Skip to content

feat: implement select_dataset() for dynamic dataset filtering - #3

Open
Kerwin0224 wants to merge 1 commit into
slime-0.2.2from
feature/dynamic-dataset-select
Open

feat: implement select_dataset() for dynamic dataset filtering#3
Kerwin0224 wants to merge 1 commit into
slime-0.2.2from
feature/dynamic-dataset-select

Conversation

@Kerwin0224

Copy link
Copy Markdown
Collaborator

Summary

Add select_dataset() to RolloutDataSource for dynamic dataset adjustment during RL training.

Changes

  • Sample: add global_index field — stable identifier, invariant across rollouts
  • Dataset: assign global_index at load time after length filtering
  • RolloutDataSource: add select_dataset(), expose full_samples, update save()/load() for checkpoint compatibility

Upstream adjust_dataset() on RolloutManager will call this method — not included in this PR.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant