Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
336 Open 2,223 Closed
336 Open 2,223 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

cp: Enable CUDA graph for ADAM optimizer (3429) into core_r0.17.0 cherry-pick Run CICD
#4142 opened Apr 3, 2026 by ko3n1g Loading… Core 0.16
2
Claude: add respond-to-issue skill docs-only documentation only (docs or docstrings)
#4141 opened Apr 3, 2026 by Phlip79 Loading…
2 tasks done
Core 0.16
2
fix: Remove fail-fast (-x) and guard distributed teardown against deadlock complexity: low
#4139 opened Apr 3, 2026 by ko3n1g Loading… Core 0.16
3
Fix layerwise expt dp size 1 and contention with element wise distributed optimizer complexity: low
#4138 opened Apr 3, 2026 by skyw Loading…
5 tasks
Core 0.16
7
Add conditions_embeddings argument to TransformerBlock, TransformerLayer for DiT (diffusion transformer) complexity: low Final Review PR is in the "final review" stage
#4134 opened Apr 3, 2026 by huvunvidia Loading…
5 tasks
Core 0.16
3
Fix incorrectly set decoupled_grad in training.py for MFSDP. complexity: low module: megatron-fsdp
#4133 opened Apr 3, 2026 by cspades Loading…
5 tasks
Core 0.16
@cspades
1
Autotune inference
#4131 opened Apr 3, 2026 by tdene • Draft
5 tasks
1
Return usage characteristics from the completions API
#4128 opened Apr 3, 2026 by sidsingh-nvidia • Draft
5 tasks
Core 0.16
2
Autotune RL lag complexity: low
#4127 opened Apr 3, 2026 by tdene Loading…
5 tasks
Core 0.16
1
RL: prevent unnecessary entry into inference mode complexity: low
#4125 opened Apr 3, 2026 by tdene Loading…
5 tasks
Core 0.16
2
[Fix] Remove manual dgrad release
#4122 opened Apr 3, 2026 by Wohox • Draft
5 tasks
1
fix the strategy in tensor_need_offloading_checker
#4121 opened Apr 3, 2026 by lhb8125 Loading…
5 tasks
1
Clip mtp grads separately when mtp_detach_heads=True
#4116 opened Apr 2, 2026 by yfw • Draft
5 tasks
2
mmiranda second batch of UG AI audit edits
#4115 opened Apr 2, 2026 by megnvidia • Draft
5 tasks
2
Add unit tests for dynamic_text_gen_server
#4112 opened Apr 2, 2026 by tdene • Draft
5 tasks
1
Add conditions_embeddings argument to TransformerBlock, TransformerLayer for DiT (diffusion transformer)
#4111 opened Apr 2, 2026 by huvunvidia Loading…
5 tasks
Core 0.16
4
Nanz/overload factor logging
#4110 opened Apr 2, 2026 by nanz-nv • Draft
5 tasks
21
fix security issue: remove weights_only=False for multimodal example complexity: low
#4104 opened Apr 1, 2026 by faradawn Loading… Core 0.16
3
Inputs refactor Run functional tests
#4103 opened Apr 1, 2026 by asolergi-nv • Draft
5 tasks
Core 0.16
4
Make MTP / prefix cache stats persist for engine lifetime
#4101 opened Apr 1, 2026 by santhnm2 • Draft
5 tasks
1
Rename MambaModel/MambaStack to HybridModel/HybridStack complexity: high
#4099 opened Apr 1, 2026 by Phlip79 Loading… Core 0.16
38
mmiranda starting AI audit of user guide
#4098 opened Apr 1, 2026 by megnvidia Loading…
5 tasks
2
Add tables and histogram for RL staleness complexity: low
#4097 opened Apr 1, 2026 by tdene Loading…
5 tasks
Core 0.16
2
Fix RL reward due to stop token complexity: low
#4096 opened Apr 1, 2026 by tdene Loading…
5 tasks
Core 0.16
2
TE fused grouped mlp with grouped bias and delayed wgrad
#4095 opened Apr 1, 2026 by ksivaman Loading…
5 tasks
Core 0.16
3
ProTip! Exclude everything labeled bug with -label:bug.