
self,
layer: torch.nn.Module,
x: torch.Tensor,
bias: torch.Tensor | None = None,
) -> torch.Tensor:
fp8_dtype = self.fp8_dtype
maybe_out_dtype = self.config.out_dtype
w, w_s, x_s, x_s_ub = self._get_layer_params(layer)
# ops.scaled_fp8_quant supports both dynamic and static quant.
# If dynamic, layer.input_scale is None and x_s computed from x.
# If static, layer.input_scale is scalar and x_s is input_scale.
# View input as 2D matrix for fp8 methods
x_2d = x.view(-1, x.shape[-1])
output_shape = [*x.shape[:-1], w.shape[1]]
out_dtype = x.dtype if maybe_out_dtype is None else maybe_out_dtype
# If input not quantized
# TODO(luka) remove this path if not used anymore
x_2d_q = x_2d
if x.dtype != fp8_dtype:
x_2d_q, x_s = self.quant_fp8(
x_2d,
x_s,
x_s_ub,
)
return self.apply_scaled_mm(
A=x_2d_q,
B=w,
out_dtype=out_dtype,
As=x_s,
Bs=w_s,
bias=bias,
output_shape=output_shape,
)
@abstractmethod
def apply_scaled_mm(
self,
*,
A: torch.Tensor,
B: torch.Tensor,
out_dtype: torch.dtype,
As: torch.Tensor,
Bs: torch.Tensor,
bias: torch.Tensor | None,
output_shape: list,
) -> torch.Tensor:
raise NotImplementedError
def get_output_padding(self) -> int | None:
return None
class Int8ScaledMMLinearKernel(
ScaledMMLinearKernel[Int8ScaledMMLinearLayerConfig, _Int8ParamsT], ABC
):
def _get_layer_params(self, layer) -> _Int8ParamsT:
w_q, w_s, i_s, i_zp, azp_adj = self.layer_param_names
return (
getattr(layer, w_q),
getattr(layer, w_s),
getattr(layer, i_s, None),
getattr(layer, i_zp, None),
getattr(layer, azp_adj, None),
)

213
vllm/model_executor/layers/quantization/kernels/scaled_mm/__init__.py
View File

@@ -2,76 +2,229 @@
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import os
from typing import TypeVar
import torch
from vllm.logger import init_logger
from vllm.model_executor.layers.quantization.kernels.scaled_mm.aiter import (
AiterScaledMMLinearKernel,
AiterInt8ScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm.cpu import (
CPUScaledMMLinearKernel,
CPUInt8ScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm.cutlass import (
CutlassScaledMMLinearKernel,
CutlassFP8ScaledMMLinearKernel,
CutlassInt8ScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm.flashinfer import (
FlashInferFP8ScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm.pytorch import (
ChannelWiseTorchFP8ScaledMMLinearKernel,
PerTensorTorchFP8ScaledMMLinearKernel,
RowWiseTorchFP8ScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm.rocm import (
ROCmFP8ScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm.ScaledMMLinearKernel import ( # noqa: E501
FP8ScaledMMLinearKernel,
FP8ScaledMMLinearLayerConfig,
Int8ScaledMMLinearKernel,
Int8ScaledMMLinearLayerConfig,
ScaledMMLinearKernel,
ScaledMMLinearLayerConfig,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm.triton import (
TritonScaledMMLinearKernel,
TritonInt8ScaledMMLinearKernel,
)
from vllm.model_executor.layers.quantization.utils.quant_utils import QuantKey
from vllm.platforms import PlatformEnum, current_platform
logger = init_logger(__name__)
# in priority/performance order (when available)
_POSSIBLE_KERNELS: dict[PlatformEnum, list[type[ScaledMMLinearKernel]]] = {
PlatformEnum.CPU: [CPUScaledMMLinearKernel],
PlatformEnum.CUDA: [CutlassScaledMMLinearKernel, TritonScaledMMLinearKernel],
PlatformEnum.ROCM: [AiterScaledMMLinearKernel, TritonScaledMMLinearKernel],
_POSSIBLE_INT8_KERNELS: dict[PlatformEnum, list[type[Int8ScaledMMLinearKernel]]] = {
PlatformEnum.CPU: [CPUInt8ScaledMMLinearKernel],
PlatformEnum.CUDA: [
CutlassInt8ScaledMMLinearKernel,
TritonInt8ScaledMMLinearKernel,
],
PlatformEnum.ROCM: [AiterInt8ScaledMMLinearKernel, TritonInt8ScaledMMLinearKernel],
}
# in priority/performance order (when available)
_POSSIBLE_FP8_KERNELS: dict[PlatformEnum, list[type[FP8ScaledMMLinearKernel]]] = {
PlatformEnum.CUDA: [
FlashInferFP8ScaledMMLinearKernel,
CutlassFP8ScaledMMLinearKernel,
PerTensorTorchFP8ScaledMMLinearKernel,
ChannelWiseTorchFP8ScaledMMLinearKernel,
],
PlatformEnum.ROCM: [
ROCmFP8ScaledMMLinearKernel,
PerTensorTorchFP8ScaledMMLinearKernel,
RowWiseTorchFP8ScaledMMLinearKernel,
ChannelWiseTorchFP8ScaledMMLinearKernel,
],
PlatformEnum.CPU: [
PerTensorTorchFP8ScaledMMLinearKernel,
ChannelWiseTorchFP8ScaledMMLinearKernel,
],
}
_KernelT = TypeVar("_KernelT", bound=ScaledMMLinearKernel)
_KernelConfigT = TypeVar("_KernelConfigT", bound=ScaledMMLinearLayerConfig)
def is_supported_and_can_implement_kernel(
kernel: type[_KernelT], config: _KernelConfigT, compute_capability: int | None
) -> tuple[bool, str]:
# TODO: Fetch `VLLM_DISABLED_KERNELS` from vllm.envs instead.
if kernel.__name__ in os.environ.get("VLLM_DISABLED_KERNELS", "").split(","):
return False, f" {kernel.__name__} is disabled by environment variable"
if compute_capability is None:
_cc = current_platform.get_device_capability()
if _cc is not None:
compute_capability = _cc[0] * 10 + _cc[1]
is_supported, failure_reason = kernel.is_supported(compute_capability)
if not is_supported:
return False, f"{kernel.__name__} {failure_reason}."
can_implement, failure_reason = kernel.can_implement(config)
if not can_implement:
return (
False,
f"{kernel.__name__} {failure_reason}.",
)
return True, ""
def choose_scaled_mm_linear_kernel(
config: ScaledMMLinearLayerConfig, compute_capability: int | None = None
) -> type[ScaledMMLinearKernel]:
config: _KernelConfigT,
possible_kernels: dict[PlatformEnum, list[type[_KernelT]]],
compute_capability: int | None = None,
force_kernel: type[_KernelT] | None = None,
) -> type[_KernelT]:
"""
Choose an ScaledMMLinearKernel that can implement the given config for the
Choose a _KernelT that can implement the given config for the
given compute capability. Attempts to choose the best kernel in terms of
performance.
Args:
config (ScaledMMLinearLayerConfig): Description of the linear layer
config (_KernelConfigT): Description of the linear layer
to be implemented.
possible_kernels (dict[PlatformEnum, list[_KernelT]]): A
dictionary of platforms and their list list of possible kernels.
compute_capability (Optional[int], optional): The compute capability of
the target device, if None uses `current_platform` to get the
compute capability. Defaults to None.
force_kernel (Optional[type[_KernelT]]): An Optional forced kernel to override
the possible_kernels if it can be implemented. If None, it will only try the
possible kernels.
Raises:
ValueError: If no kernel can implement the given config.
Returns:
type[ScaledMMLinearKernel]: Chosen kernel.
_KernelT: Chosen kernel.
"""
failure_reasons = []
for kernel in _POSSIBLE_KERNELS[current_platform._enum]:
if kernel.__name__ in os.environ.get("VLLM_DISABLED_KERNELS", "").split(","):
failure_reasons.append(f"{kernel.__name__}: disabled by env var")
continue
failure_reason_list = []
# If the current platform uses compute_capability,
# make sure the kernel supports the compute capability.
is_supported, reason = kernel.is_supported(compute_capability)
if not is_supported:
failure_reasons.append(f"{kernel.__name__}: {reason}")
continue
if force_kernel is not None:
can_implement, failure_reason = is_supported_and_can_implement_kernel(
force_kernel, config, compute_capability
)
if can_implement:
return force_kernel
can_implement, reason = kernel.can_implement(config)
if not can_implement:
failure_reasons.append(f"{kernel.__name__}: {reason}")
continue
logger.info_once(
"Tried to force %s, but the kernel couldn't be implemented",
force_kernel.__name__,
scope="global",
)
return kernel
for kernel in possible_kernels[current_platform._enum]:
is_supported_and_can_implement, failure_reason = (
is_supported_and_can_implement_kernel(kernel, config, compute_capability)
)
if is_supported_and_can_implement:
return kernel
failure_reason_list.append(failure_reason)
raise ValueError(
"Failed to find a kernel that can implement the "
"ScaledMM linear layer. Reasons: \n" + "\n".join(failure_reasons)
"ScaledMM linear layer. Reasons: \n" + "\n".join(failure_reason_list)
)
def init_fp8_linear_kernel(
activation_quant_key: QuantKey,
weight_quant_key: QuantKey,
out_dtype: torch.dtype,
force_kernel: type[FP8ScaledMMLinearKernel] | None = None,
module_name: str | None = None,
) -> FP8ScaledMMLinearKernel:
scaled_mm_linear_kernel_config = FP8ScaledMMLinearLayerConfig(
weight_quant_key=weight_quant_key,
activation_quant_key=activation_quant_key,
out_dtype=out_dtype,
)
kernel_type = choose_scaled_mm_linear_kernel(
scaled_mm_linear_kernel_config, _POSSIBLE_FP8_KERNELS, force_kernel=force_kernel
)
if module_name:
logger.info_once(
"Selected %s for %s",
kernel_type.__name__,
module_name,
scope="global",
)
return kernel_type(
scaled_mm_linear_kernel_config,
layer_param_names=["weight", "weight_scale", "input_scale", "input_scale_ub"],
)
def init_int8_linear_kernel(
is_channelwise: bool,
is_static_input_scheme: bool,
input_symmetric: bool,
module_name: str,
) -> Int8ScaledMMLinearKernel:
config = Int8ScaledMMLinearLayerConfig(
is_channelwise=is_channelwise,
is_static_input_scheme=is_static_input_scheme,
input_symmetric=input_symmetric,
)
kernel_type = choose_scaled_mm_linear_kernel(
config,
_POSSIBLE_INT8_KERNELS,
)
logger.info_once(
"Selected %s for %s",
kernel_type.__name__,
module_name,
scope="global",
)
return kernel_type(
config,
layer_param_names=[
"weight",
"weight_scale",
"input_scale",
"input_zero_point",
"azp_adj",
],
)

