Skip to content

Fixed wrong logs prefixes in KTOTrainer #1641

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

Merged
merged 2 commits into from
May 14, 2024
Merged

Fixed wrong logs prefixes in KTOTrainer #1641

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5a305f9
Fixed wrong logs prefixes in KTOTrainer
bartoszzuk May 13, 2024
d75783f
Pre-commit formating
bartoszzuk May 13, 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
14 changes: 7 additions & 7 deletions trl/trainer/kto_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1525,26 +1525,26 @@ def log(self, logs: Dict[str, float]) -> None:
"""
# logs either has 'loss' or 'eval_loss'
train_eval = "train" if "loss" in logs else "eval"
# train metrics should have no prefix, eval should have 'eval_'
prefix = "eval_" if train_eval == "eval" else ""
# accumulate average metrics from sums and lengths
for split in ["chosen", "rejected"]:
if f"count/{split}" in self._stored_metrics[train_eval]:
count_sum = torch.Tensor(self._stored_metrics[train_eval][f"count/{split}"]).sum().item()
logs[f"{train_eval}/rewards/{split}"] = (
logs[f"{prefix}rewards/{split}"] = (
torch.Tensor(self._stored_metrics[train_eval][f"rewards/{split}_sum"]).sum().item() / count_sum
)
logs[f"{train_eval}/logps/{split}"] = (
logs[f"{prefix}logps/{split}"] = (
torch.Tensor(self._stored_metrics[train_eval][f"logps/{split}_sum"]).sum().item() / count_sum
)
for key in [f"count/{split}", f"rewards/{split}_sum", f"logps/{split}_sum"]:
del self._stored_metrics[train_eval][key]
# calculate reward margin
if f"{train_eval}/rewards/chosen" in logs and f"{train_eval}/rewards/rejected" in logs:
logs[f"{train_eval}/rewards/margins"] = (
logs[f"{train_eval}/rewards/chosen"] - logs[f"{train_eval}/rewards/rejected"]
)
if f"{prefix}rewards/chosen" in logs and f"{prefix}rewards/rejected" in logs:
logs[f"{prefix}rewards/margins"] = logs[f"{prefix}rewards/chosen"] - logs[f"{prefix}rewards/rejected"]
# Add averaged stored metrics to logs
for key, metrics in self._stored_metrics[train_eval].items():
logs[f"{train_eval}/{key}"] = torch.Tensor(metrics).mean().item()
logs[f"{prefix}{key}"] = torch.Tensor(metrics).mean().item()
del self._stored_metrics[train_eval]
return super().log(logs)

Expand Down