Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
248 Open 1,215 Closed
248 Open 1,215 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Fix missing propagation of RoPE scaling and interpolation arguments in model builders community-request
#2902 opened Jan 11, 2026 by taehwakkwon Loading…
1 of 6 tasks
1
Revert "[dev] Add assertion for mxfp8 params without dp overlap (#2270)" dev2main: mbridge dev to main: this PR is needed in main for mbridge Final Review Apply this label to indicate that your PR is ready for final review.
#2901 opened Jan 10, 2026 by ko3n1g Loading…
6 tasks
Core 0.16
1
Cherrypick 1989
#2900 opened Jan 10, 2026 by ko3n1g • Draft
6 tasks
Core 0.16
2
Ko3n1g/chore/stack prs 260110
#2899 opened Jan 10, 2026 by ko3n1g Loading…
6 tasks
Core 0.16
Remove unused FlashAttention3 args
#2898 opened Jan 10, 2026 by santhnm2 Loading…
6 tasks
Core 0.16
2
Ensure that last prefill chunk is handled correctly by Mamba models Run functional tests
#2897 opened Jan 10, 2026 by santhnm2 Loading…
6 tasks
Core 0.16
3
Maanug/transfcfg generated args v2
#2896 opened Jan 10, 2026 by maanug-nv • Draft
6 tasks
1
fix(dist_ckpt): Remove step tensor from sharded param state to fix MoE/EP checkpoint validation community-request
#2895 opened Jan 10, 2026 by jreiml Loading…
1
[Megatron-FSDP] Test FP8 activations + parameter sharding with Megatron-FSDP fully-shard. Update README. Final Review Apply this label to indicate that your PR is ready for final review.
#2894 opened Jan 10, 2026 by cspades Loading…
3 of 6 tasks
Core 0.16
@cspades
1
Use different token for assign logic
#2893 opened Jan 9, 2026 by Phlip79 Loading…
6 tasks
Core 0.16
2
Add unit test for LatentMoE community-request complexity: low
#2892 opened Jan 9, 2026 by venmugil Loading…
6 tasks
Core 0.16
@venmugil
3
Support custom Router implementations in MoELayer community-request
#2891 opened Jan 9, 2026 by nschank Loading…
2 of 6 tasks
3
ci: Remove Github transition comment from CI
#2881 opened Jan 8, 2026 by chtruong814 Loading…
6 tasks
Core 0.16
Test on Muon integration Run CICD
#2880 opened Jan 8, 2026 by BoxiangW Loading…
6 tasks
Core 0.16
6
RL refit pipelining support
#2878 opened Jan 8, 2026 by wdykas Loading…
6 tasks
Core 0.16
20
Fix GPU memory leak during in-process restart
#2875 opened Jan 8, 2026 by ahmadki • Draft
6 tasks
1
handle 'step' state in checkpoint save
#2874 opened Jan 8, 2026 by ahmadki • Draft
6 tasks
1
1
Fix stale parameter index map in DistributedOptimizer
#2872 opened Jan 8, 2026 by ahmadki • Draft
6 tasks
1
1
Add a logprobs test with real gpt model. Expert Review Apply this label to indicate that your PR is ready for expert review. Run tests
#2870 opened Jan 8, 2026 by yobibyte Loading…
6 tasks
Core 0.16
2
[WIP Feat] Split-K Indexer Kernels community-request needs-follow-up Issue needs follow-up
#2869 opened Jan 8, 2026 by laixinn • Draft
7 of 17 tasks
1
Only build datasets on the required ranks
#2865 opened Jan 8, 2026 by asolergi-nv Loading…
1
Remove cross-rank synchronization during checkpoint load & deprecate torch.distributed.checkpoint.state_dict_loader.load_state_dict
#2864 opened Jan 8, 2026 by asolergi-nv Loading…
1
Update Slack user group when oncall changes
#2859 opened Jan 8, 2026 by Phlip79 Loading…
6 tasks
Core 0.16
3
Use global user buffer when the bucket size does not fit FixedPoolAllocator
#2857 opened Jan 7, 2026 by shengf-nv Loading…
6 tasks
1
Initialize M-FSDP Optimizer(DistOpt v2) with DTensor Parameter Support
#2848 opened Jan 7, 2026 by shjwudp • Draft
6 tasks
1
ProTip! What’s not been updated in a month: updated:<2025-12-11.