# PKU-Alignment/omnisafe

JMLR: OmniSafe is an infrastructural framework for accelerating SafeRL research.

Repository: https://github.com/PKU-Alignment/omnisafe
Canonical: https://ross.abutalabs.com/products/omnisafe
Homepage: https://omnisafe.readthedocs.io/en/latest/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: benchmark, pytorch, safe-reinforcement-learning, deep-reinforcement-learning, reinforcement-learning, machine-learning, constraint-rl, constraint-satisfaction-problem, deep-learning, safe-rl, saferl, safety-critical, safety-gym, safety-gymnasium
Last push: 2025-03-17T11:45:44+00:00

## Health v2 (maintenance only)
Score: 27/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 11, release rhythm 8, longevity 98
- inputs: {"age_days": 1383, "days_push": 534, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1149, forks 161 (observed 2026-08-28T04:03:46.402897+00:00)

## What it is
OmniSafe is a PyTorch-based infrastructural framework for safe reinforcement learning research, providing a unified modular toolkit and comprehensive benchmark of safe RL algorithms. It abstracts algorithm types behind Adapter and Wrapper designs to support tens of constrained/safe RL algorithms across diverse domains.

## Use cases
- benchmark safe reinforcement learning algorithms
- research constrained RL with safety constraints
- develop and compare safe RL algorithms in PyTorch
- train RL agents that minimize unsafe behavior
- evaluate algorithms on Safety-Gymnasium environments
- build new safe RL algorithms on a modular framework

## When to choose
- you need a unified, modular framework for safe/constrained RL research
- you want a reliable benchmark suite for safe RL algorithms
- you work in PyTorch and need out-of-box safe RL algorithm implementations

## When to avoid
- you only need standard (unconstrained) RL algorithms
- you need production RL deployment rather than research tooling
- you prefer non-PyTorch frameworks like JAX or TensorFlow

## Facets
- artifact type: framework
- maturity: active
- function: machine-learning, reinforcement-learning, benchmarking, simulation
- domain: reinforcement-learning, machine-learning, deep-learning, artificial-intelligence
- platform: python, windows
- tags: safe-reinforcement-learning, pytorch, constrained-rl, safety-gymnasium, benchmark, research-framework, gpu, linux, macos

## Member repositories
- PKU-Alignment/omnisafe (main) score 27

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:46.402897+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 2026-08-30T06:33:43.266502+00:00, confidence not recorded.
  - readme: https://github.com/PKU-Alignment/omnisafe (fetched 2026-08-28T04:03:46.402897+00:00, sha f0e10f7cdc7a)
  - registry_pypi: https://pypi.org/pypi/omnisafe/json (fetched 2026-08-29T12:39:07.687789+00:00, sha 502b047008f3)
- Data as of 2026-08-30T08:39:29.467469+00:00.