Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
124 Open 436 Closed
124 Open 436 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 _set_wandb_writer serialization issues bug Something isn't working module: debugging
#1806 opened Sep 11, 2025 by gakkiri Loading…
5 of 8 tasks
1
Test workflow POC
#1804 opened Sep 10, 2025 by chtruong814 Loading…
5
Add files via upload
#1801 opened Sep 10, 2025 by wenchenqian Loading…
1
Quant
#1794 opened Sep 6, 2025 by Charles2530 Loading…
1
Update README.md module: documentation
#1792 opened Sep 4, 2025 by yuyu5333 Loading…
1
fix typo in SP/TP config validation error module: documentation
#1790 opened Sep 3, 2025 by clumsy Loading…
@sbhavani
1
Add support for packing-with-padding in gpt-dataset enhancement New feature or request module: data pipeline
#1788 opened Sep 3, 2025 by terminator123 Loading…
1
Add falcon h1 2 enhancement New feature or request
#1785 opened Sep 2, 2025 by dhiaEddineRhaiem Loading…
1
bugfix: raise error if eos_token is not set in tokenizer bug Something isn't working module: data pipeline
#1774 opened Aug 27, 2025 by imomayiz Loading…
1
Fix torch_dist checkpointing ETP replica_id bug Something isn't working module: moe
#1770 opened Aug 25, 2025 by Skylion007 Loading…
1
[DRAFT] Emerging optimizer development
#1768 opened Aug 22, 2025 by skyw • Draft
14
Fix Context Parallel NaN Loss bug Something isn't working
#1765 opened Aug 21, 2025 by leoleoasd Loading…
1
1
Fix runaway Etpt in straggler detector by resetting FLOPs accumulator bug Something isn't working
#1755 opened Aug 19, 2025 by cms42 Loading…
4
perf(MoE): Use TE quant/dequant for SwiGLU fp8 input store to improve performance and stability
#1753 opened Aug 19, 2025 by xiaoxi-wangfj Loading…
1
[main][feature][under updating]zero-overhead activation offload enhancement New feature or request
#1752 opened Aug 18, 2025 by GeYuhong Loading…
26
fix: Initialize master_weight with params_dtype directly bug Something isn't working
#1748 opened Aug 15, 2025 by Mirza-Samad-Ahmed-Baig Loading…
fix loading dcp OOM bug Something isn't working
#1747 opened Aug 14, 2025 by zjjott Loading…
2
fix several typos in megatron/core/transformer/multi_token_prediction.py module: documentation
#1744 opened Aug 13, 2025 by andy-yangz Loading…
1
Hongbinl/1f1b overlap mirror 0813
#1743 opened Aug 13, 2025 by lhb8125 Loading…
ci: Add build-test-publish wheel workflow
#1742 opened Aug 13, 2025 by ko3n1g Loading…
2
Add world_size dict getter method for simple integration with W&B enhancement New feature or request
#1735 opened Aug 9, 2025 by WoosungMyung Loading…
2
export _move_new_state_to_right_device for offload/load enhancement New feature or request
#1734 opened Aug 8, 2025 by techkang Loading…
Megatron-LM changes to make Hyena/Evo 2 inference usable, especially for 40B models enhancement New feature or request
#1727 opened Aug 1, 2025 by antonvnv Loading…
fix router input jitter dtype bug Something isn't working
#1726 opened Aug 1, 2025 by chaitanyadwivedi96 Loading…
1
Add FP8 training scripts enhancement New feature or request module: transformer engine
#1723 opened Jul 31, 2025 by SDcodehub • Draft
ProTip! Add no:assignee to see everything that’s not assigned.