Run the papers100M dataset

[ZHUZHU-Whu]

Hello, the work is brilliant. It is stated in the data/OGBData file ReadMe.txt that papers100M dataset is not supported. Could you please provide the data processing interface for running papers100M? Looking forward to your reply. Thank you very much

[AiX-im]

Thank you for your interest in our work.

NeutronOrch needs 4 dataset files to run:

• your_dateset.edge，a binary edge list file, used to store the graph structure.
• your_dateset.feat，contains the feature of each node, the first number in each line indicates the node number, followed by the feature of the node.
• your_dateset.label，contains the label of each node, the first number in each line indicates the node number, followed by the classification number of the node.
• your_dateset.mask, contains the mask of each node, the first number in each line indicates the node number, followed by the mask of node (train, val, test).

We provide a python script to convert some commonly used datasets, please refer to data/generate_nts_dataset.py for details. This script supports data preprocessing for Papers100M.

If you have any other questions please let us know.