Skip to content

Pull requests: NVIDIA-NeMo/RL

New pull request New
101 Open 974 Closed
101 Open 974 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

cp: fix: log metrics that can be coerced to scalars (1723) into r0.5.0 cherry-pick CI:L1 Run doctests, unit tests, and functional tests Run CICD
#1730 opened Jan 6, 2026 by chtruong814 Loading…
1
fix: patch transformer qwen2 forward
#1728 opened Jan 6, 2026 by RayenTian • Draft
4 tasks
1
cp: fix: Disable cudnn sdpa backend when using activation checkpointing (1717) into r0.5.0 cherry-pick CI:L1 Run doctests, unit tests, and functional tests Run CICD
#1727 opened Jan 6, 2026 by chtruong814 Loading…
2
fix: apply offloading change from v2 to v1 CI:L0 Run doctests and unit tests r0.5.0
#1726 opened Jan 6, 2026 by terrykong Loading…
4 tasks
2
fix: remove seq_parallel + tp restriction in dtensor v2 CI:L1 Run doctests, unit tests, and functional tests r0.5.0
#1725 opened Jan 6, 2026 by terrykong • Draft
4 tasks
4
fix: fix several nightly tests that were flaky CI:L0 Run doctests and unit tests r0.5.0
#1724 opened Jan 6, 2026 by terrykong Loading…
1
fix: use median instead of mean for logprob error for stability in nightlies CI:L1 Run doctests, unit tests, and functional tests r0.5.0
#1722 opened Jan 6, 2026 by terrykong Loading…
4 tasks
21
fix: gemma3 27b must now have skip_tokenizer_init=False in vllm CI:L1 Run doctests, unit tests, and functional tests r0.5.0
#1721 opened Jan 6, 2026 by terrykong Loading…
4 tasks
2
fix: mcore generation config restored in nightly test r0.5.0
#1720 opened Jan 6, 2026 by terrykong Loading…
4 tasks
4
Test fixes
#1718 opened Jan 5, 2026 by terrykong • Draft
1
feat: refactor common data utilities of dtensor policy v2
#1710 opened Jan 5, 2026 by hemildesai • Draft
4 tasks
1
feat: refactor init of dtensor policy v2
#1709 opened Jan 5, 2026 by hemildesai • Draft
4 tasks
1
Update index.md documentation Improvements or additions to documentation
#1706 opened Jan 2, 2026 by snowmanwwg Loading…
4 tasks
3
Create model-support.md for NeMoRL documentation Improvements or additions to documentation
#1705 opened Jan 2, 2026 by snowmanwwg Loading…
4 tasks
3
feat: Support lora for grpo workflow CI:L1 Run doctests, unit tests, and functional tests
#1702 opened Dec 28, 2025 by RayenTian • Draft
4 tasks
1
8
[don't merge] support multiple dataloader CI:L1 Run doctests, unit tests, and functional tests
#1698 opened Dec 24, 2025 by yuki-97 • Draft
feat: RL support for custom moe models in dtensor v2 CI:L1 Run doctests, unit tests, and functional tests
#1695 opened Dec 24, 2025 by hemildesai Loading…
@hemildesai
22
[don't merge] support multiple datasets for response dataset CI:L1 Run doctests, unit tests, and functional tests
#1691 opened Dec 23, 2025 by yuki-97 • Draft
feat: Add SGLang rollout backend and tests CI:L2 Run doctests, unit tests, functional tests, and convergence tests community-request
#1674 opened Dec 21, 2025 by RolaoDenthu Loading…
4 tasks
37
Nano v3 lora
#1669 opened Dec 20, 2025 by arendu • Draft
4 tasks
fix: Add debug parameter to reduce verbose output community-request
#1664 opened Dec 19, 2025 by sahgerlad Loading…
1
refactor: Order node by IP in GRPO CI:L2 Run doctests, unit tests, functional tests, and convergence tests GB200
#1655 opened Dec 18, 2025 by guyueh1 Loading…
4 tasks
@guyueh1
7
feat: refactor mcore train/forward utilities
#1654 opened Dec 17, 2025 by ashors1 • Draft
4 tasks
chore: update Megatron-LM submodule to ed804b4
#1653 opened Dec 17, 2025 by yaoyu-33 Loading…
4 tasks
6
feat: refactor megatron data utils
#1651 opened Dec 17, 2025 by ashors1 • Draft
4 tasks
ProTip! Filter pull requests by the default branch with base:main.