NESAP Extreme Spatio-Temporal Learning

Deep learning on large spatio-temporal data, including fMRI and climate data.

Datasets

• Moving MNIST
• Brain fMRI
• Climate

Models

• PredRNN++

Package layout

The directory layout of this repo is designed to be flexible:

• Configuration files (in YAML format) go in configs/
• Dataset specifications using PyTorch's Dataset API go into datasets/
• Model implementations go into models/
• Trainer implementations go into trainers/. Trainers inherit from BaseTrainer and are responsible for constructing models as well as training and evaluating them.

All examples are run with the generic training script, train.py.

How to run

To run the examples on the Cori supercomputer, you may use the provided example SLURM batch script. Here's how to run the Hello World example on 4 Haswell nodes:

sbatch -N 4 scripts/train_cori.sh configs/hello.yaml