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 key error in GRPOTrainer #1818

Merged
merged 2 commits into from
Feb 25, 2025
Merged

Fix key error in GRPOTrainer #1818

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4ffddad
fix keyerror in GRPOTrainer
le-big-mac Feb 24, 2025
efeb7ca
check for train in _metrics
le-big-mac Feb 25, 2025
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
14 changes: 10 additions & 4 deletions unsloth/models/rl_replacements.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -164,7 +164,7 @@ def grpo_trainer__prepare_inputs(function_name, function):
# Remove _move_model_to_vllm
def grpo_trainer__move_model_to_vllm(function_name, function):
if function_name != "_move_model_to_vllm": return function

def _move_model_to_vllm(self, *args, **kwargs): return None

function = inspect.getsource(_move_model_to_vllm)
Expand Down Expand Up @@ -246,14 +246,20 @@ def compute_loss(self, model, inputs, return_outputs = False, num_items_in_batch
self, _input_ids, logits_to_keep, completion_mask, advantages,
n_chunks = self.args.unsloth_num_chunks,
)

# Log the metrics
# completion_length = self.accelerator.gather_for_metrics(completion_mask.sum(1)).float().mean().item()
self._metrics["completion_length"].append(completion_length.item())

# mean_kl = ((per_token_kl * completion_mask).sum(dim=1) / completion_mask.sum(dim=1)).mean()
# self._metrics["kl"].append(self.accelerator.gather_for_metrics(mean_kl).mean().item())
self._metrics["kl"].append(mean_kl.item())

if "train" in self._metrics:
mode = "eval" if self.control.should_evaluate else "train"
self._metrics[mode]["completion_length"].append(completion_length.item())
self._metrics[mode]["kl"].append(mean_kl.item())
else:
self._metrics["completion_length"].append(completion_length.item())
self._metrics["kl"].append(mean_kl.item())
return loss
pass

Expand Down