Generative Teaching Networks

Official PyTorch implementation of Generative Teaching Networks. For research purpose only. Support and/or new releases may be limited.

Setup

Clone the repo:

git clone https://github.com/uber-research/GTN.git && cd GTN

We use Python 3.6.2. Requirements can be installed by running:

pip install -r requirements.txt

There is a problem with torch>1.2 so we have a hard requirement on torch==1.2.

After installing the dependencies you can run pip install -e . to compile a custom torch kernel.

Training on MNIST

To train on MNIST simply run python train_cgtn.py. This command reads experiments/cgtn.json to get arguments.

Architecture Search on CIFAR10

The architecture search experiments is a separate component based on NAO which has a different LICENSE agreement. See architecture_search/ for more details.