[INC] fix w4a4 model (#50807)

Signed-off-by: mayuyuace <qiming1.zhang@intel.com>
This commit is contained in:
Qiming Zhang
2026-08-03 14:28:58 +08:00
committed by GitHub
parent f5bb701fa2
commit 2755489a26
2 changed files with 4 additions and 3 deletions
@@ -7,6 +7,7 @@ import torch
from torch.nn.parameter import Parameter
from vllm.model_executor.kernels.linear import init_mxfp4_linear_kernel
from vllm.model_executor.layers.quantization.utils.quant_utils import kMxfp4Dynamic
from vllm.model_executor.parameter import (
GroupQuantScaleParameter,
ModelWeightParameter,
@@ -29,7 +30,7 @@ class INCMxfp4LinearMethod(INCLinearScheme):
def __init__(self, layer_config: "INCLayerConfig") -> None:
self.group_size = layer_config.group_size or 32
self.kernel = init_mxfp4_linear_kernel()
self.kernel = init_mxfp4_linear_kernel(activation_quant_key=kMxfp4Dynamic)
@classmethod
def get_min_capability(cls) -> int: