Yuning You

Hi @HeyMercer, Thank you for your comments. Checkpoints can be very easily generated following the scripts https://github.com/Shen-Lab/GraphCL_Automated/tree/master/transferLearning_MoleculeNet_PPI#joao-pre-training, with an estimated time < 2 days. Feel free to pin me if...

Hi @leesoohong, Really sry for the late reply! I think this might result from the library version issue that when you call `dataset.get_num_feature()`, it does not give you `num_nodes` key....

Hi @agave233, Which experiment do you refer as the zinc_standard_agent dataset?

Hi @cjfcsjt, Sorry for the late reply. In paper we did solve it by a naive bi-level optimization if referring to section 3.2.

Hi @LA11131110128, It looks like the error comes from the mismatch between GNN and your customized data (though I am not clear where exactly it is). I would suggest to...

Hi @Struggle-Forever, Please refer to an instantiation of the subgraph augmentation https://github.com/Shen-Lab/GraphCL/issues/24#issuecomment-839406281. We would drop the non-selected nodes via, e.g. `data.x = data.x[idx_nondrop]` where the pooling function is operated on.

Hi @ZhaoYuTJPU, Sorry for the late reply. As I recall, HPO for unsupervised_TU should be learning rate in https://github.com/Shen-Lab/GraphCL/blob/master/unsupervised_TU/go.sh from {0.01, 0.001, 0.0001}.

Hi @Austinzhenghua, Thanks for your interest. For semi_TU experiments, if you fail to work with torch_geometric==1.4.0 which is out-of-dated, you can take a try an upgraded version https://github.com/Shen-Lab/GraphCL_Automated/tree/master/semisupervised_TU that should...

Yes, you can construct ur dataset by KNN.

Looks like either work for me.