ogb

See https://github.com/snap-stanford/ogb/pull/420

See https://github.com/snap-stanford/ogb/pull/368

This release introduces the following two:

• ogbl-vessel dataset (described here) @jqmcginnis
• Improved rank calculation for link prediction https://github.com/snap-stanford/ogb/pull/357 @mberr

We have included two updates:

• WikiKG90M --> WikiKG90Mv2
• PCQM4M --> PCQM4Mv2

Thanks to the DGL Team, all the LSC data is now hosted on AWS. This significantly improves the download speed around the globe! The underlying data stays exactly the same.

This release includes the three large-scale datasets for OGB-LSC at KDD Cup 2021. Details of the datasets and the KDD Cup can be found here.

The dataset downloading now uses http instead of https.

This version provides a major change in ogbg-code.

• ogbg-code has been deprecated due to prediction target (i.e., method name) leakage in input AST.
• ogbg-code2 has been introduced that fixes the issue., where the method name and its recursive definition in AST are replaced with a special token _mask_.

We sincerely thank Charles Sutton (@casutton) for finding the data leakage in our dataset.

This release fixes the dataset bug in negative samples in ogbl-wikikg and ogbl-citation and releases new versions of them: ogbl-wikikg2 and ogbl-citation2. The old versions are deprecated.

This release enhances the OGB package in the following ways.

• Made ogbn-papers100M data loading more tractable by using compressed binary files https://github.com/snap-stanford/ogb/issues/46
• Introduced DatasetSaver module for external contributors https://github.com/snap-stanford/ogb/issues/1
• Made dataset object compatible to DGL v0.5 (not backward compatible for heterogeneous graph datasets).

This release is mainly for changing the evaluation metric of ogbg-molpcba from PRC-AUC to Average Precision (AP). AP is shown to be more appropriate to summarize the non-convex nature of the Precision Recall Curve [1]. The leaderboard and our paper have been updated accordingly.

We also fix an issue and add a feature:

• Fixed an issue for saving a large library-agnostic data object. https://github.com/snap-stanford/ogb/issues/48
• Added automatic version check feature so that users will get notified when the package version is outdated.

[1] Jesse Davis and Mark Goadrich. The relationship between precision-recall and roc curves. InInternational Conference on Machine Learning (ICML), pp. 233–240, 2006.

This release fixes bugs in a dataset, evaluator, and data loader.

• Duplicated edges in ogbn-mag are removed. The updated dataset will be downloaded and processed automatically as you run your script for ogbn-mag. #40
• Evaluators for ogbl-collab and ogbl-ddi are updated. Specifically, ogbl-collab now uses Hits@50, and ogbl-ddi now uses Hits@20.
• DGL data loader bug for ogbn-mag and ogbl-biokg is fixed. #36

This is the second major release of OGB, in which we have curated many more exciting graph datasets, including heterogeneous graphs and a web-scale gigantic graph (100+ million nodes, 1+ billion edges).

First, we note that there is no change in the datasets released in version 1.1.1. Therefore, any experimental results obtained using 1.1.1 on the existing datasets are compatible to version 1.2.0.

In this new release, we have additionally released 5 new datasets listed below.

• ogbn-papers100M: Web-scale gigantic paper citation network.
• ogbn-mag: Heterogeneous academic graph.
• ogbl-biokg: Heterogeneous biomedical knowledge graph.
• ogbl-ddi: Drug-drug interaction network.
• ogbg-code: Source code Abstract Syntax Trees.

OGB package can now automatically fetch the datasets if they have been updated.

First Major Release

This is the first major release of OGB.
A number of changes have been made to the datasets, which are summarized below.

1. Re-indexed all the nodes in the node/link datasets (The graphs remain essentially the same).
2. In dataset folders for all the datasets, added mapping/ directory that contains information to map node/edge/graph/label indices to real-world entities (e.g., mapping from nodes in PPA to unique protein identifiers, mapping from molecular graphs into the SMILES strings.)
3. Deleted the ogbn-proteins node features, and put them in the species variable.
4. Deleted ogbl-reviews datasets.
5. Added 4 datasets: ogbn-arxiv, ogbl-citation, ogbl-collab, ogbl-wikikg.
6. Renamed ogbg-ppi to ogbg-ppa.
7. Renamed ogbg-mol-hiv and ogbg-mol-pcba to ogbg-molhiv and ogbg-molpcba, respectively.
8. Changed the evaluation metric of imbalanced molecule dataset (e.g., pcba) from ROC-AUC to PRC-AUC.
9. Changed the get_split_edge() interface in LinkPropPredDataset. The downloaded dataset files are also changed accordingly.
10. Added num_classes attribute for multi-class classification datasets.

Minor Changes

OGB datasets can now be imported more conveniently, e.g.:

from ogb.graphproppred import GraphPropPredDataset
from ogb.graphproppred import PygGraphPropPredDataset
from ogb.graphproppred import DglGraphPropPredDataset

Note that this will throw an ImportError if OGB can not find installations of Pyg or DGL, respectively.