53
vllm/model_executor/layers/quantization/kernels/scaled_mm/aiter.py
View File

@@ -8,60 +8,41 @@ from vllm import _custom_ops as ops
from vllm._aiter_ops import rocm_aiter_ops
from vllm.platforms import current_platform
from .cutlass import CutlassScaledMMLinearKernel
from .ScaledMMLinearKernel import ScaledMMLinearLayerConfig
from .cutlass import CutlassInt8ScaledMMLinearKernel
from .ScaledMMLinearKernel import Int8ScaledMMLinearLayerConfig
class AiterScaledMMLinearKernel(CutlassScaledMMLinearKernel):
class AiterInt8ScaledMMLinearKernel(CutlassInt8ScaledMMLinearKernel):
@classmethod
def is_supported(
cls, compute_capability: int | None = None
) -> tuple[bool, str | None]:
if not current_platform.is_rocm():
return (
False,
"AiterScaledMMLinearKernel requires `aiter` which is not "
+ "currently supported on non-ROCm platform.",
)
if compute_capability is None:
_cc = current_platform.get_device_capability()
if _cc is not None:
compute_capability = _cc.major * 10 + _cc.minor
return False, "Requires ROCm."
if compute_capability is not None and compute_capability < 90:
return False, f"requires capability 90, got {compute_capability}"
return False, "requires compute capability 90 and above."
try:
import aiter # noqa: F401 # deliberately attempt to import aiter
except Exception:
return (
False,
"AiterScaledMMLinearKernel requires `aiter` which is not "
+ "installed on ROCm.",
)
return False, "requires `aiter` to be installed."
if not rocm_aiter_ops.is_linear_enabled():
return (
False,
"AiterScaledMMLinearKernel is disabled. "
+ "Enable by setting `VLLM_ROCM_USE_AITER=1` "
"requires setting `VLLM_ROCM_USE_AITER=1` "
+ "and `VLLM_ROCM_USE_AITER_LINEAR=1`. "
+ "`VLLM_ROCM_USE_AITER_LINEAR` default is True.",
)
return True, None
@classmethod
def can_implement(cls, c: ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
def can_implement(cls, c: Int8ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
if not c.input_symmetric:
return (
False,
"AiterScaledMMLinearKernel only supports symmetric " + "quantization.",
)
return False, "supports symmetric quantization only."
return True, None
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
super().process_weights_after_loading(layer)
def apply_weights(
self,
layer: torch.nn.Module,
@@ -69,28 +50,28 @@ class AiterScaledMMLinearKernel(CutlassScaledMMLinearKernel):
bias: torch.Tensor | None = None,
) -> torch.Tensor:
"""
`AiterScaledMMLinearKernel` implements a fused version of
`AiterInt8ScaledMMLinearKernel` implements a fused version of
`output = torch.mm((scale_a * a), (scale_b * b)).to(out_dtype)`
where scale_a * a and scale_b * b are implemented using numpy-style
broadcasting.
Currently only support per-tensor-per-tensor GEMM
and per-token-per-channel GEMM through AITER
w8a8 scaled gemm. `AiterScaledMMLinearKernel` also does not support
w8a8 scaled gemm. `AiterInt8ScaledMMLinearKernel` also does not support
ATIER block scaled GEMM and mix-precision GEMM.
"""
w_q, w_s, i_s, i_zp, azp_adj = self._get_weight_params(layer)
w_q, w_s, i_s, i_zp, azp_adj = self._get_layer_params(layer)
# ops.scaled_int8_quant supports both dynamic and static quant:
# * dynamic, i_s is None and x_s computed from x.
# * static, i_s is scalar and x_s is i_s.
symmetric = azp_adj is None
assert symmetric, (
"AiterScaledMMLinearKernel only supports symmetric quantization."
"AiterInt8ScaledMMLinearKernel only supports symmetric quantization."
)
x_q, x_s, x_zp = ops.scaled_int8_quant(x, i_s, i_zp, symmetric=symmetric)
assert x_zp is None, (
"AiterScaledMMLinearKernel only supports symmetric quantization."
"AiterInt8ScaledMMLinearKernel only supports symmetric quantization."
)
out_dtype = x.dtype
@@ -117,12 +98,12 @@ class AiterScaledMMLinearKernel(CutlassScaledMMLinearKernel):
), (
"Currently only support per-tensor-per-tensor GEMM "
+ " and per-token-per-channel GEMM through AITER"
" w8a8 scaled gemm. `AiterScaledMMLinearKernel` "
" w8a8 scaled gemm. `AiterInt8ScaledMMLinearKernel` "
+ "does not support AITER block scaled GEMM."
)
# gemm_a8w8_CK(a, b, scale_a, scale_b, bias) expects
# a to be [M, K]
# b to be [N, K]
# CutlassScaledMMLinearKernel prepare weight `w_q` in [K, N] format
# CutlassInt8ScaledMMLinearKernel prepare weight `w_q` in [K, N] format
return rocm_aiter_ops.gemm_a8w8(x_q, w_q.t(), x_s, w_s, bias, out_dtype)

