Differentiable Quality Diversity

This repository is the official implementation of Differentiable Quality Diversity.

The project contains a modified version of pyribs a quality diversity optimization library. All MEGA variants are implemented in pyribs. The GradientEmitter implements both the OG-MAP-Elites and the OMG-MEGA algorithms. The GradientImprovementEmitter implements the CMA-MEGA algorithm.

See ribs/emitters/_gradient_emitter.py and ribs/emitters/_gradient_improvement_emitter.py.

Requirements

The project builds in Anaconda.

Here are the instructions to create the conda environment:

conda env create -f experiments/environment.yml

Next install the local copy of pyribs after activating conda:

conda activate dqdexps
pip3 install -e .[all]

Pretrained Models

You can download the StyleGAN pretrained models from the StyleGAN repo. Place the .pt file in the folder experiments/lsi_clip.

CLIP automatically installs with the conda environment.

Running Experiments

For each experiment you pick an identifier for the algorithm you want to run.

Linear Projection (sphere)

To run an experiment with MAP-Elites:

conda activate dqdexps
cd experiments/lin_proj

python3 lin_proj.py map_elites --objective sphere

To run a different algorithm replace map_elites with another identifier from the above table.

For additional options see:

python3 lin_proj.py --help

Linear Projection (Rastrigin)

To run an experiment with MAP-Elites:

conda activate dqdexps
cd experiments/lin_proj

python3 lin_proj.py map_elites --objective Rastrigin

To run a different algorithm replace map_elites with another identifier from the above table.

For additional options see:

python3 lin_proj.py --help

Arm Repertoire

To run an experiment with MAP-Elites:

conda activate dqdexps
cd experiments/arm

python3 arm.py map_elites

To run a different algorithm replace map_elites with another identifier from the above table.

For additional options see:

python3 arm.py --help

Latent Space Illumination (LSI)

To run an experiment with MAP-Elites:

conda activate dqdexps
cd experiments/lsi_clip

python3 lsi.py map_elites 

To run a different algorithm replace map_elites with another identifier from the above table.

For additional options see:

python3 lsi.py --help

Results

The following tables contain the reported results from the DQD paper.

Linear Projection (sphere)

Linear Projection (Rastrigin)

Arm Repertoire

Latent Space Illumination (LSI)

See the paper and supplementary materials for full data and standard error bars.

License

pyribs and this project are both released under the MIT License.

pyribs MIT License