Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
399 Open 2,564 Closed
399 Open 2,564 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 inference unit test complexity: low
#4589 opened May 2, 2026 by maanug-nv Contributor Loading…
5 tasks
@maanug-nv
2
Inference bug-fix: Re-enable EP syncs for the legacy A2A dispatcher complexity: low Final Review PR is in the "final review" stage
#4587 opened May 1, 2026 by sidsingh-nvidia Contributor Loading…
5 tasks
2
observability: add metrics instrumentation to DataParallelInferenceCoordinator community-request
#4586 opened May 1, 2026 by DhineshPonnarasan Contributor Loading…
1
1
Add opt-in nonuniform tensor parallelism community-request
#4585 opened May 1, 2026 by daiyaanarfeen Loading…
5 tasks done
1
Add argument for setting inference logits dtype Run functional tests
#4584 opened May 1, 2026 by santhnm2 Contributor • Draft
5 tasks
3
LanguageModelEmbedding: avoid kernel launch overhead for [b,s,h]→[s,b,h] in inference complexity: low Final Review PR is in the "final review" stage
#4583 opened May 1, 2026 by mathemakitten Contributor Loading…
5 tasks
2
Consolidate inference example scripts
#4581 opened May 1, 2026 by santhnm2 Contributor • Draft
5 tasks
1
Fix buffers in refit Run tests
#4580 opened May 1, 2026 by wdykas Contributor • Draft
5 tasks
4
Allow for separate mempools in CudaGraphManager complexity: low
#4579 opened May 1, 2026 by tdene Contributor • Draft
5 tasks
2
Named validation sets complexity: low
#4578 opened May 1, 2026 by RPrenger Contributor Loading…
5 tasks
4
ci: Test on ephemeral-dev
#4577 opened May 1, 2026 by chtruong814 Contributor • Draft
5 tasks
2
Fix Hang in tests Approved All necessary approvals have been made complexity: low Run tests
#4575 opened May 1, 2026 by wdykas Contributor Loading…
5 tasks
9
Update energon version requirement complexity: low
#4572 opened May 1, 2026 by maanug-nv Contributor Loading…
5 tasks
1
2
Reduce rollout broadcast memory usage and add safe optional torch.distributed import handling community-request waiting-on-customer Waiting on the original author to respond
#4571 opened May 1, 2026 by brukcodes Loading…
5 tasks
5
Enable shared expert overlap with allgatherv in inference complexity: low
#4570 opened May 1, 2026 by sidsingh-nvidia Contributor Loading…
5 tasks
5
Add the CSA/HCA prototype to HybridModel
#4569 opened May 1, 2026 by guihong-nv Contributor • Draft
5 tasks done
2
Add vLLM grouped gemm backend for MoE inference Approved All necessary approvals have been made complexity: high Run functional tests
#4566 opened Apr 30, 2026 by santhnm2 Contributor Loading…
5 tasks
19
Guard vocab reduce_scatter on TP > 1 Approved All necessary approvals have been made complexity: low
#4565 opened Apr 30, 2026 by mathemakitten Contributor Loading…
5 tasks
5
Context cpu async schedule
#4564 opened Apr 30, 2026 by lmcafee-nvidia Contributor • Draft
5 tasks
@lmcafee-nvidia
1
Update colwise data after param AG in eval
#4563 opened Apr 30, 2026 by WanZzzzzz Contributor Loading…
5 tasks
1
Update colwise data after synced param AG complexity: low Final Review PR is in the "final review" stage
#4562 opened Apr 30, 2026 by WanZzzzzz Contributor Loading…
1 of 5 tasks
2
ci: add cadence input for test filtering in CI workflows
#4561 opened Apr 30, 2026 by balasaajay Contributor • Draft
5 tasks
6
Add async scheduling for dynamic inference
#4558 opened Apr 30, 2026 by lmcafee-nvidia Contributor • Draft
1 of 5 tasks
1
Context cpu async schedule (CD-v3)
#4557 opened Apr 30, 2026 by lmcafee-nvidia Contributor • Draft
4 of 5 tasks
@lmcafee-nvidia
1
Make last_token_logits graphable Approved All necessary approvals have been made complexity: low
#4552 opened Apr 30, 2026 by tdene Contributor • Queued
5 tasks
18
ProTip! no:milestone will show everything without a milestone.