verifiers_rl recipe updates for verifiers v0.1.8 release #133
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Reorganization of data flow for the
verifiers_rlrecipe to utilize new internal structure fromverifiers>=0.1.8, including encapsulation of rollout data within aStateobject which can be more directly translated to the tinker TrajectoryGroup format.This allows maintaining only a single generic OpenAI client (rather than one per-rollout with a hook for data tracking), and adds support for groupwise rewards from verifiers
Rubricclasses.Tested on the
reverse-texttraining example.