horacepan / gnns

Collection of graph neural networks in pytorch

A collection of graph neural networks implemented in pytorch. Currently a work in progress.

Usage: To create a pickle file for the graph dataset: python make_pickle.py

python run.py -l {LEVELS} -hi {HIDDEN} -bs {BATCHSIZE} -e {EPOCHS} -lr {LEARNING_RATE}

This will run the molecular fingerprints regressor on the QM9 dataset.
Modify the PICKLEFILE variable in run.py to point to the location of the QM9 pickle
file.

To see all possible commandline arguments run: python run.py --help

References: "Convolutional Networks on Graphs for Learning Molecular Fingerprints" http://papers.nips.cc/paper/5954-convolutional-networks-on-graphs-for-learning-molecular-fingerprints.pdf

gnns's Issues

importing

For py36 instead of these:
from file_io import *
from graph import AdjGraph

is better to use the point there:
from .file_io import *
from .graph import AdjGraph