Convert formatting to use ruff instead of yapf + isort (#26247)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
@@ -9,30 +9,46 @@ from torch.nn import Parameter
|
||||
import vllm.model_executor.layers.fused_moe # noqa
|
||||
from vllm import _custom_ops as ops
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.fused_moe.config import (FusedMoEConfig,
|
||||
FusedMoEQuantConfig)
|
||||
from vllm.model_executor.layers.fused_moe.config import (
|
||||
FusedMoEConfig,
|
||||
FusedMoEQuantConfig,
|
||||
)
|
||||
from vllm.model_executor.layers.fused_moe.layer import (
|
||||
FusedMoE, FusedMoEMethodBase, FusedMoeWeightScaleSupported,
|
||||
UnquantizedFusedMoEMethod)
|
||||
from vllm.model_executor.layers.linear import (LinearBase, LinearMethodBase,
|
||||
UnquantizedLinearMethod,
|
||||
set_weight_attrs)
|
||||
FusedMoE,
|
||||
FusedMoEMethodBase,
|
||||
FusedMoeWeightScaleSupported,
|
||||
UnquantizedFusedMoEMethod,
|
||||
)
|
||||
from vllm.model_executor.layers.linear import (
|
||||
LinearBase,
|
||||
LinearMethodBase,
|
||||
UnquantizedLinearMethod,
|
||||
set_weight_attrs,
|
||||
)
|
||||
from vllm.model_executor.layers.quantization import QuantizationMethods
|
||||
from vllm.model_executor.layers.quantization.awq import (AWQConfig,
|
||||
is_layer_skipped_awq)
|
||||
from vllm.model_executor.layers.quantization.awq import AWQConfig, is_layer_skipped_awq
|
||||
from vllm.model_executor.layers.quantization.base_config import (
|
||||
QuantizationConfig, QuantizeMethodBase)
|
||||
QuantizationConfig,
|
||||
QuantizeMethodBase,
|
||||
)
|
||||
from vllm.model_executor.layers.quantization.utils import replace_parameter
|
||||
from vllm.model_executor.layers.quantization.utils.marlin_utils import (
|
||||
apply_awq_marlin_linear, awq_to_marlin_zero_points, check_marlin_supported,
|
||||
check_marlin_supports_layer, check_moe_marlin_supports_layer,
|
||||
marlin_make_empty_g_idx, marlin_make_workspace_new,
|
||||
marlin_moe_permute_scales, marlin_permute_bias, marlin_permute_scales,
|
||||
moe_awq_to_marlin_zero_points, verify_marlin_supported,
|
||||
verify_marlin_supports_shape)
|
||||
apply_awq_marlin_linear,
|
||||
awq_to_marlin_zero_points,
|
||||
check_marlin_supported,
|
||||
check_marlin_supports_layer,
|
||||
check_moe_marlin_supports_layer,
|
||||
marlin_make_empty_g_idx,
|
||||
marlin_make_workspace_new,
|
||||
marlin_moe_permute_scales,
|
||||
marlin_permute_bias,
|
||||
marlin_permute_scales,
|
||||
moe_awq_to_marlin_zero_points,
|
||||
verify_marlin_supported,
|
||||
verify_marlin_supports_shape,
|
||||
)
|
||||
from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
|
||||
from vllm.model_executor.parameter import (GroupQuantScaleParameter,
|
||||
PackedvLLMParameter)
|
||||
from vllm.model_executor.parameter import GroupQuantScaleParameter, PackedvLLMParameter
|
||||
from vllm.platforms import current_platform
|
||||
from vllm.scalar_type import scalar_types
|
||||
|
||||
@@ -48,10 +64,15 @@ class AWQMarlinConfig(QuantizationConfig):
|
||||
8: scalar_types.uint8,
|
||||
}
|
||||
|
||||
def __init__(self, weight_bits: int, group_size: int, zero_point: bool,
|
||||
lm_head_quantized: bool,
|
||||
modules_to_not_convert: Optional[list[str]],
|
||||
full_config: dict[str, Any]) -> None:
|
||||
def __init__(
|
||||
self,
|
||||
weight_bits: int,
|
||||
