[rhewett-nv]

• I, the PR author, have personally reviewed every line of this PR.

What does this PR do?

Completes the trainer telemetry lifecycle from Python startup through linked training traces and ordered exit cleanup.

This is PR 4 of 7 and is intended to be reviewed relative to PR 3.

Summary

• Captures startup timing for the GPT and MIMO Python entrypoints and records startup work beneath the trainer lifecycle.
• Links training-loop traces and final checkpoint work without changing application control flow.
• Routes normal and signal-driven cleanup through one lifecycle owner so open spans finish before telemetry providers shut down.

Issue tracking

Linked issue: N/A — follow-up integration work.

Contribution process

Pre-checks

• I have added relevant unit tests
• I have added relevant functional tests
• This PR does not add or change a GPU kernel, so no determinism manifest update is required.
• I have added proper typing to my code (typing guidelines)
• I have added relevant documentation
• I have run the autoformatter checks on my PR

Validation

• Focused telemetry unit suite at this boundary: 201 passed.
• Scoped Ruff, Black, adjacent-diff, DCO, and signature checks passed.
• No local GPU or distributed run was performed.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.