Skip to content

Pull requests: NovaSky-AI/SkyRL

New pull request New
195 Open 1,401 Closed
195 Open 1,401 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

perf(sample-support): reuse replay scores for entropy
#2004 opened Aug 7, 2026 by dyurk-lila Contributor Loading…
2
Initialize base inference before training model creation
#2003 opened Aug 7, 2026 by j316chuck Contributor Loading…
4
[fix][ci] Format Tinker engine test
#2002 opened Aug 7, 2026 by bvolpato Contributor Loading…
[tinker] Keep the multi-tenant LoRA runtime warm on last unload
#2001 opened Aug 7, 2026 by avigyabb Collaborator Loading…
3
[fix][megatron] Make multi-LoRA adapter swaps safe against colocation offload
#2000 opened Aug 7, 2026 by erictang000 Collaborator Loading…
fix(megatron): support GLM-5.2 FP8 LoRA training
#1999 opened Aug 6, 2026 by j316chuck Contributor • Draft
[fix][train] Shut down DataLoader workers on exit
#1998 opened Aug 6, 2026 by bvolpato Contributor Loading…
3
Keep the LoRA runtime warm between jobs
#1997 opened Aug 6, 2026 by j316chuck Contributor • Draft
Enable load_weights_only in checkpointing
#1996 opened Aug 6, 2026 by j316chuck Contributor Loading…
1
8
Retry transient inference proxy unavailability
#1994 opened Aug 5, 2026 by kalectory Loading…
2
fix(lora): backload optimizer before adapter swap
#1991 opened Aug 5, 2026 by pcmoritz Collaborator Loading…
@erictang000
4
[feat][SFT] Union sampling for multi-dataset runs without explicit weights
#1990 opened Aug 5, 2026 by avigyabb Collaborator • Draft
[deps] Upgrade vllm-router and fix 0.1.15 startup
#1988 opened Aug 5, 2026 by bvolpato Contributor Loading…
test(tx): scale the MoE LoRA tolerance to the output magnitude
#1987 opened Aug 5, 2026 by pranavraja99 Loading…
1
2
feat(generators): add generator.skip_failed_rollouts
#1986 opened Aug 5, 2026 by pranavraja99 Loading…
1
7
fix(train): reject a train_batch_size that truncates the prompt batch to zero
#1985 opened Aug 5, 2026 by pranavraja99 Loading…
1
2
[feat][JAX] Add Apple Silicon MPS support
#1979 opened Aug 4, 2026 by bvolpato Contributor • Draft
[perf][tinker] Resolve awaited requests with one shared batched poller
#1978 opened Aug 4, 2026 by avigyabb Collaborator Loading…
5
[perf][tinker] Add a covering index for the engine's pending scan
#1977 opened Aug 3, 2026 by avigyabb Collaborator • Draft
[chore][deps] Support GLM-5.2 configs
#1974 opened Aug 3, 2026 by bvolpato Contributor Loading…
[feat][SFT] Stream tokenization to arrow, bounding memory for large text datasets
#1971 opened Aug 3, 2026 by avigyabb Collaborator Loading…
2
[feat][SFT] Serve the tokenized-dataset cache memory-mapped
#1970 opened Aug 3, 2026 by avigyabb Collaborator Loading…
3
[fix][metrics] Reduce policy_entropy and logprob-diff metrics via sum/count
#1969 opened Aug 2, 2026 by ayucpp Loading…
1
3
[feat][fsdp] Add QLoRA and Gemma 4 training support
#1968 opened Aug 1, 2026 by bvolpato Contributor Loading…
10
[feat] 2/n Adding NVFP4 Humans& Quantization
#1967 opened Jul 31, 2026 by kailash109 Contributor • Draft
ProTip! Follow long discussions with comments:>50.