Decentralized RL Training at Scale

607 stars 98 forks 607 watchers Python Apache License 2.0
9 Open Issues Need Help Last updated: Sep 17, 2025

Open Issues Need Help

View All on GitHub
good first issue

Decentralized RL Training at Scale

Python
enhancement good first issue

Decentralized RL Training at Scale

Python

AI Summary: This issue proposes renaming the variable and configuration key `outputs_dir` to `output_dir` across the entire repository. The change is suggested to adopt a more conventional and standard naming practice commonly found in Python and Machine Learning projects.

Complexity: 3/5
enhancement good first issue

Decentralized RL Training at Scale

Python

AI Summary: This issue requests adding documentation that guides users on how to set up a GitHub SSH key. This setup is a prerequisite that needs to be completed before running the project's `install.sh` script, which presumably relies on SSH access to GitHub.

Complexity: 1/5
documentation good first issue

Decentralized RL Training at Scale

Python
enhancement good first issue

Decentralized RL Training at Scale

Python
enhancement good first issue

Decentralized RL Training at Scale

Python

AI Summary: This issue proposes saving evaluation results, including prompts, completions, and rewards, as a Hugging Face dataset to disk. This will leverage the `verifiers` `Environment.make_dataset` functionality and should be implemented within the evaluation utility function. The aim is to simplify manual assessment of evaluation outcomes.

Complexity: 2/5
enhancement good first issue

Decentralized RL Training at Scale

Python

Decentralized RL Training at Scale

Python