Interact

Interact contains implementations of several deep reinforcement learning algorithms.

Installation

Interact can be installed as follows:

git clone https://github.com/rystrauss/interact
cd interact
pip install .

If you want to use Gym environments that aren't installed by default with Gym, you'll need to install those yourself (e.g. pip install gym[atari]).

Usage

Training

An agent can be trained with the following command:

python -m interact.train --config <path_to_config_file>

This package uses Gin to configure experiments, and the --config option should be a path to a Gin config file. Algorithm-specific arguments can be found in each agent's documentation.

Some example configuration files can be found in the examples directory.

Visualizing Agents

Once an agent has been trained, it can be visualized in its environment with the following command:

python -m interact.play --agent_dir <path/to/agent/dir>

where <path/to/agent/dir> is the path to the directory that contains the agent you want to visualize (this is the directory that was created by the training script).

Implemented Algorithms

• Advantage Actor-Critic
• Deep Q-Networks
• Proximal Policy Optimization
• Phasic Policy Gradients
• Soft Actor-Critic
• Deep Deterministic Policy Gradient
• Twin Delayed Deep Deterministic Policy Gradient