Genetic Algorithms on the GPU
This is the version of gaggle used for the experiments in the paper: Gaggle: Genetic Algorithms on the GPU using Pytorch.
The code can be also accessed as a package by installing it with pip:
pip install torch-gaggle
To run the code during review, use the environment provided in the parent directory of the experiment code.
We provided a simple training example script in the examples folder.
It can be run in the following way from the examples folder:
python3 train.py --config_path ../configs/train_mnist_lenet.yml Two tutorials can be found in the tutorials folder.
The first one: introduction.ipynn covers using the GASupervisor to solve pre-built problems and get a high level overview of using Genetic Algorithms to solve problems.
The second one: research_mode.ipynb goes into a lot more depth and covers each of the main components of the inner workings of Gaggle to allow for configuration file support, reproducible experiments and custom code integration.
The paper experiment code can be found at Gaggle Experiment Code.