NVIDIA-NeMo / NVIDIA-NeMo/RL

What is the difference between nemo-rl and megatron-bridge?

Open
#2,934 5 comments 0 reactions 1 assignee Claimed by @terrykong View on GitHub
community-request Documentation enhancement waiting-on-maintainers
Dominant language
Python
Stars
2k
Forks
561
Avg merge
4d 5h
Merged PRs (30d)
145

Description

Hi, I note that there are two different receipes for model training, one in nemo-rl (with both sft and rl), the other one is megatron-bridge.

What is the point of having two different frameworks? If I intend to analyze the post-training steps or extending the post training frameworks, which one is the best option?

I think they are very similar, except for the sft stage of nemotron-super-3. Thanks a lot!

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.