Skip to content

Pull requests: NovaSky-AI/SkyRL

New pull request New
53 Open 521 Closed
53 Open 521 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

[skyrl-train] Add multi-node slurm training script
#766 opened Dec 10, 2025 by BounharAbdelaziz Loading…
7
[skyrl-train][bugfix] fix bug causing metrics + overlong filtering to use strings instead of token ids for batched generation
#764 opened Dec 10, 2025 by erictang000 Loading…
2
[algo] feat: Implemented SAPO algorithm by Qwen
#762 opened Dec 9, 2025 by BounharAbdelaziz Loading…
2 tasks done
5
[skyrl-train] default logprobs=0
#758 opened Dec 9, 2025 by agolajko Loading…
10
Weight Sync Interface 1/N: refactoring weight extraction
#753 opened Dec 8, 2025 by raulchen Loading…
10
[train][Docs] Add GSM8K recipes
#746 opened Dec 8, 2025 by SumanthRH Loading…
@tyler-griggs
2
[skyrl-train] WIP Megatron LoRA
#743 opened Dec 6, 2025 by erictang000 • Draft
5
[tx] Add top_p sampling feature and basic test implementations tx
#742 opened Dec 5, 2025 by sriranganm Loading…
3
[skyrl-train] Updated FSDP Weight Sync Implementation with Bucketing for Minimizing CUDA IPC Overhead
#741 opened Dec 5, 2025 by devpatelio Loading…
@erictang000
11
Use truncated responses ids for overlong filtering instead of strings
#730 opened Dec 2, 2025 by yarinshaked Loading…
2
[tx] Compute prompt logprobs on per-micro-batch basis tx
#727 opened Dec 1, 2025 by tyler-griggs Loading…
3
[tx] Implement /api/v1/training_runs tx
#720 opened Nov 30, 2025 by VThejas Loading…
2
typo & remove unnecessary statements
#718 opened Nov 30, 2025 by AYou0207 Loading…
6
[train][CI] GPU CI Timeout 10k -> 11k skyrl-train
#711 opened Nov 26, 2025 by CharlieFRuan Loading…
2
Skip cuda-specific dependencies on a non-Linux OS skyrl-train
#705 opened Nov 25, 2025 by raulchen Loading…
5
[update] Updated RoPE Configuration for HF Models (transformers) w. backward-compatible support for vLLM #690 skyrl-train
#703 opened Nov 25, 2025 by devpatelio Loading…
30
[tx] Custom Loss Function tx
#698 opened Nov 21, 2025 by benji-cannot-code Loading…
5
[tx] Support top_k sampling tx
#680 opened Nov 18, 2025 by agolajko Loading…
@pcmoritz
13
[tx] fsdp support for qwen3 tx
#674 opened Nov 16, 2025 by taroigarashi2001 Loading…
@tyler-griggs
22
Added support for saving shard as HuggingFace safetensors for Megatron backend (WIP) skyrl-train
#658 opened Nov 12, 2025 by zhenga1 Loading…
9
[tx] Add LLama3 support tx
#657 opened Nov 11, 2025 by atemaguer Loading…
20
[POC] Add Hugo landing page in website subdirectory
#651 opened Nov 10, 2025 by andyk • Draft
[WIP] [tx] Torch definition for qwen3 and LoRA tx
#649 opened Nov 10, 2025 by tyler-griggs Loading…
14
Multi-turn native tool use RL skyrl-train
#643 opened Nov 8, 2025 by pranavraja99 Loading…
13
Add a configuration for max tokens per microbatch skyrl-train
#640 opened Nov 7, 2025 by justinvyu Loading…
@SumanthRH @erictang000
12
ProTip! Follow long discussions with comments:>50.