71
vllm/model_executor/layers/quantization/kernels/scaled_mm/cpu.py
View File

@@ -14,24 +14,28 @@ from vllm.model_executor.layers.utils import check_cpu_sgl_kernel
from vllm.platforms import current_platform
from vllm.platforms.interface import CpuArchEnum
from .ScaledMMLinearKernel import ScaledMMLinearKernel, ScaledMMLinearLayerConfig
from .ScaledMMLinearKernel import (
Int8ScaledMMLinearKernel,
Int8ScaledMMLinearLayerConfig,
)
class CPUScaledMMLinearKernel(ScaledMMLinearKernel):
class CPUInt8ScaledMMLinearKernel(Int8ScaledMMLinearKernel):
@classmethod
def is_supported(
cls, compute_capability: int | None = None
) -> tuple[bool, str | None]:
if not current_platform.is_cpu():
return False, "Requires CPU."
return False, "requires CPU."
return True, None
@classmethod
def can_implement(cls, c: ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
def can_implement(cls, c: Int8ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
return True, None
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
weight = getattr(layer, self.w_q_name)
w_q_name, _, _, _, _ = self.layer_param_names
weight = getattr(layer, w_q_name)
dtype = weight.dtype
N, K = weight.size()
if (
@@ -49,10 +53,11 @@ class CPUScaledMMLinearKernel(ScaledMMLinearKernel):
def process_weights_for_onednn(self, layer: torch.nn.Module) -> None:
# WEIGHT
# Transpose to [K, N] for convenience
weight = getattr(layer, self.w_q_name)
w_q_name, w_s_name, i_s_name, i_zp_name, azp_adj_name = self.layer_param_names
weight = getattr(layer, w_q_name)
replace_parameter(
layer,
self.w_q_name,
w_q_name,
torch.nn.Parameter(weight.t().data, requires_grad=False),
)
@@ -61,28 +66,27 @@ class CPUScaledMMLinearKernel(ScaledMMLinearKernel):
# If we have a fused module (QKV, MLP) with per tensor scales (thus N
# scales being passed to the kernel), convert to the per-channel case.
is_fused_module = len(layer.logical_widths) > 1
weight_scale = getattr(layer, self.w_s_name)
weight_scale = getattr(layer, w_s_name)
if is_fused_module and not self.config.is_channelwise:
weight_scale = convert_to_channelwise(weight_scale, layer.logical_widths)
replace_parameter(
layer,
self.w_s_name,
w_s_name,
torch.nn.Parameter(weight_scale.data, requires_grad=False),
)
# INPUT SCALE
if self.config.is_static_input_scheme:
input_scale = getattr(layer, self.i_s_name)
input_scale = getattr(layer, i_s_name)
if self.config.input_symmetric:
replace_parameter(
layer,
self.i_s_name,
i_s_name,
torch.nn.Parameter(input_scale.max(), requires_grad=False),
)
setattr(layer, self.i_zp_name, None)
else:
input_zero_point = getattr(layer, self.i_zp_name)
input_zero_point = getattr(layer, i_zp_name)
# reconstruct the ranges
int8_traits = torch.iinfo(torch.int8)
@@ -92,20 +96,16 @@ class CPUScaledMMLinearKernel(ScaledMMLinearKernel):
scale = (range_max - range_min) / (int8_traits.max - int8_traits.min)
replace_parameter(
layer, self.i_s_name, torch.nn.Parameter(scale, requires_grad=False)
layer, i_s_name, torch.nn.Parameter(scale, requires_grad=False)
)
azp = (
(int8_traits.min - range_min / scale).round().to(dtype=torch.int32)
)
replace_parameter(
layer, self.i_zp_name, torch.nn.Parameter(azp, requires_grad=False)
layer, i_zp_name, torch.nn.Parameter(azp, requires_grad=False)
)
else:
setattr(layer, self.i_s_name, None)
setattr(layer, self.i_zp_name, None)
# Different from cutlass, oneDNN kernels only need the AZP adjustment
# term for dynamic quantization. And s_b should be folded into the
# term. Such as:
@@ -113,38 +113,37 @@ class CPUScaledMMLinearKernel(ScaledMMLinearKernel):
# s_a * (s_b * AB) - s_a * s_b * zp_a * B + bias =
# s_a * GEMM_output - s_a * zp_a * adj + bias
if not (self.config.input_symmetric and self.config.is_static_input_scheme):
weight = getattr(layer, self.w_q_name)
weight_scale = getattr(layer, self.w_s_name)
weight = getattr(layer, w_q_name)
weight_scale = getattr(layer, w_s_name)
azp_adj = weight.sum(dim=0, keepdim=True, dtype=torch.float32)
azp_adj = azp_adj * weight_scale.squeeze()
setattr(
layer,
self.azp_adj_name,
azp_adj_name,
torch.nn.Parameter(azp_adj, requires_grad=False),
)
else:
setattr(layer, self.azp_adj_name, None)
weight = getattr(layer, self.w_q_name)
weight = getattr(layer, w_q_name)
self.dnnl_handler = ops.create_onednn_scaled_mm(
weight,
getattr(layer, self.w_s_name),
getattr(layer, w_s_name),
torch.get_default_dtype(),
getattr(layer, self.i_s_name) is None,
getattr(layer, i_s_name) is None,
not self.config.input_symmetric,
32,
)
# weight is prepacked and maintained by the dnnl_handler,
# release the original weight
setattr(layer, self.w_q_name, None)
setattr(layer, w_q_name, None)
del weight
def process_weights_for_sgl(self, layer: torch.nn.Module) -> None:
w_q_name, w_s_name, _, _, _ = self.layer_param_names
# WEIGHT
weight = getattr(layer, self.w_q_name)
weight = getattr(layer, w_q_name)
packed_weight = torch.ops._C.convert_weight_packed(weight)
replace_parameter(
layer, self.w_q_name, torch.nn.Parameter(packed_weight, requires_grad=False)
layer, w_q_name, torch.nn.Parameter(packed_weight, requires_grad=False)
)
if layer.bias is not None:
@@ -156,19 +155,15 @@ class CPUScaledMMLinearKernel(ScaledMMLinearKernel):
# WEIGHT SCALE
# CPU SGL kernels only support per-channel.
# For per-tensor quant, convert to the per-channel case.
weight_scale = getattr(layer, self.w_s_name)
weight_scale = getattr(layer, w_s_name)
if not self.config.is_channelwise:
weight_scale = convert_to_channelwise(weight_scale, layer.logical_widths)
replace_parameter(
layer,
self.w_s_name,
w_s_name,
torch.nn.Parameter(weight_scale.data, requires_grad=False),
)
setattr(layer, self.i_s_name, None)
setattr(layer, self.i_zp_name, None)
setattr(layer, self.azp_adj_name, None)
def apply_weights(
self,
layer: torch.nn.Module,
@@ -187,7 +182,7 @@ class CPUScaledMMLinearKernel(ScaledMMLinearKernel):
x: torch.Tensor,
bias: torch.Tensor | None = None,
) -> torch.Tensor:
w_q, w_s, i_s, i_zp, azp_adj = self._get_weight_params(layer)
w_q, w_s, i_s, i_zp, azp_adj = self._get_layer_params(layer)
# ops.scaled_int8_quant supports both dynamic and static quant:
# * dynamic, i_s is None and x_s computed from x.
@@ -209,7 +204,7 @@ class CPUScaledMMLinearKernel(ScaledMMLinearKernel):
x: torch.Tensor,
bias: torch.Tensor | None = None,
) -> torch.Tensor:
w_q, w_s, _, _, _ = self._get_weight_params(layer)
w_q, w_s, _, _, _ = self._get_layer_params(layer)
return torch.ops._C.int8_scaled_mm_with_quant(
x,
w_q,

96
vllm/model_executor/layers/quantization/kernels/scaled_mm/cutlass.py
View File

