search

Fenglei104 / DeepPROTACs

GNU General Public License v3.0
48 stars 16 forks source link

single_prediction error #9

Closed Gordon5-AI closed 1 year ago

Gordon5-AI commented 1 year ago

Great Job! but when I run this command for prediction

python single_prediction.py single_test/

I encounter a error like this:

AttributeError: 'GCNConv' object has no attribute '_explain'

I think it may be something wrong with the torch_geometric, such as the version, but my conda environment was created according to your supplied env.yaml. Another question,is the model named test.pt you supplied in this repositories genearted by running the commad:

python main.py

I running the command, a model was generated, but I have compared the test.pt model with mine, they are different. The processed feature under the folder data just a toy dataset, yes? Looking forward to your reply! Best Regrads.

Fenglei104 commented 1 year ago

Hi, thanks for your feedback.

I met the same problem of the torch geometric version some days ago, but it was in Windows and with no GPUs, while I prepared the yaml file in Ubuntu 18.04 with an RTX 3090. So it might be the same question. You can try the following versions:

pytorch=1.8.1=py3.7_cpu_0
torch-geometric==2.0.0
torch-scatter==2.0.9
torch-sparse==0.6.12

The test.pt is trained with the data described in the paper, while here we only provide 3 examples. And I'm really sorry that the data is in use and I can not publish it now.

Gordon5-AI commented 1 year ago

Thanks for your reply, I changed the version of torch-geometric from 2.0.4 to 2.0.0, the problem was solved. Best Regrads.