# opendilab/DI-engine

OpenDILab Decision AI Engine. The Most Comprehensive Reinforcement Learning Framework B.P.

Repository: https://github.com/opendilab/DI-engine
Canonical: https://ross.abutalabs.com/products/di-engine
Homepage: https://di-engine-docs.readthedocs.io
Language: Python
License: Apache-2.0
License Family: permissive
Topics: reinforcement-learning, multiagent-reinforcement-learning, self-play, imitation-learning, inverse-reinforcement-learning, exploration-exploitation, distributed-system, python, impala, smac, atari, mujoco, minigrid, r2d2, reinforcement-learning-algorithms, pytorch-rl, offline-rl, drl, distributed-reinforcement-learning, model-based-reinforcement-learning
Last push: 2025-12-07T09:44:53+00:00

## Health v2 (maintenance only)
Score: 48/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 56, release rhythm 8, longevity 100
- inputs: {"age_days": 1886, "days_push": 269, "days_rel": 618, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3638, forks 437 (observed 2026-08-28T04:08:12.750270+00:00)

## What it is
DI-engine is an open-source reinforcement learning framework from OpenDILab that provides comprehensive implementations of deep RL algorithms built on PyTorch. It supports single-agent, multi-agent, offline, and model-based RL with distributed training capabilities across many benchmark environments.

## Use cases
- train reinforcement learning agents in Python
- run multi-agent RL experiments with self-play
- implement imitation learning and inverse RL
- train offline RL policies from datasets
- benchmark RL algorithms on Atari, MuJoCo, and SMAC
- run distributed RL training like IMPALA

## When to choose
- you need a broad collection of ready-to-use RL algorithms in PyTorch
- you want distributed or multi-agent RL training support
- you need offline RL or imitation learning out of the box

## When to avoid
- you need a lightweight single-algorithm library with minimal dependencies
- you work outside Python or PyTorch ecosystems
- you only need simple supervised deep learning

## Facets
- artifact type: framework
- maturity: active
- function: reinforcement-learning, machine-learning, deep-learning, agent-framework, simulation
- domain: reinforcement-learning, machine-learning, artificial-intelligence
- platform: python
- tags: pytorch, multi-agent-rl, offline-rl, imitation-learning, self-play, distributed-rl, decision-ai, ai-agents, game-development, linux, macos, gpu

## Member repositories
- opendilab/DI-engine (main) score 48

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:12.750270+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 2026-08-29T18:31:49.515107+00:00, confidence not recorded.
  - readme: https://github.com/opendilab/DI-engine (fetched 2026-08-28T04:08:12.750270+00:00, sha 05b63708aeaa)
  - registry_pypi: https://pypi.org/pypi/di-engine/json (fetched 2026-08-29T09:26:53.497457+00:00, sha 480d5ff469ab)
- Data as of 2026-08-30T08:39:29.467469+00:00.