group_size: int,
|
||||
zero_point: bool,
|
||||
lm_head_quantized: bool,
|
||||
modules_to_not_convert: Optional[list[str]],
|
||||
full_config: dict[str, Any],
|
||||
) -> None:
|
||||
super().__init__()
|
||||
self.pack_factor = 32 // weight_bits # packed into int32
|
||||
self.group_size = group_size
|
||||
@@ -62,21 +83,25 @@ class AWQMarlinConfig(QuantizationConfig):
|
||||
self.full_config = full_config
|
||||
|
||||
if self.weight_bits not in self.TYPE_MAP:
|
||||
raise ValueError(f"Unsupported num_bits = {self.weight_bits}. "
|
||||
f"Supported num_bits = {self.TYPE_MAP.keys()}")
|
||||
raise ValueError(
|
||||
f"Unsupported num_bits = {self.weight_bits}. "
|
||||
f"Supported num_bits = {self.TYPE_MAP.keys()}"
|
||||
)
|
||||
|
||||
self.quant_type = self.TYPE_MAP[self.weight_bits]
|
||||
|
||||
verify_marlin_supported(self.quant_type,
|
||||
group_size=self.group_size,
|
||||
has_zp=self.zero_point)
|
||||
verify_marlin_supported(
|
||||
self.quant_type, group_size=self.group_size, has_zp=self.zero_point
|
||||
)
|
||||
|
||||
def __repr__(self) -> str:
|
||||
return (f"AWQMarlinConfig(quant_type={self.quant_type}, "
|
||||
f"group_size={self.group_size}, "
|
||||
f"zero_point={self.zero_point}, "
|
||||
f"lm_head_quantized={self.lm_head_quantized}, "
|
||||
f"modules_to_not_convert={self.modules_to_not_convert})")
|
||||
return (
|
||||
f"AWQMarlinConfig(quant_type={self.quant_type}, "
|
||||
f"group_size={self.group_size}, "
|
||||
f"zero_point={self.zero_point}, "
|
||||
f"lm_head_quantized={self.lm_head_quantized}, "
|
||||
f"modules_to_not_convert={self.modules_to_not_convert})"
|
||||
)
|
||||
|
||||
@classmethod
|
||||
def get_name(cls) -> QuantizationMethods:
|
||||
@@ -99,37 +124,51 @@ class AWQMarlinConfig(QuantizationConfig):
|
||||
weight_bits = cls.get_from_keys(config, ["bits"])
|
||||
group_size = cls.get_from_keys(config, ["group_size"])
|
||||
zero_point = cls.get_from_keys(config, ["zero_point"])
|
||||
lm_head_quantized = cls.get_from_keys_or(config, ["lm_head"],
|
||||
default=False)
|
||||
lm_head_quantized = cls.get_from_keys_or(config, ["lm_head"], default=False)
|
||||
modules_to_not_convert = cls.get_from_keys_or(
|
||||
config, ["modules_to_not_convert"], None)
|
||||
return cls(weight_bits, group_size, zero_point, lm_head_quantized,
|
||||
modules_to_not_convert, config)
|
||||
config, ["modules_to_not_convert"], None
|
||||
)
|
||||
return cls(
|
||||
weight_bits,
|
||||
group_size,
|
||||
zero_point,
|
||||
lm_head_quantized,
|
||||
modules_to_not_convert,
|
||||
config,
|
||||
)
|
||||
|
||||
@classmethod
|
||||
def override_quantization_method(
|
||||
cls, hf_quant_cfg, user_quant) -> Optional[QuantizationMethods]:
|
||||
cls, hf_quant_cfg, user_quant
|
||||
) -> Optional[QuantizationMethods]:
|
||||
can_convert = cls.is_awq_marlin_compatible(hf_quant_cfg)
|
||||
is_valid_user_quant = (user_quant is None or user_quant == "marlin"
|
||||
or user_quant == "awq_marlin")
|
||||
is_valid_user_quant = (
|
||||
user_quant is None or user_quant == "marlin" or user_quant == "awq_marlin"
|
||||
)
|
||||
|
||||
if can_convert and is_valid_user_quant:
|
||||
msg = ("The model is convertible to {} during runtime."
|
||||
" Using {} kernel.".format(cls.get_name(), cls.get_name()))
|
||||
msg = (
|
||||
"The model is convertible to {} during runtime."
