search

HennyJie / BrainGB

Officially Accepted to IEEE Transactions on Medical Imaging (TMI, IF: 11.037) - Special Issue on Geometric Deep Learning in Medical Imaging.
MIT License
164 stars 42 forks source link

GAT runtime error #29

Open salmon1802 opened 1 month ago

salmon1802 commented 1 month ago

Thank you very much for your meaningful work! However, when I change the GCN to GAT in the config file, the following error is reported, I made attempts but still did not run successfully.

Below is my error log:

Processing... Done! seed for seed_everything(): 2024 2024-05-25 23:48:45,472 - Loaded dataset: ABIDE 0%| | 0/26 [00:00<?, ?it/s]edge_index shape: torch.Size([2, 1280000]) x shape: torch.Size([6400, 200]) Maximum index in edge_index: 6399 edge_attr: torch.Size([1280000]) 0%| | 0/26 [00:00<?, ?it/s] Traceback (most recent call last): File "D:/GraduateStudent/code/BrainGB-master/examples/example_main.py", line 152, in main(parser.parse_args()) File "D:/GraduateStudent/code/BrainGB-master/examples/example_main.py", line 78, in main test_micro, test_auc, test_macro = train_and_evaluate(model, train_loader, test_loader, File "D:\GraduateStudent\code\BrainGB-master\examples\train_and_evaluate.py", line 26, in train_and_evaluate out = model(data) File "D:\zj\miniconda\envs\Dual-HINet\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl return forward_call(*input, kwargs) File "D:\GraduateStudent\code\BrainGB-master\src\models\brainnn.py", line 18, in forward g = self.gnn(x, edge_index, edge_attr, batch) File "D:\zj\miniconda\envs\Dual-HINet\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl return forward_call(*input, *kwargs) File "D:\GraduateStudent\code\BrainGB-master\src\models\gat.py", line 147, in forward z = conv(z, edge_index, edge_attr) File "D:\zj\miniconda\envs\Dual-HINet\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl return forward_call(input, kwargs) File "C:\Users\28334\AppData\Local\Temp\torch_geometric.nn.sequential_f0afbb_syfjdr4d.py", line 23, in forward x = self.module_0(x, edge_index, edge_attr) File "D:\zj\miniconda\envs\Dual-HINet\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl return forward_call(*input, **kwargs) File "D:\zj\miniconda\envs\Dual-HINet\lib\site-packages\torch_geometric\nn\conv\gat_conv.py", line 341, in forward out = self.propagate(edge_index, x=x, alpha=alpha, size=size) File "D:\zj\miniconda\envs\Dual-HINet\lib\site-packages\torch_geometric\nn\conv\message_passing.py", line 538, in propagate coll_dict = self._collect(self._user_args, edge_index, File "D:\zj\miniconda\envs\Dual-HINet\lib\site-packages\torch_geometric\nn\conv\message_passing.py", line 399, in _collect self._set_size(size, dim, data) File "D:\zj\miniconda\envs\Dual-HINet\lib\site-packages\torch_geometric\nn\conv\message_passing.py", line 298, in _set_size raise ValueError( ValueError: Encountered tensor with size 1280000 in dimension 0, but expected size 6400.

Process finished with exit code 1

Gemaxis commented 3 weeks ago

It's due to your environment not matching the version. I faced the same problem, and I changed my Torch version to fix it. Btw, if you are using Windows, you can select torch=1.8.1+cu102 with the corresponding torch_geometric version.