Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
366 Open 2,462 Closed
366 Open 2,462 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(ci): re-enable scoped_cudagraph MoE test with TE 2.14 golden values complexity: medium
#4444 opened Apr 23, 2026 by buptzyb Contributor Loading… Core 0.16
2
Handle SSM sharded tensor merge OOM with CPU fallback community-request
#4442 opened Apr 23, 2026 by returnL Contributor Loading…
2 of 5 tasks
1
2
Inference: Add the embedding and output layer in the full_iteration_inference cuda graph scope for hybrid models complexity: low
#4440 opened Apr 23, 2026 by sidsingh-nvidia Contributor Loading…
5 tasks
Core 0.16
2
Added NVFP4 MTP layer support
#4439 opened Apr 23, 2026 by sanandaraj5597 Contributor Loading…
1
docs: fix broken links and anchors across READMEs and docs docs-only documentation only (docs or docstrings)
#4438 opened Apr 23, 2026 by sbhavani Contributor Loading…
1 of 5 tasks
2
docs: fix Python version requirement and uv install commands in install docs-only documentation only (docs or docstrings)
#4437 opened Apr 23, 2026 by sbhavani Contributor Loading…
5 tasks
2
chore: nightly sync main into dev (22_04_2026) complexity: high Run functional tests
#4436 opened Apr 22, 2026 by svcnvidia-nemo-ci Loading… Core 0.16
5
Draft: Refactor APIs for FSDP2 Drop-in Compatibility module: megatron-fsdp
#4435 opened Apr 22, 2026 by shjwudp Contributor • Draft
5 tasks
@shjwudp @Autumn1998
1
get rid of weights_only=False complexity: low Expert Review [deprecated] Apply this label to indicate that your PR is ready for expert review. Run functional tests
#4434 opened Apr 22, 2026 by dimapihtar Contributor Loading…
5 tasks
Core 0.16
2
fix cudagraph nonconvergence
#4433 opened Apr 22, 2026 by jiemingz Contributor • Draft
5 tasks
2
Reorder mtp_post_process after attn backward in 1F1B schedule plan
#4430 opened Apr 22, 2026 by gdengk Contributor Loading…
5 tasks
2
Adding code for Flextron complexity: high
#4429 opened Apr 22, 2026 by sheliang-nv Loading…
2 of 5 tasks
Core 0.16
@JRD971000
6
[DO NOT MERGE] Cye/mfsdp devgrad revert
#4428 opened Apr 22, 2026 by cspades Member • Draft
5 tasks
@cspades
1
[Main] Fix invisible issues related to use_decoupled_grad for Megatron-FSDP. complexity: low Final Review PR is in the "final review" stage
#4427 opened Apr 22, 2026 by cspades Member • Draft
5 tasks
Core 0.16
@cspades
3
[Dev] Fix invisible issues related to decoupled gradient for Megatron-FSDP.
#4426 opened Apr 22, 2026 by cspades Member • Draft
5 tasks
Core 0.16
@cspades
1
Add misc CUDA graph sugar to CudaGraphManager complexity: low
#4425 opened Apr 22, 2026 by tdene Contributor Loading…
5 tasks
Core 0.16
8
[DEV] fix(megatron-fsdp): compute SWiGLU/GDN split in item coordinates for non-DTensor optimizer states
#4424 opened Apr 22, 2026 by xuwchen Contributor Loading…
5 tasks
1
fix(megatron-fsdp): compute SWiGLU/GDN split in item coordinates for non-DTensor optimizer states complexity: low
#4423 opened Apr 22, 2026 by xuwchen Contributor Loading…
5 tasks
4
feat(ckpt): expose validate_access_integrity knob on dist-ckpt load complexity: low
#4422 opened Apr 22, 2026 by asolergi-nv Contributor Loading…
5 tasks
1
fix: NVRx async compatibility and defer resiliency import complexity: medium core_r0.17.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge. Expert Review [deprecated] Apply this label to indicate that your PR is ready for expert review. Final Review PR is in the "final review" stage Run functional tests
#4420 opened Apr 22, 2026 by sbak5 Contributor Loading…
5 tasks
Core 0.16
12
[Inference] Protocol Robustness Improvements in Coordinator (Follow-up to #4176) community-request
#4419 opened Apr 21, 2026 by DhineshPonnarasan Contributor Loading…
3
Unify and refactor Megatron-FSDP documentation. complexity: medium Expert Review [deprecated] Apply this label to indicate that your PR is ready for expert review. module: documentation module: megatron-fsdp
#4418 opened Apr 21, 2026 by cspades Member Loading…
5 tasks
Core 0.16
@cspades
20
style: modernize typing annotations in param_and_grad_buffer.py complexity: medium module: megatron-fsdp Run tests
#4417 opened Apr 21, 2026 by wujingyue Contributor Loading…
1 task
Core 0.16
4
refactor: simplify FixedPoolAllocator constructor inputs complexity: medium Run tests
#4415 opened Apr 21, 2026 by wujingyue Contributor Loading…
3 of 4 tasks
Core 0.16
5
Deprecate static inference
#4413 opened Apr 21, 2026 by santhnm2 Contributor • Draft
5 tasks
1
ProTip! Type g p on any issue or pull request to go back to the pull request listing page.