#7575th of 75 in ML frameworks and libraries
dopamine
Reinforcement learning research framework with compact JAX agent implementations
- Stars
- 10.9k
- License
- Apache-2.0
- Last commit
- Mar 2026
- Last release
- Sep 2019
- Language
- Jupyter Notebook
10,809 on 31 Mar 202610,926 on 20 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Dopamine is a small Python codebase for prototyping reinforcement learning algorithms, built to be read and modified directly. It ships JAX implementations of DQN, C51, Rainbow, IQN, SAC and PPO, with legacy TensorFlow versions of many, and targets Atari and Mujoco environments. Baseline results and Colab notebooks are provided, and the setup follows Machado et al. (2018) for reproducibility.
Who it is for: RL researchers prototyping algorithms on Atari or Mujoco
Strengths
- Small codebase meant to be edited directly for research experiments
- Covers DQN, C51, Rainbow, IQN, SAC and PPO in JAX
- Published baseline results and Colab notebooks for each agent
- Evaluation setup follows Machado et al. (2018) recommendations
Weaknesses
- Last tagged release is from 2019-09-26
- Only a few algorithms; not a broad RL library
- Mujoco setup references mujoco-py and a license step, which is dated
- TensorFlow agents are legacy; new agents are likely JAX-only
What it needs
- Docker
- Needs jax, tensorflow (legacy agents), ale_py, mujoco-py, openai baselines
- Models: DQN, C51, Rainbow, IQN, SAC
Also in ML frameworks and libraries
See all 75| Rank | Project | Score |
|---|---|---|
| 1 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 94 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd for deep learning | 94 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 92 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and reference training scripts | 91 out of 100 |
| #5 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 90 out of 100 |