@@ -11,35 +11,36 @@ from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
)
from vllm.platforms import current_platform
from .ScaledMMLinearKernel import ScaledMMLinearKernel, ScaledMMLinearLayerConfig
from .ScaledMMLinearKernel import (
FP8ScaledMMLinearKernel,
FP8ScaledMMLinearLayerConfig,
Int8ScaledMMLinearKernel,
Int8ScaledMMLinearLayerConfig,
)
class CutlassScaledMMLinearKernel(ScaledMMLinearKernel):
class CutlassInt8ScaledMMLinearKernel(Int8ScaledMMLinearKernel):
@classmethod
def is_supported(
cls, compute_capability: int | None = None
) -> tuple[bool, str | None]:
if not current_platform.is_cuda():
return False, "Requires CUDA."
if compute_capability is None:
_cc = current_platform.get_device_capability()
if _cc is not None:
compute_capability = _cc.major * 10 + _cc.minor
if compute_capability is not None and compute_capability < 75:
return False, f"requires capability 75, got {compute_capability}"
return False, "requires CUDA."
return True, None
@classmethod
def can_implement(cls, c: ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
def can_implement(cls, c: Int8ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
return True, None
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
w_q_name, w_s_name, i_s_name, i_zp_name, azp_adj_name = self.layer_param_names
config = self.config
# WEIGHT
# Cutlass kernels need transposed weight.
weight = getattr(layer, self.w_q_name)
weight = getattr(layer, w_q_name)
replace_parameter(
layer,
self.w_q_name,
w_q_name,
torch.nn.Parameter(weight.t().data, requires_grad=False),
)
@@ -48,28 +49,28 @@ class CutlassScaledMMLinearKernel(ScaledMMLinearKernel):
# If we have a fused module (QKV, MLP) with per tensor scales (thus N
# scales being passed to the kernel), convert to the per-channel case.
is_fused_module = len(layer.logical_widths) > 1
weight_scale = getattr(layer, self.w_s_name)
if is_fused_module and not self.config.is_channelwise:
weight_scale = getattr(layer, w_s_name)
if is_fused_module and not config.is_channelwise:
weight_scale = convert_to_channelwise(weight_scale, layer.logical_widths)
replace_parameter(
layer,
self.w_s_name,
w_s_name,
torch.nn.Parameter(weight_scale.data, requires_grad=False),
)
# INPUT SCALE
if self.config.is_static_input_scheme:
input_scale = getattr(layer, self.i_s_name)
if config.is_static_input_scheme:
input_scale = getattr(layer, i_s_name)
if self.config.input_symmetric:
if config.input_symmetric:
replace_parameter(
layer,
self.i_s_name,
i_s_name,
torch.nn.Parameter(input_scale.max(), requires_grad=False),
)
setattr(layer, self.i_zp_name, None)
setattr(layer, i_zp_name, None)
else:
input_zero_point = getattr(layer, self.i_zp_name)
input_zero_point = getattr(layer, i_zp_name)
# reconstruct the ranges
int8_traits = torch.iinfo(torch.int8)
@@ -79,38 +80,32 @@ class CutlassScaledMMLinearKernel(ScaledMMLinearKernel):
scale = (range_max - range_min) / (int8_traits.max - int8_traits.min)
replace_parameter(
layer, self.i_s_name, torch.nn.Parameter(scale, requires_grad=False)
layer, i_s_name, torch.nn.Parameter(scale, requires_grad=False)
)
# AZP loaded as int8 but used as int32
azp = (int8_traits.min - range_min / scale).to(dtype=torch.int32)
replace_parameter(
layer, self.i_zp_name, torch.nn.Parameter(azp, requires_grad=False)
layer, i_zp_name, torch.nn.Parameter(azp, requires_grad=False)
)
else:
setattr(layer, self.i_s_name, None)
setattr(layer, self.i_zp_name, None)
# azp_adj is the AZP adjustment term, used to account for weights.
# It does not depend on scales or azp, so it is the same for
# static and dynamic quantization.
# For more details, see csrc/quantization/w8a8/cutlass/Epilogues.md
# https://github.com/vllm-project/vllm/blob/main/csrc/quantization/w8a8/cutlass/Epilogues.md
if not self.config.input_symmetric:
weight = getattr(layer, self.w_q_name)
if not config.input_symmetric:
weight = getattr(layer, w_q_name)
azp_adj = weight.sum(dim=0, keepdim=True, dtype=torch.int32)
if self.config.is_static_input_scheme:
if config.is_static_input_scheme:
# cutlass_w8a8 requires azp to be folded into azp_adj
# in the per-tensor case
azp_adj = getattr(layer, self.i_zp_name) * azp_adj
azp_adj = getattr(layer, i_zp_name) * azp_adj
setattr(
layer,
self.azp_adj_name,
azp_adj_name,
torch.nn.Parameter(azp_adj, requires_grad=False),
)
else:
setattr(layer, self.azp_adj_name, None)
def apply_weights(
self,
@@ -118,7 +113,7 @@ class CutlassScaledMMLinearKernel(ScaledMMLinearKernel):
x: torch.Tensor,
bias: torch.Tensor | None = None,
) -> torch.Tensor:
w_q, w_s, i_s, i_zp, azp_adj = self._get_weight_params(layer)
w_q, w_s, i_s, i_zp, azp_adj = self._get_layer_params(layer)
# ops.scaled_int8_quant supports both dynamic and static quant:
# * dynamic, i_s is None and x_s computed from x.
@@ -145,3 +140,34 @@ class CutlassScaledMMLinearKernel(ScaledMMLinearKernel):
return ops.cutlass_scaled_mm(
x_q, w_q, scale_a=x_s, scale_b=w_s, out_dtype=x.dtype, bias=bias
)
class CutlassFP8ScaledMMLinearKernel(FP8ScaledMMLinearKernel):
@classmethod
def is_supported(
cls, compute_capability: int | None = None
) -> tuple[bool, str | None]:
if not current_platform.is_cuda():
return False, "requires CUDA."
return True, None
@classmethod
def can_implement(cls, c: FP8ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
return True, None
def apply_scaled_mm(
self,
*,
A: torch.Tensor,
B: torch.Tensor,
out_dtype: torch.dtype,
As: torch.Tensor,
Bs: torch.Tensor,
bias: torch.Tensor | None,
output_shape: list,
) -> torch.Tensor:
# Fused GEMM_DQ
output = ops.cutlass_scaled_mm(
A, B, out_dtype=out_dtype, scale_a=As, scale_b=Bs, bias=bias
)
return output.view(*output_shape)

57
vllm/model_executor/layers/quantization/kernels/scaled_mm/flashinfer.py Normal file
View File

@@ -0,0 +1,57 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import torch
from vllm.platforms import current_platform
from vllm.utils.flashinfer import flashinfer_scaled_fp8_mm, has_flashinfer
from .ScaledMMLinearKernel import (
FP8ScaledMMLinearKernel,
FP8ScaledMMLinearLayerConfig,
)
class FlashInferFP8ScaledMMLinearKernel(FP8ScaledMMLinearKernel):
@classmethod
def is_supported(
cls, compute_capability: int | None = None
) -> tuple[bool, str | None]:
if not current_platform.is_cuda():
return False, "requires CUDA."
if not has_flashinfer():
return False, "requires FlashInfer to be installed."
if compute_capability is not None and compute_capability < 100:
return False, "requires compute capability 100 and above."
return True, None
@classmethod
def can_implement(cls, c: FP8ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
per_tensor_activation_scales = (
c.activation_quant_key.scale.group_shape.is_per_tensor()
)
per_tensor_weight_scales = c.weight_quant_key.scale.group_shape.is_per_tensor()
if not (per_tensor_activation_scales and per_tensor_weight_scales):
return False, "requires per tensor activation and weight scales."
return True, None
def apply_scaled_mm(
self,
*,
A: torch.Tensor,
B: torch.Tensor,
out_dtype: torch.dtype,
As: torch.Tensor,
Bs: torch.Tensor,
bias: torch.Tensor | None,
output_shape: list,
) -> torch.Tensor:
return flashinfer_scaled_fp8_mm(
A, B, out_dtype=out_dtype, scale_a=As, scale_b=Bs, bias=bias
)

221
vllm/model_executor/layers/quantization/kernels/scaled_mm/pytorch.py Normal file
View File

@@ -0,0 +1,221 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import torch
from packaging import version
from vllm.config import CompilationMode, get_current_vllm_config
from vllm.platforms import current_platform
from .ScaledMMLinearKernel import (
FP8ScaledMMLinearKernel,
FP8ScaledMMLinearLayerConfig,
)
class TorchFP8ScaledMMLinearKernel(FP8ScaledMMLinearKernel):
"""
Base class for FP8 linear kernels using Torch.
Each subclass represents a kernel variant for
specific device capabilities and torch versions.
"""
@classmethod
def is_supported(
cls, compute_capability: int | None = None
) -> tuple[bool, str | None]:
if not (current_platform.is_cuda_alike() or current_platform.is_cpu()):
return False, "requires ROCm, CUDA or CPU."
if compute_capability is not None and compute_capability < 89:
return False, "requires compute capability 89 and above."
return True, None
def get_output_padding(self) -> int | None:
# Note: we pad the input because torch._scaled_mm is more performant
# for matrices with batch dimension > 16.
# This could change in the future.
# We also don't pad when using torch.compile,
# as it breaks with dynamic shapes.
#
# The perf gain is still relevant as of 16/1/2026
# torch version == 2.9.0. More details in the link below:
# https://github.com/vllm-project/vllm/issues/32269
vllm_config = get_current_vllm_config().compilation_config
pad_output = vllm_config.mode < CompilationMode.VLLM_COMPILE
return 17 if pad_output else None
class PerTensorTorchFP8ScaledMMLinearKernel(TorchFP8ScaledMMLinearKernel):
@classmethod
def can_implement(cls, c: FP8ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
per_tensor_activation_scales = (
c.activation_quant_key.scale.group_shape.is_per_tensor()
)
per_tensor_weight_scales = c.weight_quant_key.scale.group_shape.is_per_tensor()
if not (per_tensor_activation_scales and per_tensor_weight_scales):
return False, "requires per tensor activation and weight scales."
return True, None
def apply_scaled_mm(
self,
*,
A: torch.Tensor,
B: torch.Tensor,
out_dtype: torch.dtype,
As: torch.Tensor,
Bs: torch.Tensor,
bias: torch.Tensor | None,
output_shape: list,
) -> torch.Tensor:
output = torch._scaled_mm(
A, B, out_dtype=out_dtype, scale_a=As, scale_b=Bs, bias=bias
)
# A fix for discrepancy in scaled_mm which returns tuple
# for torch < 2.5 and a single value in torch >= 2.5
if type(output) is tuple and len(output) == 2:
output = output[0]
return torch.narrow(output, 0, 0, output_shape[0]).view(*output_shape)
class RowWiseTorchFP8ScaledMMLinearKernel(TorchFP8ScaledMMLinearKernel):
@classmethod
def is_supported(
cls, compute_capability: int | None = None
) -> tuple[bool, str | None]:
if not current_platform.is_rocm():
return False, "requires ROCm."
from vllm.platforms.rocm import on_mi3xx
if not on_mi3xx():
return False, "requires MI3xx."
if compute_capability is not None and compute_capability < 94:
return False, "requires compute capability 94 and above."
if not version.parse(torch.__version__) >= version.parse("2.7"):
return False, "requires pytorch version >=2.7."
return True, None
@classmethod
def can_implement(cls, c: FP8ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
per_tensor_activation_scales = (
c.activation_quant_key.scale.group_shape.is_per_tensor()
)
per_tensor_weight_scales = c.weight_quant_key.scale.group_shape.is_per_tensor()
if c.out_dtype == torch.float16:
# hipblaslt rowwise _scaled_mm only supports BFloat16
return False, "supports BFloat16 output data type only."
if per_tensor_activation_scales or per_tensor_weight_scales:
return False, "cannot be used with per tensor activation and weight scales."
return True, None
def apply_scaled_mm(
self,
*,
A: torch.Tensor,
B: torch.Tensor,
out_dtype: torch.dtype,
As: torch.Tensor,
Bs: torch.Tensor,
bias: torch.Tensor | None,
output_shape: list,
) -> torch.Tensor:
# Note:
# For now it has only been validated on ROCm platform.
# fp8 rowwise scaling in torch._scaled_mm is introduced in
# https://github.com/pytorch/pytorch/pull/144432 using
# hipBLASLt and ROCm 6.3, which only exists in torch 2.7 and above.
#
# For CUDA platform please validate if the torch._scaled_mm supports
# rowwise scaled GEMM before using it
# Fused GEMM_DQ Rowwise GEMM
output = torch._scaled_mm(
A,
B,
out_dtype=out_dtype,
scale_a=As,
scale_b=Bs.t(),
bias=bias,
)
return torch.narrow(output, 0, 0, output_shape[0]).view(*output_shape)
class ChannelWiseTorchFP8ScaledMMLinearKernel(TorchFP8ScaledMMLinearKernel):
@classmethod
def can_implement(cls, c: FP8ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
per_tensor_activation_scales = (
c.activation_quant_key.scale.group_shape.is_per_tensor()
)
per_tensor_weight_scales = c.weight_quant_key.scale.group_shape.is_per_tensor()
if per_tensor_activation_scales and per_tensor_weight_scales:
return False, "cannot be used with per tensor activation and weight scales."
return True, None
def apply_scaled_mm(
self,
*,
A: torch.Tensor,
B: torch.Tensor,
out_dtype: torch.dtype,
As: torch.Tensor,
Bs: torch.Tensor,
bias: torch.Tensor | None,
output_shape: list,
) -> torch.Tensor:
# Use unfused DQ due to limitations with scaled_mm
# Symmetric quantized GEMM by definition computes the following:
# C = (s_x * X) (s_w * W) + bias
# This is equivalent to dequantizing the weights and activations
# before applying a GEMM.
#
# In order to compute quantized operands, a quantized kernel
# will rewrite the above like so:
# C = s_w * s_x * (X * W) + bias
#
# For the scaled_mm fallback case, we break this down, since it
# does not support s_w being a vector.
# Input scaling factors are no longer optional in _scaled_mm starting
# from pytorch 2.5. Allocating a dummy tensor to pass as scales
dummy_tensor = torch.ones(1, dtype=torch.float32, device=A.device)
# GEMM
# This computes C = (X * W).
# Output in fp32 to allow subsequent ops to happen in-place
output = torch._scaled_mm(
A,
B,
scale_a=dummy_tensor,
scale_b=dummy_tensor,
out_dtype=torch.float32,
)
# A fix for discrepancy in scaled_mm which returns tuple
# for torch < 2.5 and a single value in torch >= 2.5
if type(output) is tuple and len(output) == 2:
output = output[0]
# Unpad (undo num_token_padding)
output = torch.narrow(output, 0, 0, output_shape[0])
x_scale = torch.narrow(As, 0, 0, output_shape[0])
# DQ
# C = sw * sx * (X * W) + bias
output = output * x_scale * Bs.t()
if bias is not None:
output = output + bias
return output.to(out_dtype).view(*output_shape)

