Skip to content

Dynamically complete the features for nodes in the graph during training #776

Open
@OysterQAQ

Description

@OysterQAQ

I plan to use graph convolution for constructing item embeddings in a recommendation system. When generating the dataset, I found that attaching all features of the items to the nodes would result in an excessively large dataset, causing IO bottlenecks and low GPU utilization. Is there a way to generate a dataset only containing node IDs and dynamically complete the features for nodes in the graph during training?

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