search

snap-stanford / GraphGym

Platform for designing and evaluating Graph Neural Networks (GNN)
Other
1.69k stars 185 forks source link

Documentation for configuration options and dataset registration. #52

Open Sann5 opened 1 year ago

Sann5 commented 1 year ago

Hello!

This project is truly amazing, thank you. That said I'm finding it difficult to apply it to my own datasets. Naturally, I would like to customize the grid search however, I'm not sure what the valid options are for each field in the configuration. The valid options I know are thanks to the examples configs and grids in the repo, but a comprehensive list for each field would be greatly appreciated. Is there any existing documentation on this matter?

I'm also unsure about how to register my datasets. At which point in the pipeline should the customized version of graphgym/contrib/loader/example.py be run? I'm guessing before the config generation script as the configs must include the dataset information. Still, I'm unsure about how this piece of code fits in the pipeline.

Thank you in advance.

HeVLF commented 1 year ago

Hey! Have you discovered how to register a dataset?

Sann5 commented 1 year ago

Hey kind of yes. But in the end i had so much trouble using graph gym that i replaced altogether with my own implementation. But if you want i can give you the link to my fork of the torch_geometric graphgym that tries to use it.

HeVLF commented 1 year ago

Hey. I'm thinking of doing the same, if the troubles persist. But if you can, please send me. Thank you!

Sann5 commented 1 year ago

https://github.com/Sann5/pytorch_geometric/tree/costum_dataset_bug

HeVLF commented 1 year ago

Thank you. Hopefully I can use it for my study

JanineCHEN commented 1 year ago

I fully agree that a comprehensive documentation is really in need for customization. I wonder if this project is still maintained? @JiaxuanYou

ChenpengZhang commented 3 months ago

I personally will recommend looking into the graphgym/loader.py file. In the file (specifically the load_pyg, load_nx function) the code is sort of self-explanatory. Looks like the model only supports (PPI, Amazon, Coauthor, KarateClub, MNISTSuperpixels, Planetoid, QM7b, TUDataset) and customized networkx datasets.