Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix glm4-9b-chat nan error on vllm 0.3.3 #11970

Merged
merged 2 commits into from
Aug 30, 2024
Merged

Fix glm4-9b-chat nan error on vllm 0.3.3 #11970

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3acce70
fix nan value
hzjane Aug 29, 2024
23709c3
update
hzjane Aug 29, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 9 additions & 1 deletion python/llm/src/ipex_llm/vllm/xpu/model_convert.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -248,7 +248,15 @@ def _ipex_llm_load_model(self) -> None:
parallel_config=self.parallel_config,
scheduler_config=self.scheduler_config)
from ipex_llm import optimize_model
optimize_model(self.model, low_bit=low_bit, torch_dtype=self.model_config.dtype)
import os
not_convert_last_mlp = os.getenv("IPEX_LLM_NOT_CONVERT_LAST_MLP", None)
if not_convert_last_mlp is not None:
# only use to avoid nan value in last mlp forward running glm4-9b-chat
modules = ["35.mlp", "36.mlp", "37.mlp", "38.mlp", "39.mlp"]
else:
modules = None
optimize_model(self.model, low_bit=low_bit, torch_dtype=self.model_config.dtype,
modules_to_not_convert=modules)
self.model = self.model.to(device=self.device_config.device,
dtype=self.model_config.dtype)

Expand Down
Loading