Fix SFT evaluation metric logging by adding avg_perplexity

[igorts-git]

Description

Fix SFT (and DPO) eval that started failing recently due to changes in PR #3664.
Now we must report eval/avg_perplexity to the metric_logger or it would fail with and Exception.

Tests

Manual execution of a local SFT run with eval enabled for a few steps with and without the fix.

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run end-to-end tests tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!