apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-20 07:24:42 +08:00
Code Issues Actions 4 Packages Projects Releases Wiki Activity

Fix down projection weight shape in fused MOE layer (#4041)

Browse Source
This commit is contained in:
Yuanle Liu
2025-09-10 12:49:03 +08:00
committed by GitHub
parent dfc94371ee
commit 7272afe3dc
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
fastdeploy/model_executor/layers/moe/fused_moe_triton_backend.py
View File

@@ -695,7 +695,7 @@ class BlockWiseFP8MoEMethod(QuantMethodBase):
layer, layer,
down_proj_weight_name, down_proj_weight_name,
layer.create_parameter( layer.create_parameter(
shape=self.up_gate_proj_weight_shape, shape=self.down_proj_weight_shape,
dtype=self.weight_dtype, dtype=self.weight_dtype,
default_initializer=paddle.nn.initializer.Constant(0), default_initializer=paddle.nn.initializer.Constant(0),
), ),