xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 01:05:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bug] Fix Unexpected Keyword Argument 'w1_bias' (#22757)

Browse Source 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
Wentao Ye 2025-08-13 00:31:47 -04:00 committed by GitHub
parent 4082338a25
commit 77a6bf07ae
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 13 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
vllm/model_executor/layers/fused_moe/layer.py
View File

@ -475,12 +475,11 @@ class UnquantizedFusedMoEMethod(FusedMoEMethodBase, CustomOp):
activation=activation,
apply_router_weight_on_input=apply_router_weight_on_input)
else:
return self.fused_experts(
# add w1_bias/w2_bias to kwargs if they exist
kwargs = dict(
hidden_states=x,
w1=layer.w13_weight,
w2=layer.w2_weight,
w1_bias=layer.w13_bias if self.has_bias else None,
w2_bias=layer.w2_bias if self.has_bias else None,
topk_weights=topk_weights,
topk_ids=topk_ids,
inplace=True,
@ -489,6 +488,17 @@ class UnquantizedFusedMoEMethod(FusedMoEMethodBase, CustomOp):
global_num_experts=global_num_experts,
expert_map=expert_map,
)
if isinstance(self.fused_experts,
FusedMoEModularKernel) and self.has_bias:
raise ValueError(
"FusedMoEModularKernel does not support bias.")
if self.has_bias:
kwargs.update({
"w1_bias": getattr(layer, "w13_bias", None),
"w2_bias": getattr(layer, "w2_bias", None),
})
return self.fused_experts(**kwargs)
def forward_cpu(
self,