# google/dopamine

Dopamine is a research framework for fast prototyping of reinforcement learning algorithms.

Repository: https://github.com/google/dopamine
Canonical: https://ross.abutalabs.com/products/dopamine
Homepage: https://github.com/google/dopamine
Language: Jupyter Notebook
License: Apache-2.0
License Family: permissive
Topics: rl, ml, ai, google, tensorflow
Last push: 2026-03-24T18:11:40+00:00

## Health v2 (maintenance only)
Score: 56/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 73, release rhythm 8, longevity 100
- inputs: {"age_days": 2960, "days_push": 162, "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 10900, forks 1391 (observed 2026-08-28T04:10:44.625197+00:00)

## What it is
Dopamine is a research framework from Google for fast prototyping of reinforcement learning algorithms, built around a small, easily readable codebase. It provides battle-tested implementations of agents like DQN, C51, Rainbow, IQN, SAC, and PPO using JAX (with legacy TensorFlow support), targeting Atari and Mujoco benchmark environments.

## Use cases
- prototype a new reinforcement learning algorithm
- run DQN or Rainbow baseline experiments on Atari
- train a SAC agent on Mujoco continuous control tasks
- reproduce RL benchmark results following Machado et al. recommendations
- learn how classic RL agents are implemented
- experiment with distributional RL ideas like C51 and IQN

## When to choose
- you need a compact, hackable codebase for speculative RL research
- you want reproducible benchmark experiments on Atari or Mujoco
- you prefer JAX-based agent implementations
- you want battle-tested reference implementations of DQN, Rainbow, SAC, or PPO

## When to avoid
- you need a production RL system rather than a research prototype
- you want a large library with many prebuilt environments and utilities
- you need the newest agents without JAX, since new agents are JAX-only
- you require non-Atari/non-Mujoco environments out of the box

## Facets
- artifact type: framework
- maturity: active
- function: reinforcement-learning, machine-learning, benchmarking
- domain: reinforcement-learning, machine-learning
- platform: python
- tags: jax, tensorflow, atari, mujoco, dqn, rainbow, sac, ppo, research-framework, research, linux, macos, docker

## Member repositories
- google/dopamine (main) score 56

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:44.625197+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 2026-08-29T17:17:25.705592+00:00, confidence not recorded.
  - readme: https://github.com/google/dopamine (fetched 2026-08-28T04:10:44.625197+00:00, sha 085344b0536b)
  - homepage: https://github.com/google/dopamine (fetched 2026-08-29T08:16:16.707267+00:00, sha 33ab3532f1c6)
- Data as of 2026-08-30T08:39:29.467469+00:00.