apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

[BugFix] Fix return value inconsistency for ep_moe_expert_combine op (#5812)

Browse Source
This commit is contained in:
Ryan
2025-12-29 16:44:00 +08:00
committed by GitHub
parent ffb3ccff74
commit eb782a0225
2 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/model_executor/layers/moe/fused_moe_cutlass_backend.py
+1 -1
View File
@@ -210,7 +210,7 @@ class CutlassMoEMethod(UnquantizedFusedMoEMethod):
None, # down_proj_bias,
False, # norm_topk_prob
1.0,
)[0]
)
else:
tmp_ffn_out = recv_x
fastdeploy/model_executor/layers/moe/fused_moe_deepgemm_backend.py
+2 -2
View File
@@ -250,7 +250,7 @@ class DeepGemmFusedMoeMethod(MoEMethodBase):
None, # down_proj_bias
False, # norm_topk_prob
1.0,
)[0]
)
else:
tmp_ffn_out = paddle.cast(recv_x[0], paddle.bfloat16)
@@ -448,6 +448,6 @@ class DeepGemmFusedMoeMethod(MoEMethodBase):
None,
False, # norm_topk_prob
1.0,
)[0]
)
return tmp_ffn_out