Graph Datasets

The goal of this repository is to store the different graph datasets currently available as benchmarks, to provide them in an homogeneous and easily loadable way. For example, to load them in PyGeometric, you can do the following:

from datasets import load_dataset

from torch_geometric.data import Data
from torch_geometric.loader import DataLoader

graphs_dataset = load_dataset("graphs-datasets/<dataset-name>")
graphs_list = [Data(graph) for graph in graphs_dataset["<train/valid/test>"]]
graphs_pygeometric = DataLoader(graph_list)

Currently available graphs are:

Open Graph Benchmark , graph classification task: ogbg-molhiv, ogbg-molpcba, ogbg-ppa, ogbg-code2
Molecular datasets , graph regression task: ZINC, AQSOL , graph classification task: MUTAG, PROTEINS, AIDS
MD17 molecular trajectory benchmark , graph regression task: aspirin, benzene, ethanol, malonaldehyde, naphthalene, salycilic acid, toluene, uracil
Social datasets , graph classification task: IMDb-B, twitch_egos, reddit_threads, deezer_ego_nets
Image datasets , graph classification task: CIFAR10, MNIST
Quantum chemistry datasets , graph classification task: alchemy
Synthetic dataset , graph classification task: CSL

• More to come!