Refactors PolicyTrainerRayProcess so that the train method uses a dict of data
#1240
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
finbarrtimbers
changed the title
PolicyRayTrainActor so that the train method uses a dict of dataPolicyTrainerRayProcess so that the train method uses a dict of data
Contributor
Documentation Changes Detected📄
|
Contributor
Documentation Changes Detected📄
|
Contributor
Documentation Changes Detected📄
|
Contributor
Documentation Changes Detected📄
|
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
hamishivi
approved these changes
Dec 3, 2025
Collaborator
hamishivi
left a comment
hamishivi
left a comment
There was a problem hiding this comment.
Generally LGTM, one nit pick around using dicts to pass stuff around.
- Rename queue_types.py to data_types.py (avoiding stdlib conflict) - Add CollatedBatchData dataclass with __getitem__ and __len__ methods - Update grpo_fast.py to use dataclass instead of dict - Update all imports across affected files 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This lets us clean it up by using comprehensions and loops instead of making repeated calls.
Also switches the variables in
trainto use Shazeer style shape suffixes.Runs:
Note
Refactors GRPO training to use a typed CollatedBatchData instead of dicts, updates the trainer/train flow and metrics, and migrates PromptRequest/Result types from queue_types to data_types across code and tests.
CollatedBatchDatathroughout (prepare_collated_data_for_workers,compute_logprobs,calculate_token_counts,train).trainsignature totrain(data_BT, pad_token_id); derive minibatching fromargs.num_mini_batchesand operate with shape-suffix variables (e.g.,*_BT).one_training_stepto passdata_BT.CollatedBatchDatadataclass with tensor lists, slicing, and length helpers.PromptRequest,GenerationResult,RequestInfo,TokenStatisticshere (used repo-wide).CollatedBatchData(field checks, iteration) and adjust imports.open_instruct.queue_typestoopen_instruct.data_typesin code and tests, includingvllm_utils.pyand benchmarking script.Written by Cursor Bugbot for commit 9f14736. This will update automatically on new commits. Configure here.