Skip to content

Pull requests: NovaSky-AI/SkyRL

New pull request New
246 Open 1,495 Closed
246 Open 1,495 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

[deps] bump transformers to 5.16.1
#2155 opened Sep 4, 2026 by erictang000 Collaborator • Draft
14 tasks
5
fix(megatron): stage common state for cloud resume
#2154 opened Sep 3, 2026 by bvolpato Contributor • Draft
fix(megatron): isolate FP32 adapter optimizer state
#2153 opened Sep 3, 2026 by bvolpato Contributor • Draft
fix(megatron): refresh masters after weights-only load
#2149 opened Sep 3, 2026 by bvolpato Contributor • Draft
fix(rollout): preserve sampled EOS probabilities
#2148 opened Sep 3, 2026 by bvolpato Contributor • Draft
fix(inference): reset requests when discarding KV cache
#2147 opened Sep 3, 2026 by bvolpato Contributor • Draft
fix(megatron): synchronize complete accumulation windows
#2146 opened Sep 3, 2026 by bvolpato Contributor • Draft
fix(megatron): finalize checkpoints before pruning
#2145 opened Sep 3, 2026 by bvolpato Contributor • Draft
perf(fsdp): limit actor LM-head work to the action window
#2144 opened Sep 3, 2026 by NenoL2001 Loading…
1
3
TITO support for SkyRL
#2143 opened Sep 3, 2026 by xinze-zheng Loading…
@SumanthRH
12
vLLM Stateful Weight Sync Sender run_h100_gpu_ci Run H100 GPU CI run_train_gpu_ci run_train_megatron_gpu_ci
#2142 opened Sep 3, 2026 by hao-aaron Collaborator Loading…
@SumanthRH
16
perf(megatron): autotune fused LM-head Triton kernels
#2141 opened Sep 2, 2026 by dyurk-lila Contributor Loading…
5
fix(tinker): align forwarding pool deadline
#2140 opened Sep 2, 2026 by hershg Contributor Loading…
1
[fix] Build DeepGEMM against selected torch
#2139 opened Sep 2, 2026 by hershg Contributor Loading…
9
[B300] Enable FlashAttention 4 on CUDA 13 Images run_train_megatron_gpu_ci_models x-trajectory
#2132 opened Sep 1, 2026 by j316chuck Contributor Loading…
@SumanthRH
6
test: add GLM 5.3 LoRA logprob parity coverage x-trajectory
#2128 opened Sep 1, 2026 by hershg Contributor • Draft
[Tinker] Keep SQLite off the forwarded-future hot path x-trajectory
#2127 opened Sep 1, 2026 by j316chuck Contributor Loading…
5
[tinker] Retry transient unavailable inference workers x-trajectory
#2126 opened Sep 1, 2026 by j316chuck Contributor Loading…
2
[tinker] Raise SQLite busy timeout under rollout load x-trajectory
#2124 opened Sep 1, 2026 by j316chuck Contributor Loading…
2
[tinker] Size the API database pool for rollout bursts x-trajectory
#2123 opened Sep 1, 2026 by j316chuck Contributor Loading…
2
[tinker] Bound non-colocated forwarding fan-out x-trajectory
#2122 opened Sep 1, 2026 by j316chuck Contributor Loading…
3
[fix][inference] Cancel abandoned generation
#2121 opened Sep 1, 2026 by bvolpato Contributor Loading…
3
perf(megatron): preserve LoRA export dtype
#2117 opened Sep 1, 2026 by hershg Contributor Loading…
5
fix(megatron): release LoRA export before reload x-trajectory
#2113 opened Sep 1, 2026 by hershg Contributor • Draft
fix(megatron): release host model replica after backload x-trajectory
#2112 opened Sep 1, 2026 by hershg Contributor Loading…
5
ProTip! Add no:assignee to see everything that’s not assigned.