fix: allow wandb config value changes on resume

[gkaplun-nvidia]

Summary

Add allow_val_change=True to wandb.config.update() calls in both WandbLogger and MLflowWandbLogger.

Problem

When resuming a wandb run, config.update() raises an error if a hyperparameter key already exists with a different value. This is common when hyperparameters are re-logged on resume (e.g., if the config is passed again during setup()).

Fix

Pass allow_val_change=True so that config values can be overwritten without raising. This matches the expected behavior for resumed runs where the same config is re-applied.

Test plan

• Resume a wandb run and verify config updates don't crash
• Verify new config values are reflected in the wandb dashboard

Summary by CodeRabbit

• Bug Fixes
  • Improved hyperparameter logging to correctly handle configuration updates when values change across experiment runs.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 78131fa3-6f77-4557-8da4-cae39c6dd7f7

📥 Commits

Reviewing files that changed from the base of the PR and between 9feb4b0 and 57b9da1.

📒 Files selected for processing (1)

• nemo_rl/utils/logger.py

---

📝 Walkthrough

Walkthrough

Updated hyperparameter logging in both WandbLogger and SwanlabLogger classes to pass allow_val_change=True parameter to the self.run.config.update() method, enabling hyperparameter value changes during repeated updates in run configurations.

Changes

Cohort / File(s)	Summary
Hyperparameter Logging Configuration nemo_rl/utils/logger.py	Updated log_hyperparams methods in WandbLogger and SwanlabLogger to pass allow_val_change=True to self.run.config.update() calls, allowing hyperparameter values to be updated when they differ on repeated configurations.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title directly addresses the main change: enabling wandb config value changes during run resume by passing allow_val_change=True to config.update() calls.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Test Results For Major Changes	✅ Passed	PR contains only minor changes (2-line bug fix) adding allow_val_change=True parameter to config.update() calls without introducing features, breaking changes, or significant refactoring.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Tip

CodeRabbit can use Trivy to scan for security misconfigurations and secrets in Infrastructure as Code files.

Add a .trivyignore file to your project to customize which findings Trivy reports.

[yuki-97]

thanks @gkaplun-nvidia , lgtm

[yuki-97]

@gkaplun-nvidia it looks DCO check failed, could you follow the guide to fix it?
https://github.com/NVIDIA-NeMo/RL/pull/2137/checks?check_run_id=68070517476

[terrykong]

lgtm. @gkaplun-nvidia could you signoff the commits to pass the DCO check?

[gkaplun-nvidia]

@terrykong, done.

[yuki-97]

/ok to test 23a9cdf

[yuki-97]

/ok to test c6eac01

[yuki-97]

/ok to test 58fb411

[yuki-97]

hi @gkaplun-nvidia , there's some unit test fail in https://github.com/NVIDIA-NeMo/RL/actions/runs/23744949284/job/69172069492?pr=2137, could you help to fix?

for the fail in fast_L1, it should be unrelated to this PR, we'll take a look.

[yuki-97]

/ok to test 20faf7f