-
Notifications
You must be signed in to change notification settings - Fork 522
Pull requests: NVIDIA-NeMo/RL
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
feat(refit): add a ModelExpress-brokered nccl_reshard transport
Documentation
Improvements or additions to documentation
#3720
opened Aug 19, 2026 by
yixinh-nv
Loading…
3 of 4 tasks
fix: Megatron Inference colocated CUDA graphs
#3718
opened Aug 19, 2026 by
tdene
Contributor
Loading…
4 tasks
feat(draft): add projected soft CE for DFlash
#3715
opened Aug 19, 2026 by
seonjinn
Contributor
Loading…
fix: add back NRL_NCCL_TIMEOUT_MINUTES to avoid megatron nccl timeout
#3714
opened Aug 19, 2026 by
hchings
Contributor
Loading…
4 tasks
feat(spec-decode): DSpark block-draft co-training (Megatron)
community-request
#3713
opened Aug 19, 2026 by
MarkXCloud
•
Draft
1 of 4 tasks
feat(spec-decode): DFlash block-draft co-training (Megatron)
community-request
#3712
opened Aug 19, 2026 by
MarkXCloud
•
Draft
1 of 4 tasks
feat(draft): add DSpark auxiliary heads
CI:megatron
#3711
opened Aug 19, 2026 by
seonjinn
Contributor
Loading…
fix: non-contiguous top-p logits, keep_top_k null, and fp32-buffer refit fidelity
CI:L1
Run doctests, unit tests, and functional tests
#3709
opened Aug 19, 2026 by
mferrato
Loading…
4 tasks done
feat(draft): support optimizer schedule overrides
#3707
opened Aug 19, 2026 by
seonjinn
Contributor
Loading…
feat(draft): preserve split-step normalization parity
#3706
opened Aug 19, 2026 by
seonjinn
Contributor
Loading…
feat(sc): log setup milestones
CI:L0
Run doctests and unit tests
#3705
opened Aug 18, 2026 by
youngeunkwon0405
Contributor
•
Draft
feat(refit): integrate ModelExpress weight synchronization
community-request
#3704
opened Aug 18, 2026 by
zhengluo-nv
Loading…
feat(draft): stream vocab-parallel soft CE
CI:L2
Run doctests, unit tests, functional tests, and convergence tests
#3703
opened Aug 18, 2026 by
seonjinn
Contributor
Loading…
fix(nemo-gym): honour pad_dynamic_image_shapes on the dedup prompt path
bug
Something isn't working
CI:L0
Run doctests and unit tests
environments
multimodal
#3702
opened Aug 18, 2026 by
DanialTaheri
Contributor
Loading…
refactor(draft): add typed speculator contract
CI:L0
Run doctests and unit tests
#3701
opened Aug 18, 2026 by
seonjinn
Contributor
Loading…
fix: Complete truncated resources limits in dev pod manifest (closes #2816)
community-request
#3700
opened Aug 18, 2026 by
botbikamordehai2-sketch
Loading…
feat(spec-decode): multi-pass TTT training for EAGLE-3 drafts (Megatron)
community-request
#3699
opened Aug 18, 2026 by
MarkXCloud
•
Draft
1 of 4 tasks
fix(megatron): free reference model GPU memory before policy buffer reloaded
community-request
#3697
opened Aug 18, 2026 by
philip-paul-mueller
•
Draft
3 of 4 tasks
feat: add checkpointing.resume_if_exists to control automatic resumption
community-request
#3692
opened Aug 18, 2026 by
VishnuR23
Contributor
Loading…
4 tasks done
fix: propagate policy CLI overrides to the distillation teacher config
community-request
#3691
opened Aug 18, 2026 by
VishnuR23
Contributor
Loading…
4 tasks done
fix: raise an actionable error when a prepare_for_* step is skipped
community-request
#3690
opened Aug 18, 2026 by
VishnuR23
Contributor
Loading…
4 tasks done
fix(megatron): create CUDA graph managers for colocated generation
community-request
#3689
opened Aug 17, 2026 by
YigongQin
Loading…
6 tasks done
feat(megatron): router replay for colocated Megatron inference
community-request
Documentation
Improvements or additions to documentation
#3687
opened Aug 17, 2026 by
YigongQin
Loading…
4 tasks done
feat(grpo): port grouped validation (pass_k) to validate_sync
community-request
waiting-on-maintainers
Waiting on maintainers to respond
#3686
opened Aug 17, 2026 by
lonexreb
Contributor
Loading…
1 of 2 tasks
Previous Next
ProTip!
Exclude everything labeled
bug with -label:bug.