117
vllm/model_executor/layers/quantization/kernels/scaled_mm/rocm.py Normal file
View File

@@ -0,0 +1,117 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import torch
import vllm.envs as envs
from vllm import _custom_ops as ops
from vllm.platforms import current_platform
from vllm.utils.platform_utils import get_cu_count
from vllm.utils.torch_utils import direct_register_custom_op
from .ScaledMMLinearKernel import (
FP8ScaledMMLinearKernel,
FP8ScaledMMLinearLayerConfig,
)
def rocm_per_tensor_float_w8a8_scaled_mm_impl(
A: torch.Tensor,
B: torch.Tensor,
out_dtype: torch.dtype,
As: torch.Tensor,
Bs: torch.Tensor,
bias: torch.Tensor,
) -> torch.Tensor:
if (
A.shape[0] == 1
and B.shape[1] % 16 == 0
and ((bias is None) or (bias.dtype == out_dtype))
):
output = ops.wvSplitKQ(
B.t(),
A,
out_dtype,
As,
Bs,
get_cu_count(),
bias,
)
# Fallback
else:
output = torch._scaled_mm(
A,
B,
out_dtype=out_dtype,
scale_a=As,
scale_b=Bs,
bias=bias,
)
return output
def rocm_per_tensor_float_w8a8_scaled_mm_fake(
A: torch.Tensor,
B: torch.Tensor,
out_dtype: torch.dtype,
As: torch.Tensor,
Bs: torch.Tensor,
bias: torch.Tensor,
) -> torch.Tensor:
return A.new_empty((*A.shape[:-1], B.shape[1]), dtype=out_dtype)
if current_platform.is_rocm():
direct_register_custom_op(
op_name="rocm_per_tensor_float_w8a8_scaled_mm_impl",
op_func=rocm_per_tensor_float_w8a8_scaled_mm_impl,
fake_impl=rocm_per_tensor_float_w8a8_scaled_mm_fake,
)
class ROCmFP8ScaledMMLinearKernel(FP8ScaledMMLinearKernel):
@classmethod
def is_supported(
cls, compute_capability: int | None = None
) -> tuple[bool, str | None]:
if not current_platform.is_rocm():
return False, "requires ROCm."
from vllm.platforms.rocm import on_mi3xx
if not on_mi3xx():
return False, "requires MI3xx."
if not envs.VLLM_ROCM_USE_SKINNY_GEMM:
return False, "requires VLLM_ROCM_USE_SKINNY_GEMM to be enabled."
return True, None
@classmethod
def can_implement(cls, c: FP8ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
per_tensor_activation_scales = (
c.activation_quant_key.scale.group_shape.is_per_tensor()
)
per_tensor_weight_scales = c.weight_quant_key.scale.group_shape.is_per_tensor()
if not (per_tensor_activation_scales and per_tensor_weight_scales):
return False, "requires per tensor activation and weight scales."
return True, None
def apply_scaled_mm(
self,
*,
A: torch.Tensor,
B: torch.Tensor,
out_dtype: torch.dtype,
As: torch.Tensor,
Bs: torch.Tensor,
bias: torch.Tensor | None,
output_shape: list,
) -> torch.Tensor:
output = torch.ops.vllm.rocm_per_tensor_float_w8a8_scaled_mm_impl(
A, B, out_dtype, As, Bs, bias
)
return torch.narrow(output, 0, 0, A.shape[0]).view(*output_shape)

41
vllm/model_executor/layers/quantization/kernels/scaled_mm/triton.py
View File

@@ -14,30 +14,35 @@ from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
)
from vllm.platforms import current_platform
from .ScaledMMLinearKernel import ScaledMMLinearKernel, ScaledMMLinearLayerConfig
from .cutlass import CutlassInt8ScaledMMLinearKernel
from .ScaledMMLinearKernel import (
Int8ScaledMMLinearLayerConfig,
)
class TritonScaledMMLinearKernel(ScaledMMLinearKernel):
class TritonInt8ScaledMMLinearKernel(CutlassInt8ScaledMMLinearKernel):
@classmethod
def is_supported(
cls, compute_capability: int | None = None
) -> tuple[bool, str | None]:
if current_platform.is_cuda_alike():
return True, None
return False, "Requires ROCm or CUDA."
return False, "requires ROCm or CUDA."
@classmethod
def can_implement(cls, c: ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
def can_implement(cls, c: Int8ScaledMMLinearLayerConfig) -> tuple[bool, str | None]:
if not c.input_symmetric:
return False, "Only symmetric input is supported."
return False, "supports symmetric input only."
return True, None
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
weight = getattr(layer, self.w_q_name)
w_q, _, i_s, _, _ = self._get_layer_params(layer)
w_q_name, w_s_name, i_s_name, i_zp_name, azp_adj_name = self.layer_param_names
replace_parameter(
layer,
self.w_q_name,
torch.nn.Parameter(weight.t().data, requires_grad=False),
w_q_name,
torch.nn.Parameter(w_q.t().data, requires_grad=False),
)
# WEIGHT SCALE
@@ -45,29 +50,29 @@ class TritonScaledMMLinearKernel(ScaledMMLinearKernel):
# If we have a fused module (QKV, MLP) with per tensor scales (thus N
# scales being passed to the kernel), convert to the per-channel case.
is_fused_module = len(layer.logical_widths) > 1
weight_scale = getattr(layer, self.w_s_name)
weight_scale = getattr(layer, w_s_name)
if is_fused_module and not self.config.is_channelwise:
weight_scale = convert_to_channelwise(weight_scale, layer.logical_widths)
replace_parameter(
layer,
self.w_s_name,
w_s_name,
torch.nn.Parameter(weight_scale.data, requires_grad=False),
)
# INPUT SCALE
if self.config.is_static_input_scheme:
input_scale = getattr(layer, self.i_s_name)
assert i_s is not None
replace_parameter(
layer,
self.i_s_name,
torch.nn.Parameter(input_scale.max(), requires_grad=False),
i_s_name,
torch.nn.Parameter(i_s.max(), requires_grad=False),
)
setattr(layer, self.i_zp_name, None)
setattr(layer, i_zp_name, None)
else:
setattr(layer, self.i_s_name, None)
setattr(layer, self.i_zp_name, None)
setattr(layer, i_s_name, None)
setattr(layer, i_zp_name, None)
setattr(layer, self.azp_adj_name, None)
setattr(layer, azp_adj_name, None)
def apply_weights(
self,
@@ -75,7 +80,7 @@ class TritonScaledMMLinearKernel(ScaledMMLinearKernel):
x: torch.Tensor,
bias: torch.Tensor | None = None,
) -> torch.Tensor:
w_q, w_s, i_s, i_zp, azp_adj = self._get_weight_params(layer)
w_q, w_s, i_s, i_zp, _ = self._get_layer_params(layer)
x_q, x_s, x_zp = ops.scaled_int8_quant(
x.contiguous(), i_s, i_zp, symmetric=True

