fix: log validation data

[parthchadha]

What does this PR do ?

Add a one line overview of what this PR aims to accomplish.

Issues

closes #1761

Usage

• You can potentially add a usage example below

# Add a code snippet demonstrating how to use this

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you run the unit tests and functional tests locally? Visit our Testing Guide for how to run tests
• Did you add or update any necessary documentation? Visit our Document Development Guide for how to write, build and test the docs.

Additional Information

• ...

Summary by CodeRabbit

Release Notes

New Features

• Added optional logging capability for validation data during training, enabling JSONL export of validation metrics
• Logger parameter is optional and backward compatible—validation works with or without logging enabled
• Improved visibility into model performance throughout the training process

Tests

• Added unit tests covering validation logging scenarios, including logger-enabled paths and configurations without a logger

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[coderabbitai]

📝 Walkthrough

Walkthrough

The changes add optional logger parameter propagation through GRPO validation functions. When a logger is provided to validate(), it writes validation data to JSONL files named val_data_step{step}.jsonl. The grpo_train and async_grpo_train functions now pass their logger instances to validate() calls, with new unit tests covering both logger-enabled and logger-free execution paths.

Changes

Cohort / File(s)	Summary
Core algorithm updates nemo_rl/algorithms/grpo.py	Updated validate() function signature to accept optional logger parameter. Modified grpo_train() and async_grpo_train() to pass logger=logger when calling validate(). Added conditional JSONL logging of validation data when logger is provided.
Unit test coverage tests/unit/algorithms/test_grpo.py	Added three new test cases for validate() function: test_validate_logs_data_when_logger_provided (verifies JSONL file creation with correct data keys), test_validate_works_without_logger (validates execution without logger), and test_validate_returns_empty_when_no_dataloader (validates empty metrics handling). Tests use mocks for environment, rollout, and configuration objects.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

🚥 Pre-merge checks | ✅ 3 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Test Results For Major Changes	⚠️ Warning	PR introduces major feature (validation data logging) but provides no documentation of testing, test results, or verification in the description.	Update PR description to document: (1) summary of three unit tests and their verification scope, (2) confirmation that existing GRPO validation behavior remains unchanged, (3) test suite results confirming no regressions introduced.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	Docstring coverage is 87.50% which is sufficient. The required threshold is 80.00%.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'fix: log validation data' directly and clearly summarizes the main change—adding logging functionality to validation data.

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

✨ Finishing touches

• 📝 Generate docstrings

---

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.}