|
||||
" Using {} kernel.".format(cls.get_name(), cls.get_name())
|
||||
)
|
||||
logger.info(msg)
|
||||
return cls.get_name()
|
||||
|
||||
if can_convert and user_quant == "awq":
|
||||
logger.info("Detected that the model can run with awq_marlin"
|
||||
", however you specified quantization=awq explicitly,"
|
||||
" so forcing awq. Use quantization=awq_marlin for"
|
||||
" faster inference")
|
||||
logger.info(
|
||||
"Detected that the model can run with awq_marlin"
|
||||
", however you specified quantization=awq explicitly,"
|
||||
" so forcing awq. Use quantization=awq_marlin for"
|
||||
" faster inference"
|
||||
)
|
||||
return None
|
||||
|
||||
def get_quant_method(self, layer: torch.nn.Module,
|
||||
prefix: str) -> Optional["QuantizeMethodBase"]:
|
||||
if (isinstance(layer, LinearBase) or
|
||||
(isinstance(layer, ParallelLMHead) and self.lm_head_quantized)):
|
||||
def get_quant_method(
|
||||
self, layer: torch.nn.Module, prefix: str
|
||||
) -> Optional["QuantizeMethodBase"]:
|
||||
if isinstance(layer, LinearBase) or (
|
||||
isinstance(layer, ParallelLMHead) and self.lm_head_quantized
|
||||
):
|
||||
if is_layer_skipped_awq(prefix, self.modules_to_not_convert):
|
||||
return UnquantizedLinearMethod()
|
||||
# Check if the layer is supported by AWQMarlin.
|
||||
@@ -138,21 +177,25 @@ class AWQMarlinConfig(QuantizationConfig):
|
||||
"Layer '%s' is not supported by AWQMarlin. Falling back to unoptimized AWQ kernels.", # noqa: E501
|
||||
prefix,
|
||||
)
|
||||
return AWQConfig.from_config(
|
||||
self.full_config).get_quant_method(layer, prefix)
|
||||
return AWQConfig.from_config(self.full_config).get_quant_method(
|
||||
layer, prefix
|
||||
)
|
||||
return AWQMarlinLinearMethod(self)
|
||||
elif isinstance(layer, FusedMoE):
|
||||
from vllm.model_executor.layers.quantization.moe_wna16 import (
|
||||
MoeWNA16Config)
|
||||
from vllm.model_executor.layers.quantization.moe_wna16 import MoeWNA16Config
|
||||
|
||||
if is_layer_skipped_awq(
|
||||
prefix, getattr(self, "modules_to_not_convert", [])):
|
||||
prefix, getattr(self, "modules_to_not_convert", [])
|
||||
):
|
||||
return UnquantizedFusedMoEMethod(layer.moe_config)
|
||||
if not check_moe_marlin_supports_layer(layer, self.group_size):
|
||||
logger.warning_once(
|
||||
f"Layer '{prefix}' is not supported by AWQMoeMarlin. "
|
||||
"Falling back to Moe WNA16 kernels.")
|
||||
return MoeWNA16Config.from_config(
|
||||
self.full_config).get_quant_method(layer, prefix)
|
||||
"Falling back to Moe WNA16 kernels."
|
||||
)
|
||||
return MoeWNA16Config.from_config(self.full_config).get_quant_method(
|
||||
layer, prefix
|
||||
)
|
||||
return AWQMoEMethod(self, layer.moe_config)
|
||||
return None
|
||||
|
||||
@@ -171,15 +214,15 @@ class AWQMarlinConfig(QuantizationConfig):
