Skip to content

Pull requests: NVIDIA-NeMo/RL

New pull request New
543 Open 2,428 Closed
543 Open 2,428 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(megatron): preserve HybridEP packed-input alignment Documentation Improvements or additions to documentation
#4079 opened Sep 10, 2026 by seonjinn Contributor • Draft
2
feat(grpo): train all Gym traces per logical rollout Documentation Improvements or additions to documentation
#4078 opened Sep 10, 2026 by esarafian Loading…
4 tasks done
7
fix(qwen3.5): freeze vision modules in text-only recipes
#4077 opened Sep 10, 2026 by seonjinn Contributor • Draft
3 of 4 tasks
1
ci: Bump Megatron-Bridge to 4386117 CI:L1 Run doctests, unit tests, and functional tests
#4075 opened Sep 10, 2026 by svcnvidia-nemo-ci Contributor Loading…
3
refactor(sc): unify capture retries and extract orchestration components CI:L1 Run doctests, unit tests, and functional tests Documentation Improvements or additions to documentation
#4073 opened Sep 10, 2026 by pthombre Contributor • 2/2 • Draft
3 of 4 tasks
1
perf(reassembler): batch group staging reads and expose fetch timings CI:L1 Run doctests, unit tests, and functional tests
#4072 opened Sep 10, 2026 by pthombre Contributor • 1/2 • Draft
3 of 4 tasks
1
1
feat: add Energon-owned SFT sequence packing CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#4070 opened Sep 10, 2026 by rohitrango Contributor Loading…
21
feat(grpo): add Binary-TV DPPO training support
#4069 opened Sep 10, 2026 by guyueh1 Contributor Loading…
1
feat(sc): add per-environment rollout telemetry
#4068 opened Sep 9, 2026 by mehraakash Contributor • Draft
2
fix(sc): regenerate buffered prompts on resume
#4067 opened Sep 9, 2026 by mehraakash Contributor • Draft
2
fix(slurm): isolate Pyxis containers per job
#4066 opened Sep 9, 2026 by mehraakash Contributor Loading…
4
feat(sc): wire refit_with_reload_api into v2 entrypoint 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
#4063 opened Sep 9, 2026 by guyueh1 Contributor Loading…
4 tasks
9
feat(spec-decode): pipeline parallelism for draft co-training via a one-sided tap channel (Megatron) community-request
#4060 opened Sep 9, 2026 by MarkXCloud Contributor • Draft
1 of 4 tasks
1
feat(spec-decode): context parallelism and sequence packing for draft co-training (Megatron) community-request
#4059 opened Sep 9, 2026 by MarkXCloud Contributor • Draft
1 of 4 tasks
1
build: bump vllm to 0.28.0 CI:L2 Run doctests, unit tests, functional tests, and convergence tests Documentation Improvements or additions to documentation
#4058 opened Sep 9, 2026 by yfw Contributor • Draft
5
feat(telemetry)!: instrument the single-controller path and propagate traces CI:L1 Run doctests, unit tests, and functional tests Documentation Improvements or additions to documentation
#4052 opened Sep 8, 2026 by rrs45 Collaborator Loading…
3 of 4 tasks
7
feat(mopd): support multiple teacher checkpoints in full-vocabulary o…
#4045 opened Sep 8, 2026 by RayenTian Contributor • Draft
4 tasks
1
fix(trtllm): report response truncation so overlong filtering can see it community-request waiting-on-maintainers Waiting on maintainers to respond
#4042 opened Sep 7, 2026 by yupengtang Loading…
3 of 4 tasks
1
fix(trtllm): honor stop_strings on both TRT-LLM generation paths community-request waiting-on-maintainers Waiting on maintainers to respond
#4041 opened Sep 7, 2026 by yupengtang Loading…
3 of 4 tasks
1
perf: use HybridEP in AutoModel MoE recipes CI:L1 Run doctests, unit tests, and functional tests
#4037 opened Sep 7, 2026 by seonjinn Contributor Loading…
2
fix(vllm): load Super-Omni RADIO final LayerNorm CI:L2 Run doctests, unit tests, functional tests, and convergence tests
#4036 opened Sep 7, 2026 by DanialTaheri Contributor Loading…
4 tasks done
4
feat(grpo): add wall-clock time-efficiency reward for NeMo-Gym rollouts
#4029 opened Sep 6, 2026 by yfw Contributor Loading…
3 of 4 tasks
10
feat: add a JSD option to the distillation loss community-request waiting-on-maintainers Waiting on maintainers to respond
#4026 opened Sep 6, 2026 by kashif Loading…
3 of 4 tasks
1
fix: avoid NaN in JSD metric when logprobs underflow to zero community-request waiting-on-maintainers Waiting on maintainers to respond
#4025 opened Sep 6, 2026 by kashif Loading…
3 of 4 tasks
1
fix(refit): enforce configured timeouts in legacy GRPO
#4019 opened Sep 5, 2026 by seonjinn Contributor Loading…
3 tasks done
2
ProTip! Filter pull requests by the default branch with base:main.