Skip to content

Add epoch summary logging and update documentation#337

Merged
sgreenbury merged 1 commit intomainfrom
timing-log-summary-docs
Apr 20, 2026
Merged

Add epoch summary logging and update documentation#337
sgreenbury merged 1 commit intomainfrom
timing-log-summary-docs

Conversation

@sgreenbury
Copy link
Copy Markdown
Contributor

This pull request improves the visibility of training timing information by adding a human-readable summary of epoch timings to the logs at the end of training. This allows users to quickly inspect timing quality without needing to load the checkpoint file.

Logging and user experience improvements:

  • Added a summary log at the end of training that reports the number of epochs, mean, minimum, and maximum epoch durations, making it easier to assess training timing directly from logs. For short runs (≤12 epochs), the per-epoch times are also listed. (src/autocast/scripts/training.py)
  • Updated documentation to describe the new timing summary log output, clarifying that users can inspect timing quality without loading the checkpoint. (docs/SCRIPTS_AND_CONFIGS.md)

@sgreenbury sgreenbury merged commit 63c671c into main Apr 20, 2026
3 checks passed
@sgreenbury sgreenbury deleted the timing-log-summary-docs branch April 20, 2026 16:56
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant