Skip to content

Fix progress bar logging #561

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
+6 −2
Open

Fix progress bar logging #561

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
8 changes: 6 additions & 2 deletions tunix/sft/peft_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -492,7 +492,8 @@ def _buffer_metrics(
step_time_delta: float = 0.0,
) -> MetricsBuffer:
"""Buffers metrics for the current step."""
loss = np.array(loss)
# Cast to fp32 for logging, because NumPy doesn't support bf16.
loss = np.array(loss).astype(np.float32)
if metrics_buffer is None:
metrics_buffer = MetricsBuffer(
step=step, losses=[loss], step_time_deltas=[step_time_delta]
Expand Down Expand Up @@ -811,4 +812,7 @@ def _default_loss_fn(

# Return the negative log likelihood (NLL) loss.
# Equivalent to: optax.softmax_cross_entropy(logits, one_hot).mean()
return -jnp.sum(jax.nn.log_softmax(logits) * one_hot) * norm_factor
loss = -jnp.sum(jax.nn.log_softmax(logits) * one_hot) * norm_factor
# Cast loss back to logits dtype.
loss = loss.astype(logits.dtype)
return loss