Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
199 Open 778 Closed
199 Open 778 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

feat: mcore trigger mbridge
#2340 opened Nov 20, 2025 by pablo-garay Loading…
6 tasks
Core 0.16
Flag to patch TE GEMMs with recalculate verification.
#2309 opened Nov 19, 2025 by kwyss-nvidia Loading…
6 tasks
2
Batch Invariance
#2308 opened Nov 19, 2025 by wdykas Loading…
6 tasks
1
Make check_large_grads non-fatal
#2307 opened Nov 19, 2025 by kwyss-nvidia Loading…
6 tasks
2
[training migration] add training config dataclass and arg generation utility Expert Review Apply this label to indicate that your PR is ready for expert review.
#2306 opened Nov 19, 2025 by maanug-nv Loading…
2 of 6 tasks
Core 0.16
2
1
hybrid-cp part 2 for main branch
#2304 opened Nov 19, 2025 by xiaoyao0115 Loading…
6 tasks
1
Draft: Megatron-FSDP and PP compatible support
#2302 opened Nov 19, 2025 by shjwudp Loading…
6 tasks
6
Improve asyncio exception handling Final Review Apply this label to indicate that your PR is ready for final review.
#2300 opened Nov 19, 2025 by tdene Loading…
6 tasks
Core 0.16
Handle UVM compile lock issues Expert Review Apply this label to indicate that your PR is ready for expert review.
#2299 opened Nov 19, 2025 by tdene Loading…
6 tasks
Core 0.16
1
Changes to support latent MoEs
#2296 opened Nov 19, 2025 by deepakn94 Loading… Core 0.16
@deepakn94
[Community][Main] feat(moe): Adding context parallel support to eager attention implementation
#2295 opened Nov 19, 2025 by yuzhongw-nvidia • Draft
6 tasks
1
Bugfix for Mamba with Chunked-Prefill
#2293 opened Nov 18, 2025 by sidsingh-nvidia Loading…
6 tasks
Core 0.16
2
add FIM dataset support Run tests
#2291 opened Nov 18, 2025 by dimapihtar Loading…
6 tasks
Core 0.16
39
build: Upgrade deps Expert Review Apply this label to indicate that your PR is ready for expert review. Run tests
#2289 opened Nov 18, 2025 by ko3n1g Loading…
6 tasks
Core 0.16
2
Fix TemporalAsyncCaller pin_memory lifetime in async checkpointing
#2288 opened Nov 18, 2025 by lvdunlin Loading…
6 tasks
2
[HOT FIX] Fix bug of hybrid-ep backend in flex-dispatcher bug Something isn't working Final Review Apply this label to indicate that your PR is ready for final review.
#2286 opened Nov 18, 2025 by Autumn1998 Loading…
6 tasks
Core 0.16
@Autumn1998
3
[Draft][Dev] fix(moe): minor refactor for fine-grained activation offloading dev branch Dev branch related issues and development
#2285 opened Nov 18, 2025 by lhb8125 • Draft
6 tasks
1
Adding support for running hybrid expert+data parallel inference via the inference coordinator.
#2283 opened Nov 18, 2025 by sidsingh-nvidia Loading…
6 tasks
Core 0.16
Hybrid Context Parallel Feature Expert Review Apply this label to indicate that your PR is ready for expert review.
#2282 opened Nov 17, 2025 by parthmannan Loading…
6 tasks
Core 0.16
3
Add support for fake distributed process groups. Expert Review Apply this label to indicate that your PR is ready for expert review.
#2280 opened Nov 17, 2025 by Victarry Loading…
10 of 16 tasks
Core 0.16
@Victarry
8
Automatically choose available ports in ZMQ
#2278 opened Nov 17, 2025 by tdene • Draft
6 tasks
2
Clean up DP coord code & unit test Expert Review Apply this label to indicate that your PR is ready for expert review. Run tests
#2277 opened Nov 17, 2025 by tdene Loading…
6 tasks
Core 0.16
2
Change ZMQ communication to use async ZMQ
#2276 opened Nov 17, 2025 by tdene • Draft
6 tasks
1
Remove dependency on megatron.training within megatron.core Final Review Apply this label to indicate that your PR is ready for final review. Run tests
#2274 opened Nov 17, 2025 by ananthsub Loading…
1 of 6 tasks
Core 0.16
3
Add assertion for mxfp8 params without dp overlap Expert Review Apply this label to indicate that your PR is ready for expert review.
#2271 opened Nov 17, 2025 by kunlunl Loading…
6 tasks
Core 0.16
1
ProTip! Type g p on any issue or pull request to go back to the pull request listing page.