|
||||
return False
|
||||
|
||||
# If we cannot find the info needed in the config, cannot convert.
|
||||
if (num_bits is None or group_size is None or zero_point is None):
|
||||
if num_bits is None or group_size is None or zero_point is None:
|
||||
return False
|
||||
|
||||
if num_bits not in cls.TYPE_MAP:
|
||||
return False
|
||||
|
||||
return check_marlin_supported(quant_type=cls.TYPE_MAP[num_bits],
|
||||
group_size=group_size,
|
||||
has_zp=zero_point)
|
||||
return check_marlin_supported(
|
||||
quant_type=cls.TYPE_MAP[num_bits], group_size=group_size, has_zp=zero_point
|
||||
)
|
||||
|
||||
|
||||
class AWQMarlinLinearMethod(LinearMethodBase):
|
||||
@@ -216,7 +259,8 @@ class AWQMarlinLinearMethod(LinearMethodBase):
|
||||
output_size_per_partition=output_size_per_partition,
|
||||
input_size_per_partition=input_size_per_partition,
|
||||
input_size=input_size,
|
||||
group_size=group_size)
|
||||
group_size=group_size,
|
||||
)
|
||||
|
||||
qweight = PackedvLLMParameter(
|
||||
data=torch.empty(
|
||||
@@ -228,7 +272,8 @@ class AWQMarlinLinearMethod(LinearMethodBase):
|
||||
output_dim=1,
|
||||
packed_dim=1,
|
||||
packed_factor=self.quant_config.pack_factor,
|
||||
weight_loader=weight_loader)
|
||||
weight_loader=weight_loader,
|
||||
)
|
||||
|
||||
num_groups = input_size_per_partition // group_size
|
||||
|
||||
@@ -242,16 +287,19 @@ class AWQMarlinLinearMethod(LinearMethodBase):
|
||||
output_dim=1,
|
||||
packed_dim=1,
|
||||
packed_factor=self.quant_config.pack_factor,
|
||||
weight_loader=weight_loader)
|
||||
weight_loader=weight_loader,
|
||||
)
|
||||
|
||||
scales = GroupQuantScaleParameter(data=torch.empty(
|
||||
num_groups,
|
||||
output_size_per_partition,
|
||||
dtype=params_dtype,
|
||||
),
|
||||
input_dim=0,
|
||||
output_dim=1,
|
||||
weight_loader=weight_loader)
|
||||
scales = GroupQuantScaleParameter(
|
||||
data=torch.empty(
|
||||
num_groups,
|
||||
output_size_per_partition,
|
||||
dtype=params_dtype,
|
||||
),
|
||||
input_dim=0,
|
||||
output_dim=1,
|
||||
weight_loader=weight_loader,
|
||||
)
|
||||
|
||||
layer.register_parameter("qweight", qweight)
|
||||
layer.register_parameter("qzeros", qzeros)
|
||||
@@ -267,12 +315,9 @@ class AWQMarlinLinearMethod(LinearMethodBase):
|
||||
# Here, we handle the repacking
|
||||
def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
|
||||
device = layer.qweight.device
|
||||
layer.qweight = torch.nn.Parameter(layer.qweight.data,
|
||||
requires_grad=False)
|
||||
layer.qzeros = torch.nn.Parameter(layer.qzeros.data,
|
||||
requires_grad=False)
|
||||
layer.scales = torch.nn.Parameter(layer.scales.data,
|
||||
requires_grad=False)
|
||||
layer.qweight = torch.nn.Parameter(layer.qweight.data, requires_grad=False)
|
||||
layer.qzeros = torch.nn.Parameter(layer.qzeros.data, requires_grad=False)
|
||||
layer.scales = torch.nn.Parameter(layer.scales.data, requires_grad=False)
|
||||
|
||||
# Allocate marlin workspace
|
||||
layer.workspace = marlin_make_workspace_new(device)
|
||||
@@ -282,7 +327,8 @@ class AWQMarlinLinearMethod(LinearMethodBase):
|
||||
layer.qweight,
|
||||
size_k=layer.input_size_per_partition,
|
||||
size_n=layer.output_size_per_partition,
|
||||
num_bits=self.quant_config.quant_type.size_bits)
|
||||
num_bits=self.quant_config.quant_type.size_bits,
|
||||
)
|
||||
replace_parameter(layer, "qweight", marlin_qweight)
|
||||
|
||||
# Permute scales from AWQ format to marlin format.
|
||||
@@ -290,7 +336,8 @@ class AWQMarlinLinearMethod(LinearMethodBase):
|
||||
layer.scales,
|
||||
size_k=layer.input_size_per_partition,
|
||||
size_n=layer.output_size_per_partition,
|
||||
group_size=self.quant_config.group_size)
|
||||
group_size=self.quant_config.group_size,
|
||||
)
|
||||
replace_parameter(layer, "scales", marlin_scales)
|
||||
|
||||
# Permute zero-points from AWQ format to marlin format.
|
||||
@@ -298,7 +345,8 @@ class AWQMarlinLinearMethod(LinearMethodBase):
|
||||
layer.qzeros,
|
||||
size_k=layer.num_groups,
|
||||
size_n=layer.output_size_per_partition,
|
||||
num_bits=self.quant_config.quant_type.size_bits)
|
||||
num_bits=self.quant_config.quant_type.size_bits,
|
||||
)
|
||||
replace_parameter(layer, "qzeros", marlin_zp)
|
||||
|
||||
# Not-used
|
||||
@@ -325,11 +373,11 @@ class AWQMarlinLinearMethod(LinearMethodBase):
|
||||
quant_type=self.quant_config.quant_type,
|
||||
output_size_per_partition=layer.output_size_per_partition,
|
||||
input_size_per_partition=layer.input_size_per_partition,
|
||||
bias=bias)
|
||||
bias=bias,
|
||||
)
|
||||
|
||||
|
||||
class AWQMoEMethod(FusedMoEMethodBase):
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
quant_config: AWQMarlinConfig,
|
||||
@@ -341,75 +389,93 @@ class AWQMoEMethod(FusedMoEMethodBase):
|
||||
raise ValueError("AWQMoEMethod only supports 4bit now.")
