ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update rl.py

Browse source
This commit is contained in:
Daniel Han 2025-02-05 05:23:19 -08:00
commit f2279837e4
1 changed files with 5 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

7
unsloth/models/rl.py
View file

@ -24,9 +24,12 @@ def PatchRL(FastLanguageModel):
@contextmanager
def unsloth_unwrap_model_for_generation(model, *args, **kwargs):
# Must use for_inference to allow inference in Unsloth
FastLanguageModel.for_inference(model)
with unwrap_model_for_generation(model, *args, **kwargs) as unwrapped_model:
yield unwrapped_model
with torch.inference_mode():
with unwrap_model_for_generation(model, *args, **kwargs) as unwrapped_model:
yield unwrapped_model
# Return back to training mode
FastLanguageModel.for_training (model)
pass