37
vllm/model_executor/layers/quantization/modelopt.py
View File

@@ -49,6 +49,9 @@ from vllm.model_executor.layers.quantization.base_config import (
QuantizationConfig,
QuantizeMethodBase,
)
from vllm.model_executor.layers.quantization.kernels.scaled_mm import (
init_fp8_linear_kernel,
)
from vllm.model_executor.layers.quantization.kv_cache import BaseKVCacheMethod
from vllm.model_executor.layers.quantization.utils.flashinfer_fp4_moe import (
build_flashinfer_fp4_cutlass_moe_prepare_finalize,
@@ -78,10 +81,12 @@ from vllm.model_executor.layers.quantization.utils.quant_utils import (
GroupShape,
cutlass_fp4_supported,
is_layer_skipped,
kFp8DynamicTokenSym,
kFp8StaticTensorSym,
kFp8StaticTokenSym,
swizzle_blockscale,
)
from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
Fp8LinearOp,
cutlass_block_fp8_supported,
requantize_with_max_scale,
)
@@ -438,8 +443,11 @@ class ModelOptFp8LinearMethod(LinearMethodBase):
def __init__(self, quant_config: ModelOptFp8Config) -> None:
self.quant_config = quant_config
self.fp8_linear = Fp8LinearOp(
act_quant_static=True, act_quant_group_shape=GroupShape.PER_TENSOR
self.fp8_linear = init_fp8_linear_kernel(
activation_quant_key=kFp8StaticTensorSym,
weight_quant_key=kFp8StaticTensorSym,
out_dtype=torch.get_default_dtype(),
module_name=self.__class__.__name__,
)
def create_weights(
@@ -507,13 +515,7 @@ class ModelOptFp8LinearMethod(LinearMethodBase):
x: torch.Tensor,
bias: torch.Tensor | None = None,
) -> torch.Tensor:
return self.fp8_linear.apply(
input=x,
weight=layer.weight,
weight_scale=layer.weight_scale,
input_scale=layer.input_scale,
bias=bias,
)
return self.fp8_linear.apply_weights(layer, x, bias)
class ModelOptFp8PcPtLinearMethod(LinearMethodBase):
@@ -527,8 +529,11 @@ class ModelOptFp8PcPtLinearMethod(LinearMethodBase):
def __init__(self, quant_config: ModelOptFp8Config) -> None:
self.quant_config = quant_config
self.fp8_linear = Fp8LinearOp(
act_quant_static=False, act_quant_group_shape=GroupShape.PER_TOKEN
self.fp8_linear = init_fp8_linear_kernel(
activation_quant_key=kFp8DynamicTokenSym,
weight_quant_key=kFp8StaticTokenSym,
out_dtype=torch.get_default_dtype(),
module_name=self.__class__.__name__,
)
def create_weights(
@@ -585,13 +590,7 @@ class ModelOptFp8PcPtLinearMethod(LinearMethodBase):
x: torch.Tensor,
bias: torch.Tensor | None = None,
) -> torch.Tensor:
return self.fp8_linear.apply(
input=x,
weight=layer.weight,
weight_scale=layer.weight_scale,
input_scale=None,
bias=bias,
)
return self.fp8_linear.apply_weights(layer, x, bias)
class ModelOptFp8PbWoLinearMethod(LinearMethodBase):

27
vllm/model_executor/layers/quantization/ptpc_fp8.py
View File

@@ -17,11 +17,13 @@ from vllm.model_executor.layers.quantization.fp8 import (
Fp8KVCacheMethod,
Fp8LinearMethod,
)
from vllm.model_executor.layers.quantization.utils.quant_utils import (
GroupShape,
is_layer_skipped,
from vllm.model_executor.layers.quantization.kernels.scaled_mm import (
init_fp8_linear_kernel,
)
from vllm.model_executor.layers.quantization.utils.quant_utils import (
is_layer_skipped,
kFp8DynamicTokenSym,
)
from vllm.model_executor.layers.quantization.utils.w8a8_utils import Fp8LinearOp
from vllm.platforms import current_platform
ACTIVATION_SCHEMES = ["static", "dynamic"]
@@ -97,9 +99,11 @@ class PTPCFp8LinearMethod(Fp8LinearMethod):
)
super().__init__(quant_config=quant_config)
# Force weight quantization
self.quant_config.is_checkpoint_fp8_serialized = False
self.fp8_linear = Fp8LinearOp(
act_quant_static=False, act_quant_group_shape=GroupShape.PER_TOKEN
self.fp8_linear = init_fp8_linear_kernel(
activation_quant_key=kFp8DynamicTokenSym,
weight_quant_key=kFp8DynamicTokenSym,
out_dtype=torch.get_default_dtype(),
module_name=self.__class__.__name__,
)
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
@@ -130,11 +134,4 @@ class PTPCFp8LinearMethod(Fp8LinearMethod):
x: torch.Tensor,
bias: torch.Tensor | None = None,
) -> torch.Tensor:
return self.fp8_linear.apply(
input=x,
weight=layer.weight,
weight_scale=layer.weight_scale,
input_scale=None,
input_scale_ub=None,
bias=bias,
)
return self.fp8_linear.apply_weights(layer, x, bias)

47
vllm/model_executor/layers/quantization/quark/schemes/quark_w8a8_fp8.py
View File

@@ -7,10 +7,18 @@ from typing import Any, cast
import torch
from torch.nn import Parameter
from vllm.logger import init_logger
from vllm.model_executor.layers.quantization.kernels.scaled_mm import (
init_fp8_linear_kernel,
)
from vllm.model_executor.layers.quantization.quark.schemes import QuarkScheme
from vllm.model_executor.layers.quantization.utils.quant_utils import GroupShape
from vllm.model_executor.layers.quantization.utils.quant_utils import (
GroupShape,
kFp8DynamicTokenSym,
kFp8StaticTensorSym,
kFp8StaticTokenSym,
)
from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
Fp8LinearOp,
normalize_e4m3fn_to_e4m3fnuz,
requantize_with_max_scale,
)
@@ -23,6 +31,8 @@ from vllm.platforms import current_platform
__all__ = ["QuarkW8A8Fp8"]
logger = init_logger(__name__)
class QuarkW8A8Fp8(QuarkScheme):
def __init__(
@@ -35,15 +45,16 @@ class QuarkW8A8Fp8(QuarkScheme):
self.is_static_input_scheme = not cast(bool, input_config.get("is_dynamic"))
self.input_qscheme = cast(str, input_config.get("qscheme"))
per_token = (
per_token_activation = (
not self.is_static_input_scheme and self.input_qscheme == "per_channel"
)
self.act_quant_group_shape = (
GroupShape.PER_TOKEN if per_token else GroupShape.PER_TENSOR
per_token_weight = self.weight_qscheme == "per_channel"
self.activation_quant_key = (
kFp8DynamicTokenSym if per_token_activation else kFp8StaticTensorSym
)
self.fp8_linear = Fp8LinearOp(
act_quant_static=self.is_static_input_scheme,
act_quant_group_shape=self.act_quant_group_shape,
self.weight_quant_key = (
kFp8StaticTokenSym if per_token_weight else kFp8StaticTensorSym
)
self.out_dtype = torch.get_default_dtype()
@@ -94,7 +105,7 @@ class QuarkW8A8Fp8(QuarkScheme):
layer.input_scale = Parameter(input_scale, requires_grad=False)
else:
weight_scale = layer.weight_scale.data
if self.act_quant_group_shape == GroupShape.PER_TOKEN:
if self.activation_quant_key.scale.group_shape == GroupShape.PER_TOKEN:
weight_scale = weight_scale.view(-1, 1)
layer.weight = Parameter(weight.t(), requires_grad=False)
# required by torch.compile to be torch.nn.Parameter
@@ -106,8 +117,6 @@ class QuarkW8A8Fp8(QuarkScheme):
# INPUT SCALE
if self.is_static_input_scheme:
layer.input_scale = Parameter(layer.input_scale.max(), requires_grad=False)
else:
layer.input_scale = None
def create_weights(
self,
@@ -163,17 +172,17 @@ class QuarkW8A8Fp8(QuarkScheme):
input_scale[:] = torch.finfo(torch.float32).min
layer.register_parameter("input_scale", input_scale)
self.fp8_linear = init_fp8_linear_kernel(
activation_quant_key=self.activation_quant_key,
weight_quant_key=self.weight_quant_key,
out_dtype=torch.get_default_dtype(),
module_name=self.__class__.__name__,
)
def apply_weights(
self,
layer: torch.nn.Module,
x: torch.Tensor,
bias: torch.Tensor | None = None,
) -> torch.Tensor:
return self.fp8_linear.apply(
input=x,
weight=layer.weight,
weight_scale=layer.weight_scale,
out_dtype=self.out_dtype,
input_scale=layer.input_scale,
bias=bias,
)
return self.fp8_linear.apply_weights(layer, x, bias)

