Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
330 Open 1,994 Closed
330 Open 1,994 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

Add torch grouped gemm bf16 and mxfp8 support w/ cuda graphed + inference_optimized MoEs
#3858 opened Mar 13, 2026 by sidsingh-nvidia • Draft
5 tasks
Core 0.16
2
Fix flakiness due to timing between shutdowns complexity: low
#3857 opened Mar 13, 2026 by tdene Loading…
5 tasks
Core 0.16
1
MFU tracking for inference
#3856 opened Mar 13, 2026 by tdene • Draft
5 tasks
1
enable async save for functional tests complexity: low Expert Review [deprecated] Apply this label to indicate that your PR is ready for expert review. Run functional tests
#3855 opened Mar 13, 2026 by dimapihtar Loading…
5 tasks
Core 0.16
2
remove legacy mpu complexity: low Expert Review [deprecated] Apply this label to indicate that your PR is ready for expert review.
#3854 opened Mar 13, 2026 by dimapihtar Loading…
5 tasks
Core 0.16
2
remove legacy data complexity: high Expert Review [deprecated] Apply this label to indicate that your PR is ready for expert review.
#3853 opened Mar 13, 2026 by dimapihtar Loading…
5 tasks
Core 0.16
2
Separate save_checkpoint into per-type execution paths community-request
#3852 opened Mar 13, 2026 by Anmol202005 Loading…
5 tasks done
2
Nemo-RL integration bugfixes for --transformer-impl inference_optimized Approved All necessary approvals have been made complexity: medium Run functional tests
#3851 opened Mar 13, 2026 by sidsingh-nvidia Loading…
5 tasks
Core 0.16
@sidsingh-nvidia
7
fix: Handle quantized CUDA tensors in async checkpoint writer complexity: low
#3845 opened Mar 12, 2026 by sbak5 Loading…
5 tasks
Core 0.16
6
Fsdp dsv3 proxy complexity: medium Run functional tests
#3844 opened Mar 12, 2026 by gautham-kollu Loading…
5 tasks
Core 0.16
5
Add activation logging and tokens per expert logging
#3842 opened Mar 12, 2026 by Mellonta • Draft
5 tasks
1
Miscellaneous inference bug fixes complexity: high Run functional tests
#3840 opened Mar 12, 2026 by santhnm2 Loading…
5 tasks
Core 0.16
8
[dev] fix no_shard training convergency and add unittest for no_shard
#3835 opened Mar 12, 2026 by wplf Loading…
1
[DEV] fix(megatron-fsdp): build expt_device_mesh only for MoE models dev branch Dev branch related issues and development Expert Review [deprecated] Apply this label to indicate that your PR is ready for expert review. module: megatron-fsdp
#3832 opened Mar 12, 2026 by xuwchen Loading…
5 tasks
Core 0.16
@xuwchen
3
fix(megatron-fsdp): build expt_device_mesh only for MoE models Expert Review [deprecated] Apply this label to indicate that your PR is ready for expert review. module: megatron-fsdp
#3831 opened Mar 12, 2026 by xuwchen • Draft
5 tasks
Core 0.16
@xuwchen
3
[dev] mHC kernel fusion complexity: high
#3828 opened Mar 12, 2026 by jingqiny-99 Loading…
2 of 5 tasks
Core 0.16
27
chore: Move to Py3.12
#3826 opened Mar 12, 2026 by ko3n1g • Draft
5 tasks
Core 0.16
2
[Dev] feat(fsdp): use TE general_gemm for mixed-precision wgrad in FSDP path
#3824 opened Mar 12, 2026 by Victarry Loading…
1 of 2 tasks
2
feat(fsdp): use TE general_gemm for mixed-precision wgrad in FSDP path complexity: low Final Review PR is in the "final review" stage
#3822 opened Mar 12, 2026 by Victarry Loading…
2 tasks done
Core 0.16
4
[Main] Support EP Overlap's Dynamic Computation Stream For Full-Iter CUDA Graph
#3820 opened Mar 12, 2026 by Wohox • Draft
5 tasks
1
Fix KV block allocation for chunked prefill
#3819 opened Mar 12, 2026 by jon-barker • Draft
5 tasks
@jon-barker
2
[Dev] Support EP Overlap's Dynamic Computation Stream For Full-Iter CUDA Graph
#3818 opened Mar 12, 2026 by Wohox • Draft
5 tasks
1
Add unit tests for speculative decoding complexity: medium
#3817 opened Mar 12, 2026 by santhnm2 Loading…
5 tasks
Core 0.16
3
Ensure that inference dummy_forward does not try to match on a cudagraph when running eager Approved All necessary approvals have been made complexity: low
#3815 opened Mar 12, 2026 by mathemakitten Loading…
5 tasks
Core 0.16
5
Add Lion optimizer support complexity: low Final Review PR is in the "final review" stage
#3813 opened Mar 11, 2026 by mchrzanowski Loading…
@mkhona-nvidia
4
ProTip! Filter pull requests by the default branch with base:main.