|
||||
self.quant_type = scalar_types.uint4
|
||||
|
||||
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):
|
||||
extra_weight_attrs.update({
|
||||
"is_transposed":
|
||||
True,
|
||||
"quant_method":
|
||||
FusedMoeWeightScaleSupported.GROUP.value,
|
||||
})
|
||||
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,
|
||||
):
|
||||
extra_weight_attrs.update(
|
||||
{
|
||||
"is_transposed": True,
|
||||
"quant_method": FusedMoeWeightScaleSupported.GROUP.value,
|
||||
}
|
||||
)
|
||||
|
||||
w13_qweight = Parameter(
|
||||
torch.empty(num_experts,
|
||||
hidden_size,
|
||||
2 * intermediate_size_per_partition //
|
||||
self.quant_config.pack_factor,
|
||||
dtype=torch.int32),
|
||||
requires_grad=False)
|
||||
torch.empty(
|
||||
num_experts,
|
||||
hidden_size,
|
||||
2 * intermediate_size_per_partition // self.quant_config.pack_factor,
|
||||
dtype=torch.int32,
|
||||
),
|
||||
requires_grad=False,
|
||||
)
|
||||
layer.register_parameter("w13_qweight", w13_qweight)
|
||||
set_weight_attrs(w13_qweight, extra_weight_attrs)
|
||||
|
||||
w2_qweight = Parameter(torch.empty(num_experts,
|
||||
intermediate_size_per_partition,
|
||||
hidden_size //
|
||||
self.quant_config.pack_factor,
|
||||
dtype=torch.int32),
|
||||
requires_grad=False)
|
||||
w2_qweight = Parameter(
|
||||
torch.empty(
|
||||
num_experts,
|
||||
intermediate_size_per_partition,
|
||||
hidden_size // self.quant_config.pack_factor,
|
||||
dtype=torch.int32,
|
||||
),
|
||||
requires_grad=False,
|
||||
)
|
||||
layer.register_parameter("w2_qweight", w2_qweight)
|
||||
set_weight_attrs(w2_qweight, extra_weight_attrs)
|
||||
|
||||
num_groups_w13 = hidden_size // self.quant_config.group_size
|
||||
num_groups_w2 = (intermediate_size_per_partition //
|
||||
self.quant_config.group_size)
|
||||
num_groups_w2 = intermediate_size_per_partition // self.quant_config.group_size
|
||||
|
||||
# WEIGHT_SCALES
|
||||
# Allocate 2 scales for w1 and w3 respectively.
|
||||
w13_scales = Parameter(torch.empty(num_experts,
|
||||
num_groups_w13,
|
||||
intermediate_size_per_partition * 2,
|
||||
dtype=params_dtype),
|
||||
requires_grad=False)
|
||||
w13_scales = Parameter(
|
||||
torch.empty(
|
||||
num_experts,
|
||||
num_groups_w13,
|
||||
intermediate_size_per_partition * 2,
|
||||
dtype=params_dtype,
|
||||
),
|
||||
requires_grad=False,
|
||||
)
|
||||
layer.register_parameter("w13_scales", w13_scales)
|
||||
set_weight_attrs(w13_scales, extra_weight_attrs)
|
||||
|
||||
w2_scales = Parameter(torch.empty(num_experts,
|
||||
num_groups_w2,
|
||||
hidden_size,
|
||||
dtype=params_dtype),
|
||||
requires_grad=False)
|
||||
w2_scales = Parameter(
|
||||
torch.empty(num_experts, num_groups_w2, hidden_size, dtype=params_dtype),
|
||||
requires_grad=False,
|
||||
)
|
||||
layer.register_parameter("w2_scales", w2_scales)
|
||||
set_weight_attrs(w2_scales, extra_weight_attrs)
