-
Notifications
You must be signed in to change notification settings - Fork 3.9k
Pull requests: NVIDIA/Megatron-LM
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
chore: nightly sync main into dev (08_05_2026)
Run functional tests
Run MBridge tests
Attach this for testing this PR against MBridge main
#4708
opened May 8, 2026 by
svcnvidia-nemo-ci
•
Draft
chore: Update golden values for various functional tests
complexity: medium
Run functional tests
#4706
opened May 8, 2026 by
balasaajay
Contributor
•
Queued
5 tasks
Fix issue where parameter groups with different min/max LRs get overridden at checkpoint load time
complexity: low
#4705
opened May 8, 2026 by
jstjohn
Contributor
Loading…
5 tasks
chore(codeowners): add megatron/inference/ ownership
complexity: low
#4704
opened May 8, 2026 by
ko3n1g
Contributor
Loading…
Simple and stable Inference APIs
complexity: high
#4697
opened May 8, 2026 by
YangFei1990
Contributor
Loading…
4 of 5 tasks
ci: Remove unnecessary taint node job
Approved
All necessary approvals have been made
complexity: low
#4696
opened May 8, 2026 by
chtruong814
Contributor
Loading…
5 tasks
Add a knob to throttle the max allowed inflight offload in fine grained offloading
complexity: low
#4692
opened May 8, 2026 by
nanz-nv
Contributor
Loading…
5 tasks
Fix unit tests
complexity: medium
#4689
opened May 8, 2026 by
shanmugamr1992
Contributor
Loading…
5 tasks
Fix weights/opt memory estimation
complexity: medium
#4687
opened May 7, 2026 by
YangFei1990
Contributor
Loading…
5 tasks
feat(gpt): add output postprocess hook
Final Review
PR is in the "final review" stage
#4686
opened May 7, 2026 by
Glitchfix
Loading…
3 of 5 tasks
Guard omegaconf imports
complexity: low
#4685
opened May 7, 2026 by
maanug-nv
Contributor
Loading…
5 tasks
fix: updated paged stashing hook on_save_for_backward().
#4683
opened May 7, 2026 by
rapatel
Contributor
Loading…
5 tasks
Update transformer-engine dependency to version 2.15.0
Run functional tests
#4682
opened May 7, 2026 by
balasaajay
Contributor
•
Draft
5 tasks
fix legacy torch save when tensor_model_parallel_size > expert_model_parallel_size * expert_tensor_parallel_size
complexity: low
#4678
opened May 7, 2026 by
dimapihtar
Contributor
Loading…
5 tasks
Allow selecting top-k used for group scoring
complexity: low
#4667
opened May 7, 2026 by
janEbert
Contributor
Loading…
fix(fsdp): recognize legacy GDN TP metadata
Final Review
PR is in the "final review" stage
module: megatron-fsdp
#4664
opened May 7, 2026 by
Glitchfix
Loading…
3 of 5 tasks
[Megatron-FSDP] Add conditional param.grad dereferencing logic to support CUDA graphability.
complexity: low
Expert Review
[deprecated] Apply this label to indicate that your PR is ready for expert review.
module: megatron-fsdp
#4663
opened May 6, 2026 by
cspades
Member
Loading…
5 tasks
chore: nightly sync main into dev (06_05_2026)
complexity: high
Run functional tests
Run MBridge tests
Attach this for testing this PR against MBridge main
#4659
opened May 6, 2026 by
svcnvidia-nemo-ci
Loading…
Previous Next
ProTip!
Find all pull requests that aren't related to any open issues with -linked:issue.