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

Fix qwen 14b run into register attention fwd (#11128)

Browse source 
* fix qwen 14b
This commit is contained in:
Zhao Changmin 2024-05-24 14:45:07 +08:00 committed by GitHub
parent 373f9e6c79
commit 65f4212f89
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 2 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
python/llm/src/ipex_llm/transformers/convert.py
View file

@ -1249,7 +1249,8 @@ def _optimize_post(model, lightweight_bmm=False):
from ipex_llm.transformers.models.qwen import qwen_mlp_forward from ipex_llm.transformers.models.qwen import qwen_mlp_forward
from ipex_llm.transformers.models.chatglm2 import chatglm_rms_norm_forward from ipex_llm.transformers.models.chatglm2 import chatglm_rms_norm_forward
from ipex_llm.transformers.models.qwen import qwen_model_forward from ipex_llm.transformers.models.qwen import qwen_model_forward
if model.config.max_position_embeddings == 8192: if model.config.max_position_embeddings == 8192 \
and model.config.hidden_size == 4096:
convert_forward(model, convert_forward(model,
module.QWenAttention, module.QWenAttention,
qwen_attention_forward_registered qwen_attention_forward_registered