Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
328 Open 2,155 Closed
328 Open 2,155 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: wait for async P2P send before deallocating output tensor
#4047 opened Mar 28, 2026 by ZhiyuLi-Nvidia • Draft
5 tasks
2
Fix HybridDeviceOptimizer KeyError after mixed-precision param replacement community-request Final Review PR is in the "final review" stage
#4046 opened Mar 28, 2026 by ma-ben Loading…
1
2
Fix edge case where prefix caching could create prefill chunk of length 1
#4044 opened Mar 27, 2026 by santhnm2 • Draft
5 tasks
1
Docs: improve docstrings and comments in example training loop community-request
#4041 opened Mar 27, 2026 by DhineshPonnarasan • Draft
1
1
[Kernel] Fused Indexer Loss Kernel community-request
#4039 opened Mar 27, 2026 by laixinn Loading…
5 tasks
1
Fix unnecessary permute padding for non-quantized MoE dispatch community-request needs-follow-up Issue needs follow-up
#4038 opened Mar 27, 2026 by xiaoxi-wangfj Loading…
5 tasks
2
Add stacklevel to warnings.warn() calls across megatron/core community-request
#4036 opened Mar 26, 2026 by Bias92 Loading…
3 of 5 tasks
2
docs: add AGENT.md with CI/CD and failure navigation guidance docs-only documentation only (docs or docstrings)
#4035 opened Mar 26, 2026 by ko3n1g Loading…
3 tasks
Core 0.16
2
Implement scatter_gather_tensors_in_pipeline logic community-request
#4032 opened Mar 26, 2026 by QingZhou-YangHY Loading… Core 0.16
1
4
Sync main into dev
#4031 opened Mar 25, 2026 by ilml Loading…
7 tasks done
Core 0.16
5
Miscellaneous inference fixes complexity: medium
#4030 opened Mar 25, 2026 by santhnm2 Loading…
5 tasks
Core 0.16
4
Megatron-FSDP: Make _pre_forward_param_unshard and _register_post_backward_hook formal complexity: low module: megatron-fsdp
#4029 opened Mar 25, 2026 by shjwudp • Draft
5 tasks
@shjwudp
2
add support for DCP and FSDP async save
#4027 opened Mar 25, 2026 by dimapihtar • Draft
5 tasks
2
fix fine_grained_callables with fused rmsnorm residual complexity: low dev2main: mbridge dev to main: this PR is needed in main for mbridge Final Review PR is in the "final review" stage
#4026 opened Mar 25, 2026 by CarlosGomes98 Loading…
5 tasks
Core 0.16
@CarlosGomes98
3
m-fsdp: wire use_precision_aware_optimizer from ddp_config to ParamAn… complexity: low Final Review PR is in the "final review" stage module: megatron-fsdp
#4024 opened Mar 25, 2026 by rapatel Loading…
5 tasks
Core 0.16
@rapatel
10
Gk/1node fsdp conv
#4023 opened Mar 25, 2026 by gautham-kollu • Draft
5 tasks
1
[DO NOT MERGE] Combined MiMo non-colocated changes for MBridge integration
#4022 opened Mar 24, 2026 by yashaswikarnati • Draft
1
Allow the evaluation batch size to differ from the training batch size
#4014 opened Mar 24, 2026 by michal2409 Loading…
4 of 5 tasks
2
Fix spike-no-more embedding weight decay regression (#3945) community-request
#4012 opened Mar 24, 2026 by DhineshPonnarasan • Draft
2
Rename moe_router_force_biased to moe_router_force_biased_std and only apply if non-zero
#4011 opened Mar 24, 2026 by denys-fridman Loading…
5 tasks
2
fix: improve text generation CLI validation and clean up duplicate imports community-request Final Review PR is in the "final review" stage needs-follow-up Issue needs follow-up
#4010 opened Mar 24, 2026 by nathon-lee Loading…
3
feat(energy-monitor): Add per-GPU power CSV logging via --log-energy-csv-dir
#4008 opened Mar 23, 2026 by sriduttb • Draft
3 of 5 tasks
2
Adding perf knobs complexity: low Run functional tests
#4007 opened Mar 23, 2026 by gautham-kollu Loading…
5 tasks
Core 0.16
5
NVFP4 native weights for DDP 26.04 this PR is high priority and should be merged asap community-request complexity: high Final Review PR is in the "final review" stage needs-follow-up Issue needs follow-up
#4005 opened Mar 23, 2026 by WanZzzzzz Loading…
5 tasks done
Core 0.16
13
debug: Unit tests for merge-main-into-dev
#4004 opened Mar 23, 2026 by chtruong814 Loading…
5 tasks
Core 0.16
6
ProTip! Type g i on any issue or pull request to go back to the issue listing page.