Code release for our paper here: https://arxiv.org/abs/2311.04986
conda env create -f environment.yaml
For Moving MNIST (https://www.cs.toronto.edu/~nitish/unsupervised_video/):
cd datasets/MNIST && chmod +x download.sh && ./download.sh
For Kinetics (https://www.crcv.ucf.edu/data/UCF101.php):
cd datasets/UCF101 && chmod +x download.sh && ./download.sh
To start a training run, run the following (leave the comma behind the 0,
):
PYTHONPATH=. python main.py --base <path_to_config> --devices 0,