Skip to content

Pull requests: NVIDIA-NeMo/RL

New pull request New
534 Open 2,343 Closed
534 Open 2,343 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(sc): stream PPO actor and critic minibatches community-request Documentation Improvements or additions to documentation
#3963 opened Sep 2, 2026 by tianyi-zhang-02 Contributor • Draft
4 tasks done
1
feat(sc): support multi-reward GDPO rollouts community-request Documentation Improvements or additions to documentation
#3962 opened Sep 2, 2026 by tianyi-zhang-02 Contributor • Draft
4 tasks done
1
feat(refit): add digest verification for vLLM IPC weight transfers CI:L1 Run doctests, unit tests, and functional tests Documentation Improvements or additions to documentation
#3961 opened Sep 2, 2026 by NolenLiang Contributor Loading…
3 of 4 tasks
3
feat(automodel): speculative decoding draft co-training
#3960 opened Sep 2, 2026 by yuekaizhang Contributor Loading…
26
chore: Bump Gym, Megatron-Bridge, vllm for superv3.5
#3958 opened Sep 2, 2026 by yfw Contributor Loading…
4 tasks
4
fix: merge LoRA into refit weights under activation checkpointing
#3953 opened Sep 1, 2026 by anubhutivyas Loading…
4 tasks
1
fix(megatron): make linear CE fusion work through VLM wrappers community-request
#3951 opened Sep 1, 2026 by dafu-wu Contributor Loading…
3 of 4 tasks
1
fix(async): honor max_num_epochs in async GRPO community-request
#3948 opened Sep 1, 2026 by bzantium Loading…
3 of 4 tasks
1
perf(docker): hardlink worker venvs into layers Documentation Improvements or additions to documentation
#3947 opened Sep 1, 2026 by tdene Contributor Loading…
4 tasks
3
feat(modelopt): add format-general real-quant vLLM rollout Documentation Improvements or additions to documentation
#3945 opened Sep 1, 2026 by mxinO Contributor • Draft
1
fix(sglang): rebind checkpoint-engine refit peers across engine restarts Documentation Improvements or additions to documentation
#3944 opened Sep 1, 2026 by Kh4L Contributor • Draft
1
perf(distillation): defer teacher top-k Ray payloads community-request
#3943 opened Sep 1, 2026 by yuchenwang3 Loading…
4 tasks done
1
2
fix: update ray.sub
#3942 opened Aug 31, 2026 by hchings Contributor Loading…
4 tasks
@hchings
1
pjin/super-v3.5-dev-20260826 Documentation Improvements or additions to documentation
#3941 opened Aug 31, 2026 by pjin-nvidia Contributor • Draft
4
fix(vlm): match training-side image tiling to the rollout's placeholder runs bug Something isn't working community-request waiting-on-customer Waiting on the original author to respond
#3940 opened Aug 31, 2026 by pulkitkumar95 Loading…
4
feat(dynamo): support prefill decode disaggregation CI:L1 Run doctests, unit tests, and functional tests Documentation Improvements or additions to documentation
#3939 opened Aug 31, 2026 by jthomson04 Contributor • Draft
18
fix(async): recover from permanently-failed prompt groups instead of hanging community-request
#3938 opened Aug 31, 2026 by pulkitkumar95 Loading…
4
feat: sampling mask replay
#3937 opened Aug 31, 2026 by zyzhou5 Contributor Loading…
4 tasks
1
feat: fp32 LM head in vLLM and MCore + fix max-length handling
#3936 opened Aug 31, 2026 by HeyyyyyyG Contributor Loading…
5
fix(automodel): align checkpoint lifecycle with consolidated save modes Documentation Improvements or additions to documentation
#3935 opened Aug 31, 2026 by jQizhang Contributor Loading…
1
2
perf(sequence-packing): keep packed-sequence boundaries on CPU CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#3933 opened Aug 31, 2026 by CarlosGomes98 Contributor Loading…
4 tasks
5
perf(sequence-packing): use CPU sequence boundaries CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#3932 opened Aug 31, 2026 by CarlosGomes98 Contributor Loading…
4 tasks
5
perf: compute selected token logprobs directly CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version) waiting-on-customer Waiting on the original author to respond
#3930 opened Aug 31, 2026 by CarlosGomes98 Contributor Loading…
4 tasks
7
feat(single-controller): pause generation during refit 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
#3929 opened Aug 31, 2026 by youngeunkwon0405 Contributor Loading…
@youngeunkwon0405
26
fix(sc): apply advantage clipping and overlong filtering 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
#3927 opened Aug 31, 2026 by yfw Contributor Loading…
13
ProTip! Type g p on any issue or pull request to go back to the pull request listing page.