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 06:56:12 -08:00
commit 8506d6be95
1 changed files with 9 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

16
unsloth/models/rl.py
View file

@ -39,13 +39,15 @@ def PatchRL(FastLanguageModel):
@contextmanager
def unsloth_unwrap_model_for_generation(model, accelerator):
# Must use for_inference to allow inference in Unsloth
with unwrap_model_for_generation(model, accelerator) as unwrapped_model:
FastLanguageModel.for_inference(unwrapped_model)
try:
yield unwrapped_model
finally:
# Finally return back training
FastLanguageModel.for_training(model)
with torch.inference_mode():
with unwrap_model_for_generation(model, accelerator) as unwrapped_model:
FastLanguageModel.for_inference(unwrapped_model)
try:
yield unwrapped_model
finally:
# Finally return back training
FastLanguageModel.for_training(model)
pass
pass
pass
pass