Skip to content

Pull requests: NVIDIA-NeMo/RL

New pull request New
498 Open 2,230 Closed
498 Open 2,230 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: complete dev pod resource limits dict (closes #2816) community-request
#3796 opened Aug 24, 2026 by botbikamordehai2-sketch Loading…
1
1
ci: Bump Megatron-Bridge to 846216e CI:L1 Run doctests, unit tests, and functional tests
#3795 opened Aug 24, 2026 by svcnvidia-nemo-ci Contributor Loading…
3
fix(grpo): stop corrupting total_reward across dynamic-sampling rounds community-request
#3794 opened Aug 24, 2026 by khazic Loading…
1
fix(grpo): pass rewards into the GRPO train batch for VAPO's NLL term community-request
#3792 opened Aug 24, 2026 by khazic Loading…
1
fix(loss): normalize the VAPO positive-example NLL by the global token count Documentation Improvements or additions to documentation
#3791 opened Aug 24, 2026 by khazic Loading…
1
fix(megatron): add fa4, deepgemm, and dependency community-request
#3790 opened Aug 24, 2026 by YigongQin Loading…
2 of 4 tasks
1
fix(megatron): install finalizer on runtime model config
#3789 opened Aug 24, 2026 by rohitrango Contributor Loading…
4 tasks done
2
fix(dtensor): count TP-replicated gradients once in get_grad_norm community-request
#3788 opened Aug 24, 2026 by khazic Loading…
1
fix(sc): apply advantage clipping and reward scaling in the advantage stage community-request
#3787 opened Aug 24, 2026 by tianyi-zhang-02 Contributor Loading…
3 of 4 tasks
1
fix(sc): build sample_mask from the completion instead of all ones community-request
#3786 opened Aug 24, 2026 by tianyi-zhang-02 Contributor Loading…
3 of 4 tasks
2
fix(gym): make the gym venv prefetch dry run pass again CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#3785 opened Aug 24, 2026 by yfw Contributor Loading…
4 tasks
11
docs(sc): checkpointing is supported; only validation is not community-request Documentation Improvements or additions to documentation
#3784 opened Aug 24, 2026 by tianyi-zhang-02 Contributor Loading…
4 tasks done
1
fix(sc): keep the watchdog armed while the train pump drains community-request
#3783 opened Aug 24, 2026 by tianyi-zhang-02 Contributor Loading…
3 of 4 tasks
1
fix: Warn on expandable segments at IPC sender community-request
#3782 opened Aug 24, 2026 by benthecarman Loading…
4 tasks done
2
fix: Hoist val metrics dict in validate() community-request
#3781 opened Aug 24, 2026 by benthecarman Loading…
3 of 4 tasks
2
docs: Update performance numbers for v0.7 release CI:docs Run doctest Documentation Improvements or additions to documentation
#3779 opened Aug 24, 2026 by parthmannan Contributor Loading…
4 tasks
2
fix(loss): narrow the prev-side reductions when prev_logprobs is curr_logprobs community-request
#3778 opened Aug 23, 2026 by tianyi-zhang-02 Contributor Loading…
3 of 4 tasks
1
fix(vllm): patch flashinfer_trtllm moe backend to support refit
#3774 opened Aug 23, 2026 by guyueh1 Contributor Loading…
1
feat(sc): support PPO in single controller CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version) Documentation Improvements or additions to documentation
#3773 opened Aug 23, 2026 by yuki-97 Contributor • Draft
4 tasks done
39
feat(gym): apply effort-level reward shaping in the SingleController CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#3770 opened Aug 23, 2026 by RayenTian Contributor Loading…
4 tasks
5
feat(async-grpo): add length reward shaping CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version) Documentation Improvements or additions to documentation
#3769 opened Aug 22, 2026 by ehosseiniasl Contributor Loading…
@ehosseiniasl
4
feat: add MOPD to single-controller text path CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version) Documentation Improvements or additions to documentation
#3768 opened Aug 22, 2026 by yfw Contributor Loading…
4 tasks
26
fix(grpo): support environment sample masking in single controller
#3766 opened Aug 22, 2026 by macandro96 Contributor Loading…
4 tasks
1
feat: log toolcall and thinktag violation rate into single controller CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#3765 opened Aug 21, 2026 by arnavk-nvidia Contributor Loading…
2 of 4 tasks
1
refactor(dynamo): move native token response handling to Gym CI:L1 Run doctests, unit tests, and functional tests Documentation Improvements or additions to documentation
#3763 opened Aug 21, 2026 by jthomson04 Contributor Loading…
6
ProTip! Adding no:label will show everything without a label.