Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
384 Open 2,539 Closed
384 Open 2,539 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 release tests: remove --global-batch-size conflicting with --step-batch-size-schedule complexity: low
#4545 opened Apr 30, 2026 by deepakn94 Contributor Loading…
1 task
1
Fixes for Nemotron3 Super release test config
#4544 opened Apr 29, 2026 by maanug-nv Contributor • Draft
5 tasks
1
revert: replace rampup batch size scheduler with custom step batch size schedules (#4411) complexity: high
#4543 opened Apr 29, 2026 by ko3n1g Contributor Loading…
3
docs: use @file-path notation for file references in skills docs-only documentation only (docs or docstrings)
#4542 opened Apr 29, 2026 by ko3n1g Contributor Loading…
1
Avoid nsys profile crash with CUDA graphs
#4541 opened Apr 29, 2026 by tdene Contributor • Draft
5 tasks
1
ci: add base_sha to codecov/codecov-action upload step complexity: low
#4540 opened Apr 29, 2026 by chtruong814 Contributor • Queued
5 tasks
3
Add --model-recipe CLI flag for HybridModel
#4539 opened Apr 29, 2026 by Phlip79 Member • Draft
3 tasks
1
Nemotron-3 Nano recipe with HybridModel DSL
#4538 opened Apr 29, 2026 by Phlip79 Member • Draft
2 tasks
2
Python DSL recipe API for HybridModel
#4537 opened Apr 29, 2026 by Phlip79 Member • Draft
3 tasks
6
Skip cu_seqlens broadcast when packed sequences are not in use complexity: low
#4536 opened Apr 29, 2026 by tomlifu Contributor Loading…
5 tasks
@tomlifu
2
Async schedule GPU/CPU overlap
#4535 opened Apr 29, 2026 by lmcafee-nvidia Contributor • Draft
@lmcafee-nvidia
1
mmiranda working on another set of broken links
#4534 opened Apr 29, 2026 by megnvidia Contributor Loading…
5 tasks
7
Move policy epoch stats to the message object complexity: low Final Review PR is in the "final review" stage
#4533 opened Apr 29, 2026 by ArEsKay3 Contributor Loading…
5 tasks
2
ci: Update CI workflow conditions to include merge group handling
#4532 opened Apr 29, 2026 by balasaajay Contributor • Draft
5 tasks
3
[Split 1/N of #3430] feat(mHC): basic pytorch implementation of manifold hyper connection complexity: high Run functional tests
#4531 opened Apr 29, 2026 by Connor-XY Loading…
3 of 4 tasks
9
fix: update fine-grained offload assertion for full-iteration cuda gr…
#4525 opened Apr 29, 2026 by rapatel Contributor Loading…
5 tasks
1
bump emerging-optimizer package for deepseek v4 coefficients type complexity: medium
#4523 opened Apr 29, 2026 by FDecaYed Contributor Loading…
5 tasks
2
additional tests for nvrx complexity: low Expert Review [deprecated] Apply this label to indicate that your PR is ready for expert review.
#4522 opened Apr 29, 2026 by dimapihtar Contributor Loading…
5 tasks
4
Allow optimizer CG to share the same pool as full-iter CG
#4521 opened Apr 29, 2026 by nanz-nv Contributor Loading…
5 tasks
2
ModelOpt QAD test for CICD
#4520 opened Apr 29, 2026 by AAnoosheh Contributor • Draft
1 of 5 tasks
@AAnoosheh
1
[feat] enable cuda graph for the language_model sub module in VLM models
#4519 opened Apr 29, 2026 by shifangx Contributor • Draft
5 tasks
2
[dev] [DeepSeek-v4] Part 3: MTP support with mHC and new mHC contract dev branch Dev branch related issues and development
#4518 opened Apr 29, 2026 by hxbai Contributor • Draft
5 tasks
@hxbai
1
Add DSV3 mxfp8 proj/logits gemm support
#4516 opened Apr 29, 2026 by mmarcinkiewicz • Draft
5 tasks
2
Add a knob to throttle the max allowed inflight offload in fine grained offloading
#4514 opened Apr 29, 2026 by nanz-nv Contributor Loading…
5 tasks
3
fix: update fine-grained offload assertion for full-iteration cuda gr…
#4513 opened Apr 29, 2026 by rapatel Contributor Loading…
5 tasks
1
ProTip! Find all pull requests that aren't related to any open issues with -linked:issue.