NVIDIA / NVIDIA/Megatron-LM

Sequence Packing + LR adjustment for RL

Open
#3,078 0 comments 0 reactions 1 assignee Claimed by @yobibyte View on GitHub
bug module: rl
Dominant language
Python
Stars
17.9k
Forks
4.5k
Avg merge
4d 6h
Merged PRs (30d)
271

Description

I think, sequence packing's RL code path does not adjust the learning rate properly. I suspect this is because we adjust learning rate by the number of bins, and not by the number of actual sequences we processed.

I will have a look, but @jalbericiola for visibility.

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.