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

[Bugfix] Fix GPT-OSS on AMD after #28603 (#28816)

Browse Source 
Signed-off-by: zhewenli <zhewenli@meta.com>
This commit is contained in:
Zhewen Li
2025-11-17 10:15:26 -08:00
committed by GitHub
parent e42bd8c2e3
commit f8b19c0ffd
2 changed files with 7 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/model_executor/layers/quantization/mxfp4.py
View File

@@ -755,8 +755,8 @@ class Mxfp4MoEMethod(FusedMoEMethodBase):
self.w13_weight = w13_weight
self.w2_weight = w2_weight
layer.w13_weight = Parameter(w13_weight.data, requires_grad=False)
layer.w2_weight = Parameter(w2_weight.data, requires_grad=False)
layer.w13_weight = Parameter(w13_weight.storage.data, requires_grad=False)
layer.w2_weight = Parameter(w2_weight.storage.data, requires_grad=False)
else:
raise ValueError(f"Unsupported backend: {self.mxfp4_backend}")