[Lai-Kai]

The streaming trainer already writes end-of-pass evaluation metrics to .metrics.jsonl, but the trajectory builder currently only reads text logs. Add JSONL input support so users can export and plot those evaluation metrics with the existing trajectory outputs.

JSONL records use global_step for the evaluation axis and filter metrics by the selected task. The last matching record wins for repeated steps, including resumed runs and final evaluation. Invalid JSONL or a missing task returns a clear error. Existing text-log input remains supported, and no new dependencies are required.

Validation:

• python3 -m unittest discover -s recommendation/scripts/tests -p 'test_build_ne_auc_trajectory.py' -v — 4 tests passed.
• CLI checks with small synthetic JSONL and text logs verified JSON/CSV contents without matplotlib.
• Malformed JSONL returned exit code 2 with a line number and created no output files.

[github-actions]

MLCommons CLA bot:
Thank you very much for your submission; we really appreciate it. Before we can accept your contribution,
we ask that you sign the MLCommons CLA (Apache 2). Please submit your GitHub ID to our onboarding form to initiate
authorization. If you are from a MLCommons member organization, we will request that you be added to the CLA.
If you are not from a member organization, we will email you a CLA to sign. For any questions, please contact
support@mlcommons.org.
0 out of 1 committers have signed the MLCommons CLA.
❌ @Lai-Kai
_{You can retrigger this bot by commenting recheck in this Pull Request}

[Lai-Kai]

recheck