Skip to content

Fix misleading preemption logs in checkpointing #4124

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
lukebaumann wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix misleading preemption logs in checkpointing #4124

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
10 changes: 3 additions & 7 deletions src/maxtext/common/checkpointing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -973,15 +973,11 @@ def maybe_save_checkpoint(checkpoint_manager, state, config, data_iterator, step
except elastic_utils.manager.ScaleUpSignalError as e:
if config.elastic_enabled:
max_logging.log(f"Elastic event detected, letting exception bubble up: {e}")
raise
else:
raise exceptions.StopTraining("Job is preempted.") from e
raise
except jax.errors.JaxRuntimeError as e:
if config.elastic_enabled:
max_logging.log(f"Elastic event detected, letting exception bubble up: {e}")
raise
else:
raise exceptions.StopTraining("Job is preempted.") from e
raise
except Exception as e:
raise exceptions.StopTraining(f"Checkpointing failed. {str(e)}") from e

Expand All @@ -991,7 +987,7 @@ def maybe_save_checkpoint(checkpoint_manager, state, config, data_iterator, step

# Raise exception upon preemption
if checkpoint_manager.reached_preemption(actual_step):
raise exceptions.StopTraining("Job is preempted.")
raise exceptions.StopTraining("Job received termination signal (SIGTERM).")


def save_checkpoint(checkpoint_manager, step, state, config=None, data_iterator=None, force=False):
Expand Down
Loading