Continuous Control With Ensemble Deep Deterministic Policy Gradients

This repository is the official implementation of Continuous Control With Ensemble Deep Deterministic Policy Gradients.

Requirements

Before installation, please make sure you have MuJoCo engine set up on your machine. We use mujoco150 in order to be comparable with previous benchmarks on v2 environments. See this issue

To install requirements:

pip install -r requirements.txt

Training

To train the model(s) in the paper, run this command:

python run.py <experiment_specification path>

Logger automatically stops training and evaluates current policy every log_every environment interactions. The data is printed to standard output and stored on drive.

We include specifications for our most important experiments.

Results

Our model achieves the following performance on the MuJoCo suite: