-
Notifications
You must be signed in to change notification settings - Fork 491
Pull requests: NVIDIA-NeMo/RL
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
chore: loosen L1 eval upper bound to avoid flaky in CI
CI:L1
Run doctests, unit tests, and functional tests
#3491
opened Aug 4, 2026 by
yuki-97
Contributor
Loading…
feat: colocated Megatron reshard
CI:L1
Run doctests, unit tests, and functional tests
#3490
opened Aug 4, 2026 by
tdene
Contributor
Loading…
4 tasks
feat: [DONT MERGE] Rebasing check for opt/dev
Documentation
Improvements or additions to documentation
ci: Bump Megatron-Bridge to 3ecb452
CI:L1
Run doctests, unit tests, and functional tests
#3487
opened Aug 4, 2026 by
svcnvidia-nemo-ci
Contributor
Loading…
fix(nemo-gym): refresh final-token route across turns
CI:Lfast
Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#3485
opened Aug 4, 2026 by
zpqiu
Contributor
Loading…
feat(megatron): expose training CUDA Graph config
CI:Lfast
Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#3483
opened Aug 4, 2026 by
seonjinn
Contributor
Loading…
cp: Merge main to super-v3.5-posttraining - 2026-08-03
CI:Lfast
Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
CI
Relating to CI
Documentation
Improvements or additions to documentation
#3482
opened Aug 3, 2026 by
yfw
Contributor
Loading…
4 tasks
docs(ci-cd): add guide for adding tests and new suites to CI
Documentation
Improvements or additions to documentation
#3481
opened Aug 3, 2026 by
kajalj22
Contributor
Loading…
2 tasks
feat(sc): recover replay buffer from native TQ checkpoints
#3480
opened Aug 3, 2026 by
macandro96
Contributor
•
Draft
5 of 9 tasks
perf(vllm): batch MXFP8 MoE refit shuffles
CI:Lfast
Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#3478
opened Aug 3, 2026 by
seonjinn
Contributor
Loading…
feat(refit): support BF16 to MXFP8 NCCL reshard
CI:L1
Run doctests, unit tests, and functional tests
#3477
opened Aug 3, 2026 by
seonjinn
Contributor
Loading…
Order packed microbatches largest-first per DP rank
Documentation
Improvements or additions to documentation
feat(sc): restart dead generation shards and re-admit them at the next refit
#3473
opened Aug 3, 2026 by
asolergi-nv
Loading…
3 of 4 tasks
feat(sc): keep training when a generation shard dies
#3472
opened Aug 3, 2026 by
asolergi-nv
Loading…
3 of 4 tasks
feat(sc): track generation fleet health and route NeMo-Gym through NeMo-RL
#3471
opened Aug 3, 2026 by
asolergi-nv
Loading…
3 of 4 tasks
feat(sc): contain rollout failures in the SingleController path
CI:L1
Run doctests, unit tests, and functional tests
#3470
opened Aug 3, 2026 by
asolergi-nv
Loading…
3 of 4 tasks
test(grpo): reproduce LOO dynamic-sampling group split
#3469
opened Aug 3, 2026 by
gkaplun-nvidia
Contributor
Loading…
feat: add inflight batchsize and kvcache size metrics and for trtllm rollout
CI:Lfast
Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#3468
opened Aug 3, 2026 by
Superjomn
Contributor
Loading…
4 tasks
test(grpo): reproduce masked samples affecting group stats
#3467
opened Aug 3, 2026 by
gkaplun-nvidia
Contributor
Loading…
3 of 4 tasks
feat(draft): support sequence packing with online eagle3 draft training
CI:Lfast
Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)
#3463
opened Aug 3, 2026 by
yuekaizhang
Contributor
Loading…
feat(sc): gate-authoritative token capture on the Gym tokidcap stack
CI:L1
Run doctests, unit tests, and functional tests
Documentation
Improvements or additions to documentation
feat: nccl v2.30.7 bump
CI:L2
Run doctests, unit tests, functional tests, and convergence tests
#3453
opened Jul 31, 2026 by
youngeunkwon0405
Contributor
Loading…
4 tasks
Previous Next
ProTip!
Mix and match filters to narrow down what you’re looking for.