Open Graph Benchmark (OGB)

A collection of benchmark datasets, data-loaders and evaluators for graph machine learning in PyTorch. Data loaders are fully compatible with PyTorch Geometric and Deep Graph Library (DGL). The goal is to have an easily-accessible standardized large-scale benchmark datasets to drive research in graph machine learning.

Datasets available

Benchmark datasets are broadly classified into three categories. Datasets that are currently available are also listed (more to come soon).

Node property prediction : Prediction on single nodes.
- Prediction of protein functionality in a protein-protein association network.
Link property prediction : Prediction on pairs of nodes.
- Prediction of protein-protein association and type in a protein-protein association network.
Graph property prediction : Prediction on an entire graph/subgraph.
- Prediction of chemical properties of molecules (12 kinds of datasets available).

Installation

You can install OGB using Python's package manager pip. To avoid any conflict with your existing Python setup, it is suggested to work in a virtual environment with virtualenv. To install virtualenv:

pip install --upgrade virtualenv
virtualenv venv
source venv/bin/activate

Requirements

Python 3.7
PyTorch>=1.2
DGL>=0.4.1 or torch-geometric>=1.3.1
Numpy>=1.16.0
pandas>=0.24.0
urllib3>=1.24.0
scikit-learn>=0.20.0

Pip install

The recommended way to install OGB is using Python's package manager pip:

pip install ogb

From source

You can also install OGB from source. This is recommended if you want to contribute to OGB.

git clone https://github.com/snap-stanford/ogb
cd ogb
python setup.py install

Example

We highlight two key features of OGB, namely, (1) easy-to-use data loaders, and (2) standardized evaluators.

(1) Data loaders

We prepare easy-to-use PyTorch Geometric and DGL data loaders. We handle dataset downloading as well as standardized dataset splitting. Below, on PyTorch Geometric, we see that a few lines of code is sufficient to prepare and split the dataset! Needless to say, you can enjoy the same convenience for DGL!

from ogb.graphproppred.dataset_pyg import PygGraphPropPredDataset
from torch_geometric.data import DataLoader

dataset = PygGraphPropPredDataset(name = "ogbg-mol-tox21") 
splitted_idx = dataset.get_idx_split() 

train_loader = DataLoader(dataset[splitted_idx["train"]], batch_size=32, shuffle=True)
valid_loader = DataLoader(dataset[splitted_idx["valid"]], batch_size=32, shuffle=False)
test_loader = DataLoader(dataset[splitted_idx["test"]], batch_size=32, shuffle=False)

(2) Evaluators

We also prepare standardized evaluators for easy evaluation and comparison of different methods. The evaluator takes input_dict (a dictionary whose format is specified in evaluator.expected_input_format) as input, and returns a dictionary storing the performance metric appropriate for the given dataset. The standardized evaluation protocol allows researchers to reliably compare their methods.

from ogb.graphproppred import Evaluator

evaluator = Evaluator(name = "ogbg-mol-tox21")
# We can learn the input and output format specification of the evaluator as follows.
# print(evaluator.expected_input_format) 
# print(evaluator.expected_output_format) 
input_dict = {"y_true": y_true, "y_pred": y_pred}
result_dict = evaluator.eval(input_dict) # E.g., {"ap": 0.3421, "rocauc": 0.7321}

Citing OGB

Coming soon.

fgerzer/ogb