add trajectory_id to TrajectoryStep #675
Merged
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Outdated suggestions cannot be applied.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.
Description
Adds
trajectory_idtoTrajectoryStep, which is useful for training multi-agent systems. In those, there are several mostly independent, multi-turn rollouts. We want to interleave the trajectory steps for each of those rollouts for efficiency, but must keep the rollouts of the different agents separate. This PR adds atrajectory_idto enable the correct grouping and interleaving of rollouts for training.Type of Change
Testing
uv run pytestlocally.Checklist
Additional Notes
Note
Adds per-step
trajectory_idto enable grouping/interleaving of trajectories and wires it through state and environments.TrajectoryStepwithtrajectory_id: strstate["trajectory_id"] = uuid4().hexinEnvironment.init_statetrajectory_idto steps inMultiTurnEnv.add_model_responseand RLM sub-LLM steps (batch_id_request_id); preserveis_truncatedtrajectory_idin allTrajectoryStepconstructions and adjust token fields; update docs to reflect new fieldsWritten by Cursor Bugbot for commit 1b27db6. This will update automatically on new commits. Configure here.