Skip to content

Pull requests: NVIDIA/Megatron-LM

New pull request New
370 Open 2,286 Closed
370 Open 2,286 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] Use MSC for checking checkpoint existence community-request
#4251 opened Apr 10, 2026 by pavelgein Loading…
1 of 5 tasks
2
feat(moe): add DenseMixer dense forward pass for improved router grad… community-request
#4250 opened Apr 10, 2026 by meinie0826 Loading…
4 of 5 tasks
10
Fix RL to once again work with --skip-train complexity: low
#4249 opened Apr 10, 2026 by tdene Contributor Loading…
5 tasks
Core 0.16
1
Overload factor logging (squashed)
#4248 opened Apr 10, 2026 by nanz-nv Contributor • Draft
5 tasks
1
Paged Stashing
#4247 opened Apr 10, 2026 by nanz-nv Contributor • Draft
5 tasks
1
[Dev] Add runnable Qwen3.5 FSDP+EP in examples
#4245 opened Apr 10, 2026 by BestJuly Contributor • Draft
5 tasks
1
Add the YARN support for mamba_model
#4244 opened Apr 9, 2026 by guihong-nv • Draft
5 tasks
Core 0.16
2
Estimate speed-of-light decode latency complexity: medium
#4243 opened Apr 9, 2026 by santhnm2 Contributor Loading…
5 tasks
Core 0.16
2
Enable dbias_dprob triton kernel in TE GroupedLinear
#4242 opened Apr 9, 2026 by vasunvidia Contributor Loading…
5 tasks
3
RL: Offload optimizer via torch_memory_saver complexity: medium Final Review PR is in the "final review" stage
#4241 opened Apr 9, 2026 by tdene Contributor • Draft
5 tasks
Core 0.16
1
optimize attention init
#4240 opened Apr 9, 2026 by wdykas Contributor • Draft
5 tasks
1
fix: enforce correct pass thresholds for deterministic and approximate tests Run functional tests
#4238 opened Apr 9, 2026 by ko3n1g Contributor • Draft
2 tasks
Core 0.16
5
docs: add data loading best practices for large-scale training
#4236 opened Apr 9, 2026 by sbhavani Contributor • Draft
1 of 5 tasks
2
RL: Onload optimizer after logprobs computation complexity: low
#4235 opened Apr 9, 2026 by tdene Contributor Loading…
5 tasks
Core 0.16
3
Fix MoE checkpoint save skipping TP ranks when TP > EP * ETP (#4200) community-request
#4234 opened Apr 9, 2026 by aarushisingh04 Loading…
1 of 5 tasks
1
1
Graphed update_requests
#4233 opened Apr 9, 2026 by tdene Contributor • Draft
5 tasks
1
[Main][FSDP] TE Partial CUDA Graph + Megatron-FSDP compatibility
#4232 opened Apr 9, 2026 by buptzyb Contributor • Draft
5 tasks
1
[Dev][FSDP] TE Partial CUDA Graph + Megatron-FSDP compatibility
#4231 opened Apr 9, 2026 by buptzyb Contributor • Draft
5 tasks
1
[dev] fix(ssm): handle alignment padding in GDN packed seq + CP community-request
#4230 opened Apr 9, 2026 by yxs Loading… Core 0.16
13
feat(optimizer): add FlashAdamW optimizer integration community-request Final Review PR is in the "final review" stage
#4229 opened Apr 9, 2026 by meinie0826 Loading…
4 of 5 tasks
1
5
build: bump DeepEP to 34152ae Approved All necessary approvals have been made complexity: low core_r0.17.0 Auto-cherrypick to release branch. Apply before merge; cherrypick happens after merge.
#4228 opened Apr 9, 2026 by ko3n1g Contributor Loading…
1 task
Core 0.16
2
[training migration] Add container class for config dataclasses
#4227 opened Apr 9, 2026 by maanug-nv Contributor • Draft
5 tasks
1
Minor improvements for Dynamic-cp
#4226 opened Apr 9, 2026 by xiaoyao0115 Contributor Loading…
5 tasks
1
Extract args init to launch scripts complexity: low
#4225 opened Apr 9, 2026 by maanug-nv Contributor Loading…
5 tasks
Core 0.16
6
Clean up remaining HAVE_TE gating in GPT experimental specs and inference controller community-request
#4223 opened Apr 9, 2026 by returnL Contributor Loading…
1 of 5 tasks
2
ProTip! Updated in the last three days: updated:>2026-04-07.