EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

fix awq_dequantize import (#5669)

Browse Source
This commit is contained in:
Yineng Zhang
2025-04-23 01:36:26 -07:00
committed by GitHub
parent 15fabcc07f
commit fbb5f229d4
1 changed files with 5 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
python/sglang/srt/layers/quantization/awq.py
View File

@@ -3,7 +3,6 @@ import logging
from typing import Any, Dict, List, Optional from typing import Any, Dict, List, Optional
import torch import torch
from sgl_kernel import awq_dequantize
from sglang.srt.layers.linear import ( from sglang.srt.layers.linear import (
LinearBase, LinearBase,
@@ -12,6 +11,11 @@ from sglang.srt.layers.linear import (
) )
from sglang.srt.layers.parameter import GroupQuantScaleParameter, PackedvLLMParameter from sglang.srt.layers.parameter import GroupQuantScaleParameter, PackedvLLMParameter
from sglang.srt.layers.quantization.base_config import QuantizationConfig from sglang.srt.layers.quantization.base_config import QuantizationConfig
from sglang.srt.utils import is_cuda
_is_cuda = is_cuda()
if _is_cuda:
from sgl_kernel import awq_dequantize
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)