|
||||
|
||||
# WEIGHT_ZERO_POINT
|
||||
# Allocate 2 zero points for w1 and w3 respectively.
|
||||
w13_qzeros = Parameter(
|
||||
torch.empty(num_experts,
|
||||
num_groups_w13,
|
||||
2 * intermediate_size_per_partition //
|
||||
self.quant_config.pack_factor,
|
||||
dtype=torch.int32),
|
||||
requires_grad=False)
|
||||
torch.empty(
|
||||
num_experts,
|
||||
num_groups_w13,
|
||||
2 * intermediate_size_per_partition // self.quant_config.pack_factor,
|
||||
dtype=torch.int32,
|
||||
),
|
||||
requires_grad=False,
|
||||
)
|
||||
layer.register_parameter("w13_qzeros", w13_qzeros)
|
||||
set_weight_attrs(w13_qzeros, extra_weight_attrs)
|
||||
|
||||
w2_qzeros = Parameter(torch.empty(num_experts,
|
||||
num_groups_w2,
|
||||
hidden_size //
|
||||
self.quant_config.pack_factor,
|
||||
dtype=torch.int32),
|
||||
requires_grad=False)
|
||||
w2_qzeros = Parameter(
|
||||
torch.empty(
|
||||
num_experts,
|
||||
num_groups_w2,
|
||||
hidden_size // self.quant_config.pack_factor,
|
||||
dtype=torch.int32,
|
||||
),
|
||||
requires_grad=False,
|
||||
)
|
||||
layer.register_parameter("w2_qzeros", w2_qzeros)
|
||||
set_weight_attrs(w2_qzeros, extra_weight_attrs)
|
||||
|
||||
@@ -469,14 +535,16 @@ class AWQMoEMethod(FusedMoEMethodBase):
|
||||
layer.w13_qzeros,
|
||||
size_k=layer.w13_qzeros.shape[1],
|
||||
size_n=layer.w13_qzeros.shape[2] * self.quant_config.pack_factor,
|
||||
num_bits=self.quant_config.weight_bits)
|
||||
num_bits=self.quant_config.weight_bits,
|
||||
)
|
||||
replace_parameter(layer, "w13_qzeros", marlin_w13_zp)
|
||||
|
||||
marlin_w2_zp = moe_awq_to_marlin_zero_points(
|
||||
layer.w2_qzeros,
|
||||
size_k=layer.w2_qzeros.shape[1],
|
||||
size_n=layer.w2_qzeros.shape[2] * self.quant_config.pack_factor,
|
||||
num_bits=self.quant_config.weight_bits)
|
||||
num_bits=self.quant_config.weight_bits,
|
||||
)
|
||||
replace_parameter(layer, "w2_qzeros", marlin_w2_zp)
|
||||
|
||||
if hasattr(layer, "w13_bias") and layer.w13_bias is not None:
|
||||
@@ -486,7 +554,8 @@ class AWQMoEMethod(FusedMoEMethodBase):
|
||||
layer.w2_bias.data = marlin_permute_bias(layer.w2_bias)
|
||||
|
||||
def get_fused_moe_quant_config(
|
||||
self, layer: torch.nn.Module) -> Optional[FusedMoEQuantConfig]:
|
||||
self, layer: torch.nn.Module
|
||||
) -> Optional[FusedMoEQuantConfig]:
|
||||
return None
|
||||
|
||||
def apply(
|
||||
@@ -515,8 +584,7 @@ class AWQMoEMethod(FusedMoEMethodBase):
|
||||
assert self.fused_experts is None
|
||||
|
||||
if enable_eplb:
|
||||
raise NotImplementedError(
|
||||
"EPLB not supported for `AWQMoEMethod` yet.")
|
||||
raise NotImplementedError("EPLB not supported for `AWQMoEMethod` yet.")
|
||||
|
||||
assert activation == "silu", "Only SiLU activation is supported."
|
||||
|
||||
@@ -532,7 +600,8 @@ class AWQMoEMethod(FusedMoEMethodBase):
|
||||
scoring_func=scoring_func,
|
||||
routed_scaling_factor=routed_scaling_factor,
|
||||
e_score_correction_bias=e_score_correction_bias,
|
||||
indices_type=self.topk_indices_dtype)
|
||||
indices_type=self.topk_indices_dtype,
|
||||
)
|
||||
|
||||
return torch.ops.vllm.fused_marlin_moe(
|
||||
x,
|
||||
@@ -551,4 +620,5 @@ class AWQMoEMethod(FusedMoEMethodBase):
|
||||
expert_map=expert_map,
|
||||
w1_zeros=layer.w13_qzeros,
|
||||
w2_zeros=layer.w2_qzeros,
|
||||
workspace=layer.workspace)
|
||||
workspace=layer.workspace,
|
||||
)
|
||||
|
||||
Reference in New Issue
Block a user