Skip to content

feat(slime): Add integration of slime + rollout gateway. - #106

Open
Linbo-Liu wants to merge 1 commit into
awslabs:mainfrom
Linbo-Liu:slime
Open

feat(slime): Add integration of slime + rollout gateway.#106
Linbo-Liu wants to merge 1 commit into
awslabs:mainfrom
Linbo-Liu:slime

Conversation

@Linbo-Liu

Copy link
Copy Markdown
Contributor

Issue #, if available:

Description of changes:

Summary

  • Adds backends/experimental/slime/integration/ with two slime hook entry points:
    • rollout.generate (--custom-generate-function-path): per-sample ACR rollout via RolloutGateway + RolloutClient; maps trajectory leaves to slime Samples; failures return ABORTED with remove_sample=True
    • rewards.normalize_episode_rewards (--custom-reward-post-process-path): GRPO normalization that groups by group_index + dedups by rollout_id, fixing slime's positional reshape when a rollout forks into multiple leaves
  • Adds GSM8K example (examples/math_agent/) with train.sh, config.yaml.example, SETUP.md, and install_slime.sh

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

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.

2 participants