🗂 Graph Learning Indexer: a contributor-friendly and metadata-rich platform for graph learning benchmarks. Dataloading, Benchmarking, Tagging, and more!
When saving datatset splits like train_set, we now use dtype of np.int64 for an consistent behavior with torch, as torch indexing requires long tensor for some version.
Description
When saving datatset splits like
train_set, we now use dtype of np.int64 for an consistent behavior with torch, as torch indexing requires long tensor for some version.Related Issue
This pr fixes #474
Motivation and Context
This pr is motivated by the issue found in #472.
How Has This Been Tested?
See automated tests.