Skip to content

Pull requests: NVIDIA-NeMo/RL

New pull request New
71 Open 870 Closed
71 Open 870 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

feat: per-worker active/idle timeline + IFB size logging CI:L1 Run doctests, unit tests, and functional tests
#1534 opened Nov 18, 2025 by youngeunkwon0405 • Draft
4 tasks
@youngeunkwon0405
6
feat: Support qwen3-next, mcore path
#1530 opened Nov 17, 2025 by ahmadki Loading…
1 task
11
feat: force on-policy ratio to 1
#1529 opened Nov 17, 2025 by yfw • Draft
4 tasks
perf: perf script change for qwen30b-a3b CI:docs Run doctest
#1526 opened Nov 15, 2025 by youngeunkwon0405 Loading…
4 tasks
@youngeunkwon0405
4
feat: RL sampler [WIP]
#1522 opened Nov 14, 2025 by pjin-nvidia • Draft
4 tasks
feat: Add moe load balancing metrics
#1520 opened Nov 13, 2025 by yfw • Draft
4 tasks
1
feat: refactor dtensor policy v2 into modular functions CI:L0 Run doctests and unit tests
#1511 opened Nov 12, 2025 by hemildesai • Draft
4 tasks
16
feat: Automodel init for DTensorPolicyV2 CI:L2 Run doctests, unit tests, functional tests, and convergence tests
#1509 opened Nov 12, 2025 by adil-a Loading…
14
feat: add llama3.3 nemotron super 49b recipes CI:L1 Run doctests, unit tests, and functional tests documentation Improvements or additions to documentation
#1506 opened Nov 11, 2025 by yuki-97 Loading…
24
build: Use dynamic engine for generate.
#1502 opened Nov 11, 2025 by shanmugamr1992 Loading…
4 tasks
5
refactor: refactor config matching from megatron-bridge
#1501 opened Nov 10, 2025 by youngeunkwon0405 • Draft
4 tasks
@youngeunkwon0405
3
feat: pipeline-rl style # of inflight prompt regulation CI:L1 Run doctests, unit tests, and functional tests documentation Improvements or additions to documentation
#1499 opened Nov 10, 2025 by youngeunkwon0405 Loading…
4 tasks
@youngeunkwon0405
11
fix: Support vLLM DP+EP in async engine via Ray-level data parallelism community-request
#1495 opened Nov 10, 2025 by clumsy Loading…
4 tasks done
10
feat: allow uv-less execution and fingerprint the environment CI:L1 Run doctests, unit tests, and functional tests CI Relating to CI documentation Improvements or additions to documentation
#1491 opened Nov 9, 2025 by terrykong Loading…
1
4
fix: Megatron static inference and adapt to mcore engine API changes CI:L1 Run doctests, unit tests, and functional tests r0.4.0
#1488 opened Nov 7, 2025 by shanmugamr1992 Loading…
4 tasks
10
feat: Add AceMathRL recipe
#1484 opened Nov 6, 2025 by ffrujeri • Draft
4 tasks
feat: fp16 for DTensor policies
#1474 opened Nov 5, 2025 by adil-a Loading…
4
Mmanohara/merge grpo helpsteer cp tp community-request
#1472 opened Nov 4, 2025 by nv-mmanohara Loading…
4 tasks
19
feat: DTensorPolicyV2 GPT-OSS support CI:L0 Run doctests and unit tests
#1470 opened Nov 4, 2025 by adil-a Loading…
30
build: Ensure automodel has deepep and TE
#1456 opened Oct 31, 2025 by chtruong814 Loading…
4 tasks
feat: Random dataset with specified input and output sequence length CI:L0 Run doctests and unit tests
#1453 opened Oct 31, 2025 by guyueh1 Loading…
4 tasks
1
10
feat: Add GPT-OSS support via mcore
#1452 opened Oct 31, 2025 by ashors1 • Draft
4 tasks
1
2
feat: [draft do not merge] Fp8 moe rollout CI:L0 Run doctests and unit tests documentation Improvements or additions to documentation
#1446 opened Oct 29, 2025 by guyueh1 Loading…
4 tasks
2
@guyueh1
8
fix: add theoretical TFlops for H200 GPU CI:L0 Run doctests and unit tests
#1422 opened Oct 24, 2025 by roclark Loading…
4 tasks done
3
DRAFT: feat: Enable simulated user for multi-turn GRPO
#1412 opened Oct 22, 2025 by ahmadki Loading…
4 tasks
16
ProTip! no:milestone will show everything without a milestone.