🧶 feat: Add WeaveCallback for W&B Weave integration

[parambharat]

What does this PR do?

This PR adds a new WeaveCallback that integrates TRL trainers with W&B Weave to trace of model predictions and evaluations during training.

Key features:

• Tracing Mode: Logs prompts and model completions to weave to simplify debugging.
• Evaluation Mode: Logs predictions with custom scoring functions and summary metrics
• Uses Weave's EvaluationLogger for consistent, structured data logging

Changes:

• Added WeaveCallback class in trl/trainer/callbacks.py
• Added is_weave_available() utility in trl/import_utils.py
• Exported WeaveCallback in trl/trainer/__init__.py

Usage:

# Tracing mode (predictions only)
weave_callback = WeaveCallback(trainer=trainer)

# Evaluation mode (predictions + custom scoring)
weave_callback = WeaveCallback(
    trainer=trainer,
    scorers={"accuracy": accuracy_scorer}
)
trainer.add_callback(weave_callback)

The callback requires trainers to have an evaluation dataset with a "prompt" column and generates completions during evaluation steps for logging to Weave.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline, Pull Request section?
• Was this discussed/approved via a GitHub issue? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. - @qgallouedec cc: @ayulockin

[qgallouedec]

LGTM!

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.