Skip to content

[Optimization] Allow for subtypes of int for edge indexes #24

@GuillaumeBroggi

Description

@GuillaumeBroggi

elif edges_indexes.dtype != int:
raise RuntimeError('Edges indexes matrix is not a numpy.array '
'of dtype int.')

Current behavior enforces int type for edge indexes. Many graphs do not require storing so many indexes; using smaller footprint types can significantly decrease memory usage. For instance, we save 10 GB over ~60k graphs in the cylinder flow dataset when using in16.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions