L2P-GNN pre-training GNN architecture on chem dataset

pre-training GNN architecture on chem dataset

Open wubo2180 opened this issue 3 years ago • 3 comments

Hi, do you have the pre-trained GNN model file on chem dataset

Jan 08 '22 08:01 wubo2180

Hi, we conducted experiments on the Chem dataset, but we didn't save the results and the pre-trained GNN model. If you are interested in the performance on the Chem dataset, you can directly run the model on it.

Thx.

Jan 08 '22 08:01 rootlu

Thanks. However, there are some mistakes when the GNN type is gcn, graphsage or gat (gin is OK). Error:

/model_chem.py , line 376, in forward

self_loop_attr[:, 7] = 1  # for bio dataset
IndexError: index 7 is out of bounds for dimension 1 with size 2

Should I comment out this line of code when I conduct experiment on chem dataset?

Jan 08 '22 13:01 wubo2180

By the way, there are some mistaks in util.py `L2P-GNN/code/util.py", line 31, in call

i, _, k = structured_negative_sampling(data.edge_index) NameError: name structured_negative_sampling is not defined`

Jan 08 '22 13:01 wubo2180

L2P-GNN L2P-GNN copied to clipboard

pre-training GNN architecture on chem dataset

L2P-GNN
L2P-GNN copied to clipboard