MishaLaskin/rad
A research implementation of Reinforcement Learning with Augmented Data (RAD), supporting image-based RL agents (SAC, PPO, CURL) on DM-Control and OpenAI Gym.
Not currently ranked — collecting fresh signals.
star history
This repository provides the official implementation of the RAD paper (Laskin et al., 2020), which combines data augmentations with model-free deep reinforcement learning to improve sample efficiency. It includes implementations of SAC, PPO, and CURL agents, with configurable augmentation pipelines (crop, rotate, flip, etc.) for image-based observations. The codebase trains agents on environments from DM-Control and OpenAI Gym using PyTorch, with configurable hyperparameters via command-line arguments.
Frequently asked
- What is MishaLaskin/rad?
- A research implementation of Reinforcement Learning with Augmented Data (RAD), supporting image-based RL agents (SAC, PPO, CURL) on DM-Control and OpenAI Gym.
- Is rad open source?
- Yes — MishaLaskin/rad is an open-source project tracked on heatdrop.
- What language is rad written in?
- MishaLaskin/rad is primarily written in Jupyter Notebook.
- How popular is rad?
- MishaLaskin/rad has 418 stars on GitHub.
- Where can I find rad?
- MishaLaskin/rad is on GitHub at https://github.com/MishaLaskin/rad.