Acme
A library of reinforcement learning components and agents
Overview
Acme is a library of reinforcement learning (RL) building blocks designed for researchers. It provides simple, efficient, and readable reference implementations and baseline agents for algorithm performance. Acme is flexible and can be used as a starting point for novel research. It also supports running agents at multiple scales, such as single-stream or distributed agents.
Features
- Exposes simple, efficient, and readable RL agents
- Provides strong baselines for algorithm performance
- Can be used as a starting block for novel research
- Supports running agents at multiple scales
Summary
Acme is a research framework for reinforcement learning that provides simple and efficient RL agents. It offers strong baselines for algorithm performance and supports running agents at multiple scales. The library is designed to be used by researchers and provides reference implementations for novel research. It can be installed in a Python virtual environment and requires additional dependencies such as JAX or TensorFlow.