30
vllm/model_executor/layers/quantization/quark/schemes/quark_w8a8_int8.py
View File

@@ -7,8 +7,7 @@ import torch
from vllm.logger import init_logger
from vllm.model_executor.layers.quantization.kernels.scaled_mm import (
ScaledMMLinearLayerConfig,
choose_scaled_mm_linear_kernel,
init_int8_linear_kernel,
)
from vllm.model_executor.layers.quantization.quark.schemes import QuarkScheme
from vllm.model_executor.parameter import (
@@ -22,8 +21,6 @@ logger = init_logger(__name__)
class QuarkW8A8Int8(QuarkScheme):
_kernel_backends_being_used: set[str] = set()
def __init__(
self,
qscheme: str,
@@ -50,18 +47,13 @@ class QuarkW8A8Int8(QuarkScheme):
):
layer.logical_widths = output_partition_sizes
scaled_mm_linear_kernel_config = ScaledMMLinearLayerConfig(
self.kernel = init_int8_linear_kernel(
is_channelwise=(self.qscheme == "per_channel"),
is_static_input_scheme=(self.is_static_input_scheme is True),
input_symmetric=(self.input_symmetric is True),
module_name=self.__class__.__name__,
)
kernel_type = choose_scaled_mm_linear_kernel(scaled_mm_linear_kernel_config)
if kernel_type.__name__ not in self._kernel_backends_being_used:
logger.info("Using %s for QuarkW8A8Int8", kernel_type.__name__)
self._kernel_backends_being_used.add(kernel_type.__name__)
# WEIGHT
weight = ModelWeightParameter(
data=torch.empty(
@@ -102,25 +94,21 @@ class QuarkW8A8Int8(QuarkScheme):
layer.register_parameter("weight_zero_point", weight_zero_point)
# INPUT SCALE
input_zero_point = None
input_scale = None
if self.is_static_input_scheme:
input_scale = BasevLLMParameter(
data=torch.empty(1, dtype=torch.float32), weight_loader=weight_loader
)
layer.register_parameter("input_scale", input_scale)
input_zero_point = BasevLLMParameter(
data=torch.empty(1, dtype=torch.int8), weight_loader=weight_loader
)
layer.register_parameter("input_zero_point", input_zero_point)
self.kernel = kernel_type(
c=scaled_mm_linear_kernel_config,
w_q_param_name="weight",
w_s_param_name="weight_scale",
i_s_param_name="input_scale",
i_zp_param_name="input_zero_point",
azp_adj_param_name="azp_adj",
)
layer.register_parameter("input_scale", input_scale)
layer.register_parameter("input_zero_point", input_zero_point)
if not hasattr(layer, "azp_adj"):
layer.register_parameter("azp_adj", None)
# Checkpoints are serialized in quark format, which is
# different from the format the kernel may want. Handle repacking here.

3
vllm/model_executor/layers/quantization/utils/quant_utils.py
View File

@@ -123,6 +123,9 @@ kFp8StaticTensorSym = QuantKey(FP8_DTYPE, kStaticTensorScale, symmetric=True)
kDynamicTensorScale = ScaleDesc(torch.float32, False, GroupShape.PER_TENSOR)
kFp8DynamicTensorSym = QuantKey(FP8_DTYPE, kDynamicTensorScale, symmetric=True)
kStaticTokenScale = ScaleDesc(torch.float32, True, GroupShape.PER_TOKEN)
kFp8StaticTokenSym = QuantKey(FP8_DTYPE, kStaticTokenScale, symmetric=True)
kDynamicTokenScale = ScaleDesc(torch.float32, False, GroupShape.PER_TOKEN)
kFp8DynamicTokenSym = QuantKey(FP8_DTYPE, kDynamicTokenScale, symmetric=True)

