Skip to content

Pull requests: NVIDIA-NeMo/RL

New pull request New
337 Open 1,994 Closed
337 Open 1,994 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

build(sglang): bump sglang to 0.5.13.post1
#3344 opened Jul 25, 2026 by Kh4L • Draft
1
chore: Bump Gym
#3343 opened Jul 25, 2026 by yfw Contributor Loading…
4 tasks
2
fix: reserve MOPD teacher clusters before Gym startup CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#3340 opened Jul 24, 2026 by yfw Contributor Loading…
4 tasks
3
lora 2 community-request Documentation Improvements or additions to documentation
#3338 opened Jul 24, 2026 by xrsrke Loading…
4 tasks
19
ci: Bump Megatron-Bridge to 8645cda CI:L1 Run doctests, unit tests, and functional tests
#3335 opened Jul 24, 2026 by svcnvidia-nemo-ci Contributor Loading…
3
[DO NOT MERGE] NCCL grow API demo
#3334 opened Jul 24, 2026 by youngeunkwon0405 Contributor Loading…
4 tasks
1
feat: standardize mcore generation config defaults and expose cuda_gr… CI:L1 Run doctests, unit tests, and functional tests
#3331 opened Jul 23, 2026 by shanmugamr1992 Contributor Loading…
4 tasks
2
feat(sglang): add end-to-end training and refit integration Documentation Improvements or additions to documentation
#3330 opened Jul 23, 2026 by Kh4L • Draft
4 of 6 tasks
3
feat(tq): train prefetch microbatches nccl
#3328 opened Jul 23, 2026 by zyzhou5 Contributor Loading…
4 tasks
1
Multimodal nemo gym compatible grpo pipeline
#3325 opened Jul 23, 2026 by rohitrango Contributor Loading…
2 of 4 tasks
1
fix: harden multimodal Gym training lifecycle
#3324 opened Jul 23, 2026 by JeffPengCoder Loading…
6 tasks done
6
fix(eagle3): isolate draft grad clipping
#3323 opened Jul 23, 2026 by yuekaizhang Contributor Loading…
2
fix: Megatron async checkpoint CUDA memory retention CI:L1 Run doctests, unit tests, and functional tests
#3322 opened Jul 23, 2026 by NolenLiang Contributor Loading…
3 tasks done
1
2
fix(ppo): enable DTensor value dynamic batching with equivalence coverage CI:L1 Run doctests, unit tests, and functional tests CI Relating to CI
#3321 opened Jul 23, 2026 by fujial-code Contributor Loading…
1
2
feat: support sequence packing and context parallelism for DTensor value models CI:L1 Run doctests, unit tests, and functional tests CI Relating to CI
#3320 opened Jul 23, 2026 by fujial-code Contributor Loading…
1
2
perf(loss): avoid full-vocab student log-softmax in same-vocab top-k KD community-request waiting-on-maintainers Waiting on maintainers to respond
#3314 opened Jul 22, 2026 by tianyi-zhang-02 Contributor Loading…
3 tasks done
1
1
feat: fault tolerant generation Documentation Improvements or additions to documentation
#3313 opened Jul 22, 2026 by ashors1 Contributor • Draft
4 tasks
1
Standalone Critic Pretraining (stepwise) Documentation Improvements or additions to documentation
#3310 opened Jul 21, 2026 by jkyi-nvidia Loading…
1 of 4 tasks
1
feat(flops): add NemotronH MoE-aware FLOPs tracking for MFU community-request waiting-on-maintainers Waiting on maintainers to respond
#3309 opened Jul 21, 2026 by dafu-wu Loading…
4 tasks done
1
perf: avoid CPU round trip during NVFP4 refit CI:L1 Run doctests, unit tests, and functional tests
#3308 opened Jul 21, 2026 by seonjinn Contributor Loading…
7
feat: implement ASFT community-request waiting-on-maintainers Waiting on maintainers to respond
#3302 opened Jul 21, 2026 by pengdurice Contributor Loading…
3 of 4 tasks
1
perf(grpo): make non-colocated refit offload optional
#3300 opened Jul 21, 2026 by wdykas Loading…
4 tasks
1
feat(vllm): refit-idempotent BF16 FlashInfer-TRTLLM MoE (vLLM 0.20.x stopgap)
#3296 opened Jul 20, 2026 by seonjinn Contributor Loading…
2
feat(grpo): reuse generation logprobs as IS prev-logprobs CI:L2 Run doctests, unit tests, functional tests, and convergence tests
#3295 opened Jul 20, 2026 by seonjinn Contributor Loading…
@seonjinn
16
perf: Reduce MXFP8 refit weight-transfer overheads CI:L1 Run doctests, unit tests, and functional tests
#3294 opened Jul 20, 2026 by seonjinn Contributor Loading…
@seonjinn
10
ProTip! Mix and match filters to narrow down what you’re looking for.