From NVIDIA Megatron-LM for visibility#18
Open
RaymondLi0 wants to merge 6333 commits intobigcode-project:multi-query-attentionfrom
Open
From NVIDIA Megatron-LM for visibility#18RaymondLi0 wants to merge 6333 commits intobigcode-project:multi-query-attentionfrom
RaymondLi0 wants to merge 6333 commits intobigcode-project:multi-query-attentionfrom
Commits
This pull request is big! We're only showing the most recent 250 commits
Commits on Jan 26, 2026
- authored
- authored
- authored
- authored
- andauthored
- authored
- authored
- authored
- authored
Commits on Jan 27, 2026
- authored
- committed
- authored
- authored
- authored
- authored
- andauthored
- authored
- authored
- authored
- authored
- authored
- authored
Commits on Jan 28, 2026
- authored
- committed
- authored
- authored
- andauthored
- authored
- andauthored
- andauthored
- authored
- committed
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?v=4&size=32){kind=avatar}
Commits on Jan 29, 2026
- committed
- authored
- authored
- andauthored
- andauthored
- authored
- andauthored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
Commits on Jan 30, 2026
- committed
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- committed
- andauthored
- authored
- authored
Commits on Jan 31, 2026
Commits on Feb 1, 2026
Commits on Feb 2, 2026
- andauthored
- authored
- andauthored
- authored
- andauthored
- committed
- committed
- andauthored
- authored
Commits on Feb 3, 2026
- andauthored
- authored
- authored
- authored
- andauthored
- authored
- authored
- authored
- andauthored
- authored
Add a unit test to check that RL `get_logprobs` will reuse training cudagraphed forward pass (#3209)
authored- authored
- authored
- authored
- authored
- authored
Commits on Feb 4, 2026
- committed
- authored
- authored
- authored
- authored
- authored
- andauthored
- authored
- authored
- andauthored
- authored
- authored
- committed
- committed
- authored
- authored
Commits on Feb 5, 2026
- andauthored
- committed
- authored
- andauthored
- andauthored
- authored
- authored
- authored
Commits on Feb 6, 2026
- committed
- authored
- authored
- authored
- authored
- authored
- authored
Commits on Feb 7, 2026
- authored
- authored
- committed
- authored
- andauthored
- andauthored
- andauthored
- authored
- authored
Commits on Feb 9, 2026
- andauthored
- andauthored
- andauthored
- andauthored
- andauthored
- authored
- authored
- authored
- authored
Commits on Feb 10, 2026
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- andauthored
- authored
- authored
- authored
- authored
Commits on Feb 11, 2026
- committed
- authored
- andauthored
- andauthored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
- andauthored
- authored
- authored
- authored
- authored
Commits on Feb 12, 2026
- committed
- andauthored
- authored
- authored
- authored
- authored
- andauthored
- andauthored
- authored
- authored
Commits on Feb 13, 2026
- committed
- authored
- committed
- authored
- authored
- authored
- authored
- authored
- andauthored
- authored
- andauthored
Commits on Feb 14, 2026
Commits on Feb 16, 2026
Commits on Feb 17, 2026
- andauthored
- authored
- authored
- authored
- authored
- authored
- authored
- authored
Commits on Feb 18, 2026
- committed
- committed
- authored
- andauthored
- authored
- andauthored
- authored
- authored
- authored
- authored
Commits on Feb 19, 2026
- authored
- authored
- andauthored
- authored
- authored
- authored
- committed
- authored
- authored
- authored
- authored
Commits on Feb 20, 2026
[OMNIML-3232] ModelOpt: add full TE spec option and wire Mamba stack / post-training scripts (#3393)
authored- authored
- committed
- authored
- andauthored
- authored
- authored
- authored
