Skip to content

Pull requests: NovaSky-AI/SkyRL

New pull request New
151 Open 1,191 Closed
151 Open 1,191 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

[megatron] add aux loss coeff explicitly and add post init to keep config default kwargs even when overriden
#1708 opened May 26, 2026 by erictang000 Collaborator Loading…
3
[metrics] track min/max logprob diff in trainer
#1707 opened May 26, 2026 by erictang000 Collaborator Loading…
2
[logging] Wrap run in try/except + surface exceptions to wandb if supported
#1706 opened May 26, 2026 by erictang000 Collaborator Loading…
1 task
3
[logging] Add MoE auxiliary loss
#1705 opened May 26, 2026 by erictang000 Collaborator Loading…
1 task
2
Removed dead timeout parameter from DistributedStrategy subclasses
#1699 opened May 26, 2026 by jamesbraza Contributor Loading…
feat(tracking): plumb wandb tags through Tracking init
#1698 opened May 26, 2026 by dinhxuanvu Loading…
4 tasks done
11
Add Trackio rollout trace logging
#1697 opened May 22, 2026 by abidlabs Loading…
5
[feat] Support training callbacks in SkyRL
#1696 opened May 22, 2026 by SumanthRH Member Loading…
7
fix: prevent process hang when vLLM engine dies during async generation
#1691 opened May 19, 2026 by dinhxuanvu Loading…
2 of 3 tasks
6
feat: add max_training_steps config for CI smoke tests
#1690 opened May 19, 2026 by dinhxuanvu Loading…
3 of 4 tasks
3
[examples][infra] Anyscale 2-node Qwen3 launch flow
#1689 opened May 18, 2026 by xyuzh • Draft
4 tasks
5
[workers] Clean teardown on SIGTERM: drain CUDA + destroy process group
#1688 opened May 18, 2026 by xyuzh Loading…
2
[tinker] bump async db pool default 5+10→20+40 for concurrent multi-LoRA
#1686 opened May 18, 2026 by j316chuck • Draft
[bug] Switch to reload_weights API for loading weights in legacy inference codepath run_train_old_inference_gpu_ci
#1685 opened May 18, 2026 by SumanthRH Member Loading…
1
2
[ci] H100 CI run_train_gpu_ci run_train_megatron_gpu_ci_models run_train_megatron_gpu_ci
#1679 opened May 16, 2026 by hao-aaron Collaborator Loading…
8
vllm metrics docs
#1662 opened May 13, 2026 by hao-aaron Collaborator Loading…
10
[vllm] test out ray v2 executor run_train_gpu_ci run_train_megatron_gpu_ci_models run_train_megatron_gpu_ci run_train_old_inference_gpu_ci
#1660 opened May 13, 2026 by erictang000 Collaborator Loading…
3
[tx] add retries for getting random port used for JAX coordinator address
#1652 opened May 12, 2026 by andrewsykim Contributor Loading…
2
Avoid blocking Ray registry lookups for policy loss resolution
#1644 opened May 10, 2026 by taivu1998 Loading…
1
[tinker] Support additional LoRA config parameters
#1643 opened May 10, 2026 by taivu1998 Loading…
1
8
Fix multi-turn generator token budget accounting
#1642 opened May 10, 2026 by taivu1998 Loading…
1
3
Add opt-in handling for failed SkyRLGym rollouts
#1641 opened May 10, 2026 by taivu1998 Loading…
1
4
Stabilize Qwen3 MoE LoRA mixed-scale test
#1640 opened May 10, 2026 by taivu1998 Loading…
1
Remove legacy skyrl-train and skyrl-tx folders
#1639 opened May 10, 2026 by taivu1998 Loading…
1
2
[skyrl] Pass through profiler_config to vLLM engine
#1622 opened May 5, 2026 by pcmoritz Collaborator Loading…
3
ProTip! Add no:assignee to see everything that’s not assigned.