An error when finetuning on graph classification dataset about gcc HOT 3 CLOSED

I am wondering whether this is because the super().init() of the GraphClassificationDataset is missed (see gcc/datasets/graph_dataset.py line 312).

from gcc.

Hi @Kqiii ,

Thanks for pointing out. This is a bug created during us refactoring the code. Please change the following to fix it:

Change

to

self.dataset = data_util.create_graph_classification_dataset(dataset)
self.graphs = self.dataset.graph_lists

This should fix the problem. We're really sorry for the trouble. Please kindly follow up the results if possible.

from gcc.

Hi @Kqiii ,

Thanks for pointing out. This is a bug created during us refactoring the code. Please change the following to fix it:

Change

GCC/gcc/datasets/graph_dataset.py

Line 330 in dec36b7

self.graphs = data_util.create_graph_classification_dataset(dataset).graph_lists

to

self.dataset = data_util.create_graph_classification_dataset(dataset)
self.graphs = self.dataset.graph_lists

This should fix the problem. We're really sorry for the trouble. Please kindly follow up the results if possible.

Hi,

Thank you for your prompt reply. This works well for me. Btw, your GCC work is really informative and insightful. Actually, I have read many papers on self-supervised graph representation learning recently, and I think GCC is definitely among the best.

Thanks!

from gcc.