378
vllm/model_executor/layers/quantization/utils/w8a8_utils.py
View File

@@ -1,34 +1,11 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from collections.abc import Callable
import torch
from packaging import version
from vllm import _custom_ops as ops
from vllm import envs
from vllm.config import CompilationMode, get_current_vllm_config
from vllm.model_executor.layers.quantization.input_quant_fp8 import QuantFP8
from vllm.model_executor.layers.quantization.utils.quant_utils import GroupShape
from vllm.platforms import current_platform
from vllm.utils.flashinfer import flashinfer_scaled_fp8_mm, has_flashinfer
from vllm.utils.platform_utils import get_cu_count
from vllm.utils.torch_utils import direct_register_custom_op
# Input scaling factors are no longer optional in _scaled_mm starting
# from pytorch 2.5. Allocating a dummy tensor to pass as input_scale
TORCH_DEVICE_IDENTITY = None
# The condition to determine if it is on a platform that supports
# torch._scaled_mm rowwise feature.
# The condition is determined once as the operations
# are time-consuming.
USE_ROWWISE_TORCH_SCALED_MM = (
current_platform.is_rocm()
and version.parse(torch.__version__) >= version.parse("2.7")
and current_platform.has_device_capability(94)
)
def sparse_cutlass_supported() -> bool:
@@ -140,361 +117,6 @@ def requantize_with_max_scale(
return max_w_scale, weight
def maybe_create_device_identity():
# Allocate dummy ones tensor for torch._scaled_mm
global TORCH_DEVICE_IDENTITY
if TORCH_DEVICE_IDENTITY is None:
TORCH_DEVICE_IDENTITY = torch.ones(1, dtype=torch.float32)
def cutlass_w8a8_scaled_mm(
*,
qinput: torch.Tensor,
weight: torch.Tensor,
out_dtype: torch.dtype,
scale_a: torch.Tensor,
scale_b: torch.Tensor,
bias: torch.Tensor,
output_shape: list,
**kwargs,
) -> torch.Tensor:
# Fused GEMM_DQ
output = ops.cutlass_scaled_mm(
qinput, weight, out_dtype=out_dtype, scale_a=scale_a, scale_b=scale_b, bias=bias
)
return output.view(*output_shape)
def flashinfer_w8a8_scaled_mm(
*,
qinput: torch.Tensor,
weight: torch.Tensor,
out_dtype: torch.dtype,
scale_a: torch.Tensor,
scale_b: torch.Tensor,
bias: torch.Tensor,
output_shape: list,
**kwargs,
) -> torch.Tensor:
return flashinfer_scaled_fp8_mm(
qinput, weight, out_dtype=out_dtype, scale_a=scale_a, scale_b=scale_b, bias=bias
)
def rocm_per_tensor_w8a8_scaled_mm_impl(
qinput: torch.Tensor,
weight: torch.Tensor,
out_dtype: torch.dtype,
scale_a: torch.Tensor,
scale_b: torch.Tensor,
bias: torch.Tensor,
) -> torch.Tensor:
from vllm.platforms.rocm import on_mi3xx
if (
envs.VLLM_ROCM_USE_SKINNY_GEMM
and on_mi3xx()
and qinput.shape[0] == 1
and qinput.shape[1] % 16 == 0
and ((bias is None) or (bias.dtype == out_dtype))
):
output = ops.wvSplitKQ(
weight.t(),
qinput,
out_dtype,
scale_a,
scale_b,
get_cu_count(),
bias,
)
else:
output = torch._scaled_mm(
qinput,
weight,
out_dtype=out_dtype,
scale_a=scale_a,
scale_b=scale_b,
bias=bias,
)
return output
def rocm_per_tensor_w8a8_scaled_mm_fake(
qinput: torch.Tensor,
weight: torch.Tensor,
out_dtype: torch.dtype,
scale_a: torch.Tensor,
scale_b: torch.Tensor,
bias: torch.Tensor,
) -> torch.Tensor:
return qinput.new_empty((*qinput.shape[:-1], weight.shape[1]), dtype=out_dtype)
def rocm_per_tensor_w8a8_scaled_mm(
*,
qinput: torch.Tensor,
weight: torch.Tensor,
out_dtype: torch.dtype,
scale_a: torch.Tensor,
scale_b: torch.Tensor,
bias: torch.Tensor,
output_shape: list,
) -> torch.Tensor:
output = torch.ops.vllm.rocm_per_tensor_w8a8_scaled_mm_impl(
qinput, weight, out_dtype, scale_a, scale_b, bias
)
return torch.narrow(output, 0, 0, qinput.shape[0]).view(*output_shape)
direct_register_custom_op(
op_name="rocm_per_tensor_w8a8_scaled_mm_impl",
op_func=rocm_per_tensor_w8a8_scaled_mm_impl,
fake_impl=rocm_per_tensor_w8a8_scaled_mm_fake,
)
def torch_per_tensor_w8a8_scaled_mm(
*,
qinput: torch.Tensor,
weight: torch.Tensor,
out_dtype: torch.dtype,
scale_a: torch.Tensor,
scale_b: torch.Tensor,
bias: torch.Tensor,
output_shape: list,
) -> torch.Tensor:
output = torch._scaled_mm(
qinput, weight, out_dtype=out_dtype, scale_a=scale_a, scale_b=scale_b, bias=bias
)
# A fix for discrepancy in scaled_mm which returns tuple
# for torch < 2.5 and a single value in torch >= 2.5
if type(output) is tuple and len(output) == 2:
output = output[0]
return torch.narrow(output, 0, 0, qinput.shape[0]).view(*output_shape)
def torch_per_token_w8a8_scaled_mm(
*,
qinput: torch.Tensor,
weight: torch.Tensor,
out_dtype: torch.dtype,
scale_a: torch.Tensor,
scale_b: torch.Tensor,
bias: torch.Tensor,
output_shape: list,
**kwargs,
) -> torch.Tensor:
# Note: Callers of this function should check USE_ROWWISE_TORCH_SCALED_MM
# when using it.
# For now it has only been validated on ROCm platform.
# fp8 rowwise scaling in torch._scaled_mm is introduced in
# https://github.com/pytorch/pytorch/pull/144432 using
# hipBLASLt and ROCm 6.3, which only exists in torch 2.7 and above.
#
# For CUDA platform please validate if the torch._scaled_mm supports
# rowwise scaled GEMM before using it
# Fused GEMM_DQ Rowwise GEMM
output = torch._scaled_mm(
qinput,
weight,
out_dtype=out_dtype,
scale_a=scale_a,
scale_b=scale_b.t(),
bias=bias,
)
output = torch.narrow(output, 0, 0, qinput.shape[0])
output = output.view(*output_shape)
return output
def torch_channelwise_w8a8_scaled_mm(
*,
qinput: torch.Tensor,
weight: torch.Tensor,
out_dtype: torch.dtype,
scale_a: torch.Tensor,
scale_b: torch.Tensor,
bias: torch.Tensor,
output_shape: list,
**kwargs,
) -> torch.Tensor:
# Use unfused DQ due to limitations with scaled_mm
# Symmetric quantized GEMM by definition computes the following:
# C = (s_x * X) (s_w * W) + bias
# This is equivalent to dequantizing the weights and activations
# before applying a GEMM.
#
# In order to compute quantized operands, a quantized kernel
# will rewrite the above like so:
# C = s_w * s_x * (X * W) + bias
#
# For the scaled_mm fallback case, we break this down, since it
# does not support s_w being a vector.
# GEMM
# This computes C = (X * W).
# Output in fp32 to allow subsequent ops to happen in-place
output = torch._scaled_mm(
qinput,
weight,
scale_a=TORCH_DEVICE_IDENTITY,
scale_b=TORCH_DEVICE_IDENTITY,
out_dtype=torch.float32,
)
# A fix for discrepancy in scaled_mm which returns tuple
# for torch < 2.5 and a single value in torch >= 2.5
if type(output) is tuple and len(output) == 2:
output = output[0]
# Unpad (undo num_token_padding)
output = torch.narrow(output, 0, 0, qinput.shape[0])
x_scale = torch.narrow(scale_a, 0, 0, qinput.shape[0])
# DQ
# C = sw * sx * (X * W) + bias
output = output * x_scale * scale_b.t()
if bias is not None:
output = output + bias
return output.to(out_dtype).view(*output_shape)
def dispatch_w8a8_scaled_mm(
preferred_backend: str, per_tensor_weights: bool, per_tensor_activations: bool
) -> Callable[..., torch.Tensor]:
if per_tensor_weights and per_tensor_activations:
if preferred_backend == "rocm":
return rocm_per_tensor_w8a8_scaled_mm
if preferred_backend == "flashinfer":
return flashinfer_w8a8_scaled_mm
if preferred_backend == "cutlass":
return cutlass_w8a8_scaled_mm
return torch_per_tensor_w8a8_scaled_mm
# cutlass_scaled_mm supports per tensor/channel W and per tensor/token A
if preferred_backend == "cutlass" or preferred_backend == "flashinfer":
return cutlass_w8a8_scaled_mm
# If torch.scaled_mm supports per-channel (weights) per-token (inputs)
if (
not per_tensor_weights
and not per_tensor_activations
and USE_ROWWISE_TORCH_SCALED_MM
):
return torch_per_token_w8a8_scaled_mm
# Normally, torch.scaled_mm supports per tensor weights + activations only
# so fallback to naive if per channel or per token
return torch_channelwise_w8a8_scaled_mm
# TODO(luka): follow similar pattern for marlin and block-fp8-linear
# https://github.com/vllm-project/vllm/issues/14397
class Fp8LinearOp:
"""
This class executes a FP8 linear layer using cutlass if supported and
torch.scaled_mm otherwise.
It needs to be a class instead of a method so that config can be read
in the __init__ method, as reading config is not allowed inside forward.
"""
def __init__(
self,
act_quant_static: bool,
act_quant_group_shape: GroupShape = GroupShape.PER_TENSOR,
pad_output: bool | None = None,
):
if current_platform.is_rocm():
self.preferred_backend = "rocm"
elif current_platform.is_cuda() and cutlass_fp8_supported():
if has_flashinfer() and current_platform.has_device_capability(100):
self.preferred_backend = "flashinfer"
else:
self.preferred_backend = "cutlass"
else:
self.preferred_backend = "torch"
# Note: we pad the input because torch._scaled_mm is more performant
# for matrices with batch dimension > 16.
# This could change in the future.
# We also don't pad when using torch.compile,
# as it breaks with dynamic shapes.
if pad_output is None:
config = get_current_vllm_config().compilation_config
pad_output = (
config.mode < CompilationMode.VLLM_COMPILE
and self.preferred_backend == "torch"
)
self.output_padding = 17 if pad_output else None
self.act_quant_static = act_quant_static
self.act_quant_group_shape = act_quant_group_shape
self.quant_fp8 = QuantFP8(
static=act_quant_static,
group_shape=act_quant_group_shape,
num_token_padding=self.output_padding,
)
def apply(
self,
input: torch.Tensor,
weight: torch.Tensor,
weight_scale: torch.Tensor,
out_dtype: torch.dtype | None = None,
input_scale: torch.Tensor | None = None,
input_scale_ub: torch.Tensor | None = None,
bias: torch.Tensor | None = None,
) -> torch.Tensor:
# ops.scaled_fp8_quant supports both dynamic and static quant.
# If dynamic, layer.input_scale is None and x_scale computed from x.
# If static, layer.input_scale is scalar and x_scale is input_scale.
# View input as 2D matrix for fp8 methods
input_2d = input.view(-1, input.shape[-1])
output_shape = [*input.shape[:-1], weight.shape[1]]
if out_dtype is None:
out_dtype = input.dtype
# If input not quantized
# TODO(luka) remove this path if not used anymore
if input.dtype != current_platform.fp8_dtype():
qinput, x_scale = self.quant_fp8(
input_2d,
input_scale,
input_scale_ub,
)
else:
qinput, x_scale = input_2d, input_scale
# Must have dim() conditions
# In per-token quant scenario, when the number of token is 1,
# the scale will only have 1 elements.
# Without checking the dim(),
# we cannot distingushes between per-tensor and per-token quant.
# Example:
# When the number of token is 1, per-token scale is [[1]]
# When per-tensor scale is [1] or ().
per_tensor_weights = weight_scale.numel() == 1
per_tensor_activations = (x_scale.numel() == 1) and x_scale.dim() < 2
# TODO(luka) do this dispatch during init (after ScaledMM refactor)
w8a8_scaled_mm_func = dispatch_w8a8_scaled_mm(
self.preferred_backend, per_tensor_weights, per_tensor_activations
)
return w8a8_scaled_mm_func(
qinput=qinput,
weight=weight,
out_dtype=out_dtype,
scale_a=x_scale,
scale_b=weight_scale,
bias=bias,
output_shape=output_shape,
)
def normalize_e4m3fn_to_e4m3fnuz(
weight: torch.Tensor,
weight_scale: torch.Tensor,