ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

hotfix qwen2-7b weight setting (#11991)

Browse source
This commit is contained in:
binbin Deng 2024-09-02 18:11:08 +08:00 committed by GitHub
parent a40ea7038d
commit 2f3d1bd0ec
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 1 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1
python/llm/src/ipex_llm/transformers/npu_models/qwen2_mp.py
View file

@ -587,6 +587,7 @@ def run_decode(
(attn_layer.v_proj.weight, attn_layer.v_proj.scale), (attn_layer.v_proj.weight, attn_layer.v_proj.scale),
(attn_layer.o_proj.weight, attn_layer.o_proj.scale), (attn_layer.o_proj.weight, attn_layer.o_proj.scale),
(mlp_layer.gate_proj.weight, mlp_layer.gate_proj.scale), (mlp_layer.gate_proj.weight, mlp_layer.gate_proj.scale),
(mlp_layer.up_proj.weight, mlp_layer.up_proj.scale),
(mlp_layer.down_proj_0.weight, mlp_layer.down_proj_0.scale), (mlp_layer.down_proj_0.weight, mlp_layer.down_proj_0.scale),
(mlp_layer.down_proj_1.weight, mlp_layer.down_proj_1.scale) (mlp_layer.down_proj_1.weight, mlp_layer.down_proj_1.scale)
] ]