# NovaSky-AI/SkyRL

SkyRL: A Modular Full-stack RL Library for LLMs

Repository: https://github.com/NovaSky-AI/SkyRL
Canonical: https://ross.abutalabs.com/products/skyrl
Homepage: https://docs.skyrl.ai/docs
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-08-24T20:44:14+00:00

## Health v2 (maintenance only)
Score: 80/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 81, longevity 35
- inputs: {"age_days": 498, "days_push": 9, "days_rel": 48, "gap_med": 52.0, "n_releases_24m": 7}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2201, forks 410 (observed 2026-08-28T04:06:25.527538+00:00)

## What it is
SkyRL is a modular full-stack reinforcement learning library for post-training large language models, combining a training framework (skyrl-train), a Tinker API-compatible backend (skyrl-tx), an agent layer for long-horizon tasks (skyrl-agent), and a Gymnasium-style environment library (skyrl-gym). It supports FSDP and Megatron training backends with vLLM inference on your own NVIDIA GPUs, including LoRA and full-parameter fine-tuning.

## Use cases
- train llm with reinforcement learning on my own gpus
- run tinker api training scripts locally without code changes
- fine-tune llm with grpo on gsm8k
- build custom gym environments for rlhf training
- train multi-turn tool-use agents with rl
- rl post-training for sql and search agents
- supervised fine-tuning of large language models with lora
- train vision-language models with reinforcement learning

## When to choose
- you want to run RL post-training (GRPO, PPO, DAPO) for LLMs on your own NVIDIA hardware
- you need Tinker API compatibility with zero code changes on local GPUs
- you are training long-horizon, multi-turn tool-use agents
- you want a modular RL stack with pluggable environments, trainers, and inference engines
- you need to scale from 0.5B to 200B+ parameter models with FSDP or Megatron

## When to avoid
- you only need simple SFT without RL and prefer lightweight fine-tuning tools
- you have no NVIDIA GPUs or CUDA 13.0-capable drivers
- you want a fully managed hosted training service rather than self-managed infrastructure
- you need RL for non-LLM domains like robotics control

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, llm-training, agent-framework, gpu-computing, sdk
- domain: large-language-models, reinforcement-learning, machine-learning, gpu-computing
- platform: python, cloud
- tags: reinforcement-learning, rlhf, grpo, tinker-api, fsdp, megatron, vllm, ray, lora, post-training, gymnasium-environments, multi-turn-agents, ai-agents, gpu, docker, linux

## Member repositories
- NovaSky-AI/SkyRL (main) score 80

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:25.527538+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:47:03.409855+00:00, confidence not recorded.
  - readme: https://github.com/NovaSky-AI/SkyRL (fetched 2026-08-28T04:06:25.527538+00:00, sha a789e897cf5f)
  - homepage: https://docs.skyrl.ai/docs (fetched 2026-08-29T10:27:15.786154+00:00, sha 5f840523ae09)
  - site_page: https://docs.skyrl.ai/docs/tinker/architecture (fetched 2026-08-29T10:27:15.803771+00:00, sha 9c84a192532d)
  - site_page: https://docs.skyrl.ai/docs/tinker/cookbook (fetched 2026-08-29T10:27:15.805568+00:00, sha 609885930091)
  - site_page: https://docs.skyrl.ai/docs/tutorials/new_env (fetched 2026-08-29T10:27:15.808908+00:00, sha 226c5827a321)
  - site_page: https://docs.skyrl.ai/docs/tinker/overview (fetched 2026-08-29T10:27:15.795075+00:00, sha 6109706116f7)
  - site_page: https://docs.skyrl.ai/docs/getting-started/installation (fetched 2026-08-29T10:27:15.796842+00:00, sha 33b8ffed345c)
  - site_page: https://docs.skyrl.ai/docs/getting-started/quickstart (fetched 2026-08-29T10:27:15.798695+00:00, sha f2bd0cd5ac8d)
  - site_page: https://docs.skyrl.ai/docs/getting-started/overview (fetched 2026-08-29T10:27:15.800408+00:00, sha e8e0b7772ec3)
  - site_page: https://docs.skyrl.ai/docs/tinker/quickstart (fetched 2026-08-29T10:27:15.801999+00:00, sha 77452d3188d2)
- Data as of 2026-08-30T08:39:29.467469+00:00.