Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
318 Open 1,831 Closed
318 Open 1,831 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

Change Review Process
#3659 opened Mar 2, 2026 by Phlip79 Loading…
6 tasks
1
Update setup.py community-request complexity: low Expert Review Apply this label to indicate that your PR is ready for expert review.
#3658 opened Mar 2, 2026 by sakgoyal Loading… Core 0.16
2
Prefix caching | Mamba memory only.
#3657 opened Mar 2, 2026 by lmcafee-nvidia Loading…
4 tasks
@lmcafee-nvidia
1
add the logic to enable chunked mlp during training community-request
#3656 opened Mar 2, 2026 by pengdurice Loading…
3 of 6 tasks
1
add mix_hidden_states option in conversion Expert Review Apply this label to indicate that your PR is ready for expert review. Final Review PR is in the "final review" stage
#3655 opened Mar 2, 2026 by yeyu-nvidia Loading…
6 tasks
Core 0.16
Implement responses API for inference
#3654 opened Mar 2, 2026 by tdene • Draft
6 tasks
1
Prevent double serialization inside Flask server Expert Review Apply this label to indicate that your PR is ready for expert review.
#3653 opened Mar 2, 2026 by tdene Loading…
6 tasks
Core 0.16
19
feat(checkpoint): zero-copy storage sharing in CheckpointWithoutOutput complexity: low Final Review PR is in the "final review" stage
#3649 opened Mar 2, 2026 by Victarry Loading…
5 tasks done
@Victarry
1
Offload Flask frontend to separate process
#3648 opened Mar 1, 2026 by santhnm2 Loading…
6 tasks
Core 0.16
2
Fix TransformerConfig validation for mixed dense/MoE upcycling community-request
#3647 opened Mar 1, 2026 by rkteddy Loading…
1 of 6 tasks
3
Fix upcycling state dict conversion for mixed dense/MoE models community-request Expert Review Apply this label to indicate that your PR is ready for expert review.
#3646 opened Mar 1, 2026 by rkteddy Loading…
1 of 6 tasks
2
Enhance and fix NVTX for training complexity: low Expert Review Apply this label to indicate that your PR is ready for expert review.
#3642 opened Feb 28, 2026 by yaox12 Loading…
6 tasks
3
Prefork Dataloader worker before CUDA Init
#3640 opened Feb 27, 2026 by hexinw-nvidia • Draft
2
refactor to support emerging optimizers beyond muon
#3638 opened Feb 27, 2026 by FDecaYed Loading…
6 tasks
1
[Draft][main] enable manual_dgrad_release for tst Run MBridge tests Attach this for testing this PR against MBridge main
#3637 opened Feb 27, 2026 by Wohox • Draft
6 tasks
1
[Draft][dev] enable manual_dgrad_release for tst Run MBridge tests Attach this for testing this PR against MBridge main
#3636 opened Feb 27, 2026 by Wohox • Draft
6 tasks
Core 0.16
2
feat: support unified model bagel with generation and understanding community-request
#3635 opened Feb 27, 2026 by sophiayyya • Draft
1 of 6 tasks
1
Fix illegal memory access with mamba inference Expert Review Apply this label to indicate that your PR is ready for expert review.
#3631 opened Feb 26, 2026 by tdene Loading…
6 tasks
Core 0.16
3
fix: handle zero-size tensors in MoE token dispatchers community-request Expert Review Apply this label to indicate that your PR is ready for expert review. needs-follow-up Issue needs follow-up
#3626 opened Feb 26, 2026 by callum-ward-inflection Loading…
6 tasks done
1
3
Fix token dispatched cudagraph_attrs Final Review PR is in the "final review" stage
#3625 opened Feb 26, 2026 by asolergi-nv Loading…
6 tasks
Core 0.16
2
Correctly generate state dict in MultiTokenPredictionBlock Final Review PR is in the "final review" stage
#3624 opened Feb 26, 2026 by asolergi-nv Loading…
6 tasks
Core 0.16
3
[dev] refactor to support emerging optimizers beyond muon
#3618 opened Feb 26, 2026 by FDecaYed Loading…
6 tasks
Core 0.16
19
Upgrade GitHub Actions to latest versions community-request needs-follow-up Issue needs follow-up
#3609 opened Feb 26, 2026 by salmanmkc Loading…
1
Fix cp and not per token loss calculation in schedules.py
#3607 opened Feb 26, 2026 by wplf Loading…
6 tasks
8
[feature] MegaScope Tensor Tracer community-request
#3606 opened Feb 26, 2026 by superay-a Loading…
4 of 6 tasks
2
ProTip! Mix and match filters to narrow down what you’re looking for.