Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
163 Open 609 Closed
163 Open 609 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

fp8 param cuda graph support main
#2088 opened Nov 2, 2025 by kunlunl Loading…
6 tasks
1
[dev] fp8 param cuda graph support
#2087 opened Nov 2, 2025 by kunlunl Loading…
6 tasks
1
[Dev] [Draft] FP8 params support for megatron-fsdp
#2086 opened Nov 2, 2025 by kunlunl Loading…
6 tasks
1
Fix ambiguous tensor truth-value check in train_rl.loss_func (use .it…
#2085 opened Nov 2, 2025 by vignesh1507 Loading…
1
Fix NameError in pretrain_retro.py (add import_module), remove unused…
#2084 opened Nov 2, 2025 by vignesh1507 Loading…
1
Prevent summary jobs from running in forks
#2083 opened Nov 1, 2025 by tdene Loading…
6 tasks
Core 0.16
ci(hotfix): Bypass approvalbot in merge-queue
#2082 opened Nov 1, 2025 by ko3n1g Loading…
6 tasks
Core 0.16
Records time to first token in the dynamic inference coordinator
#2081 opened Nov 1, 2025 by sidsingh-nvidia • Draft
6 tasks
2
Add minor log update
#2080 opened Nov 1, 2025 by gautham-kollu Loading…
6 tasks
Core 0.16
2
(release dev branch) Nemotron nano v2 vl
#2079 opened Oct 31, 2025 by cuichenx Loading…
6 tasks
Nemotron nano v2 vl changes for Megatron Bridge
#2078 opened Oct 31, 2025 by cuichenx Loading…
6 tasks
Core 0.16
Add extra RL files Final Review Apply this label to indicate that your PR is ready for final review.
#2077 opened Oct 31, 2025 by tdene • Queued
6 tasks
Core 0.16
7
(Cherry pick) fix yarn attn factor r0.15.0
#2076 opened Oct 31, 2025 by cuichenx Loading…
6 tasks
3
Remove MoE throughput test
#2075 opened Oct 31, 2025 by mathemakitten Loading…
1 of 6 tasks
Core 0.16
Add BytesIO to safe_globals Expert Review Apply this label to indicate that your PR is ready for expert review.
#2074 opened Oct 31, 2025 by tdene Loading…
6 tasks
Core 0.16
1
Inference throughput tests: refactor goldens to be in list format
#2072 opened Oct 31, 2025 by mathemakitten Loading…
1 of 6 tasks
Core 0.16
1
Move data_samplers.py from legacy to core.datasets & add DistributedSignalHandler to DataLoader workers
#2068 opened Oct 31, 2025 by asolergi-nv Loading…
3 of 6 tasks
Core 0.16
2
chore: Update local tooling
#2066 opened Oct 31, 2025 by ko3n1g • Queued
6 tasks
Core 0.16
Fix Qwen3-Next hang on Blackwell, add a flag to control torch.compile
#2058 opened Oct 31, 2025 by yuzhongw-nvidia Loading…
6 tasks
Core 0.16
4
Make PipelineParallelLayout always return str from __repr__ Final Review Apply this label to indicate that your PR is ready for final review.
#2055 opened Oct 30, 2025 by ananthsub Loading…
3 of 6 tasks
Core 0.16
4
Hybrid Data x Context Parallelism Feature enhancement New feature or request Expert Review Apply this label to indicate that your PR is ready for expert review.
#2054 opened Oct 30, 2025 by parthmannan Loading…
6 tasks
Core 0.16
1
M4 + Dist Checkpoint: Replace global parallel state with explicit group parameters
#2053 opened Oct 30, 2025 by dimapihtar • Draft
6 tasks
1
ci: POC: Gridify test cases
#2049 opened Oct 30, 2025 by ko3n1g Loading…
6 tasks
Core 0.16
Siddharth/torch symm fused
#2048 opened Oct 30, 2025 by sidsingh-nvidia • Draft
6 tasks
Core 0.16
@sidsingh-nvidia
1
Optimizer refactor
#2047 opened Oct 30, 2025 by deepakn94 Loading… Core 0.16
1
ProTip! Find all pull requests that aren't related to any open issues with -linked:issue.