Skip to content

Pull requests: NovaSky-AI/SkyRL

New pull request New
107 Open 972 Closed
107 Open 972 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

[train] Fix double-serialization of TensorBatch in pickle
#1379 opened Mar 24, 2026 by erictang000 Loading…
3
[router replay] downcast expert router indices to uint8/int16 to reduce space
#1378 opened Mar 24, 2026 by erictang000 Loading…
5
[train] Prefix-aware merge for step-wise trajectories (#1277)
#1377 opened Mar 24, 2026 by deepsheth3 Loading…
9
[trainer] refactor dispatch_from_staged to individually serialize DP chunks to avoid materializing whole batch on all workers
#1376 opened Mar 24, 2026 by erictang000 Loading…
3 tasks
5
[bug] Move server creation and server start in the same thread
#1375 opened Mar 23, 2026 by hao-aaron Loading…
Bump vLLM to 0.18
#1374 opened Mar 23, 2026 by hao-aaron Loading…
6
[SkyRL] Introduce /render endpoint to the new http inference client
#1373 opened Mar 23, 2026 by nithinvc • Draft
2 tasks done
6
Ulysses position_ids pre-gather, NUMA rewrite, and operational improvements
#1371 opened Mar 23, 2026 by ashutoshuiuc Loading…
9
FSDP2/MoE core training stability fixes
#1370 opened Mar 23, 2026 by ashutoshuiuc Loading…
21
Add Anthropic Messages API (/v1/messages) endpoint
#1369 opened Mar 23, 2026 by ashutoshuiuc Loading…
1
16
Improved LoRA weight swap and robust transitions_to_training_data
#1368 opened Mar 23, 2026 by ashutoshuiuc Loading…
1
10
[train] LoRA-only weight syncing for Megatron
#1360 opened Mar 20, 2026 by hao-aaron Loading…
1
@erictang000
30
feat: add multi-node WandB system metrics aggregation
#1358 opened Mar 20, 2026 by penfever Loading…
3 tasks
7
feat: make max_logprobs configurable for inference engines
#1357 opened Mar 20, 2026 by penfever Loading…
2 tasks
fix: stagger vLLM engine startup to avoid EADDRINUSE
#1356 opened Mar 20, 2026 by penfever Loading…
2 tasks
3
Fix/causal conv1d cuda128
#1354 opened Mar 20, 2026 by dzorlu Loading…
8
[bug][examples] Fix lcb dataset issue in examples dir
#1349 opened Mar 19, 2026 by s-chundi Loading…
1
5
[train][2/N] Support for Megatron PP + CP for R3
#1335 opened Mar 17, 2026 by devpatelio Loading…
25
[tx] Enable MPS backend for Apple Silicon on Jax backend tx
#1332 opened Mar 17, 2026 by pcmoritz Loading…
3 tasks
2
[train] Multi-modal inputs support in FSDP2
#1331 opened Mar 17, 2026 by nithinvc Loading…
@SumanthRH
28
Glm r3
#1326 opened Mar 14, 2026 by erictang000 • Draft
WIP Qwen 3.5 Dense FSDP
#1313 opened Mar 11, 2026 by CharlieFRuan • Draft
[bug][train] Fix max_seq_len calculation
#1303 opened Mar 10, 2026 by tamoghnokandar Loading…
1
@SumanthRH
7
MoE/FSDP2 training fixes, Anthropic Messages API, and robustness improvements
#1298 opened Mar 9, 2026 by ashutoshuiuc Loading…
4 tasks done
2
26
[skyrl-train] Implement loss reduction via advantage normalization and fix token_mean reduction strategy
#1296 opened Mar 9, 2026 by justinvyu Loading…
10
ProTip! no:milestone will show everything without a milestone.