[data] feat: dump train/test example as JSON

[wantbook-book]

What does this PR do?

This PR adds functionality to save one training and one testing example as JSON files for reference, making it easier to inspect dataset formatting and preprocessing.
Related to potential future debugging and reproducibility improvements.

Checklist Before Starting

• Search for similar PRs. Paste at least one query link here: ...
• Format the PR title as [{modules}] {type}: {description} (This will be checked by the CI)
  • {modules} include fsdp, megatron, sglang, vllm, rollout, trainer, ci, training_utils, recipe, hardware, deployment, ray, worker, single_controller, misc, perf, model, algo, env, tool, ckpt, doc, data
  • If this PR involves multiple modules, separate them with , like [megatron, fsdp, doc]
  • {type} is in feat, fix, refactor, chore, test
  • If this PR breaks any API (CLI arguments, config, function signature, etc.), add [BREAKING] to the beginning of the title.
  • Example: [BREAKING][fsdp, megatron] feat: dynamic batching

Test

Manually verified that two files train_example.json and test_example.json are saved correctly in the specified local_dir.

API and Usage Example

This change does not alter the public API.

Design & Code Changes

• Added code to save train_dataset[0] and test_dataset[0] as JSON files in local_dir

• Helps with quick inspection and reproducibility of dataset inputs

Checklist Before Submitting

Important

Please check all the following items before requesting a review, otherwise the reviewer might deprioritize this PR for review.

• Read the Contribute Guide.
• Apply pre-commit checks: pre-commit install && pre-commit run --all-files --show-diff-on-failure --color=always
• Add / Update the documentation.
• Add unit or end-to-end test(s) to the CI workflow to cover all the code. If not feasible, explain why: easy code
• Once your PR is ready for CI, send a message in the ci-request channel in the verl Slack workspace.

[gemini-code-assist]

Code Review

This PR adds a useful feature for inspecting dataset examples. I've identified a potential IndexError if the train or test datasets are empty and suggested a fix to make the script more robust.

[gemini-code-assist]

The current implementation accesses train_dataset[0] and test_dataset[0] directly. This will raise an IndexError and crash the script if either dataset is empty. An empty dataset is a valid edge case, for example, if the source dataset has empty splits or if the mapping/filtering results in an empty dataset.

To make the script more robust, you should add a check to ensure the datasets are not empty before attempting to access their elements.

Suggested change

	example = train_dataset[0]
	with open(os.path.join(local_dir, "train_example.json"), "w") as f:
	json.dump(example, f, indent=2)
	example = test_dataset[0]
	with open(os.path.join(local_dir, "test_example.json"), "w") as f:
	json.dump(example, f, indent=2)
	if len(train_dataset) > 0:
	example = train_dataset[0]
	with open(os.path.join(local_dir, "train_example.json"), "w") as f:
	json.dump(example, f, indent=2)
	if len(test_dataset) > 0:
	example = test_dataset[0]
	with open(os.path.join(local_dir, "test_example.json"), "w") as f:
	json.dump(example, f, indent=2)