RuntimeError: Expected object of device type cuda but got device type cpu for argument #2 'mat2' in call to _th_mm

[debinz]

Hi

I am using your great code! I have slightly modified your models.py to fit my regression task, but when I run the train.py, error occurred:

Traceback (most recent call last):
File "train.py", line 1273, in
main()
File "train.py", line 1250, in main
benchmark_task(prog_args, writer=writer)
File "train.py", line 1024, in benchmark_task
writer=writer,
File "train.py", line 190, in train
ypred, att_adj = model(h0, adj, batch_num_nodes, assign_x=assign_input) #use the 'forward' that is in model class
File "/home/omnisky/venv/lib/python3.7/site-packages/torch/nn/modules/module.py", line 541, in call
result = self.forward(*input, **kwargs)
File "/home/omnisky/Documents/individual_variance/gnn-model-explainer/models.py", line 570, in forward
self.assign_pred_modulesi
File "/home/omnisky/venv/lib/python3.7/site-packages/torch/nn/modules/module.py", line 541, in call
result = self.forward(*input, **kwargs)
File "/home/omnisky/venv/lib/python3.7/site-packages/torch/nn/modules/linear.py", line 87, in forward
return F.linear(input, self.weight, self.bias)
File "/home/omnisky/venv/lib/python3.7/site-packages/torch/nn/functional.py", line 1372, in linear
output = input.matmul(weight.t())
RuntimeError: Expected object of device type cuda but got device type cpu for argument #2 'mat2' in call to _th_mm

Can you give some advice about that? Thank you for your kind assistance!

Zeng

[iamzhangyunlu]

Hi, have you already solved the problem?

[RexYing]

Thanks for pointing out.
Since the line number doesn't correspond, wonder if you the argument method is set to "base"?
The default should be base.

I'll refactor out some unused code so that it doesn't accidentally call an unused model.

[jiank2]

I have the same problem. Probably, the reason is that prog_args.gpu is set to False, not sure why it is False since const is True and action is 'store_const' (at least this is what happened on my PC). My solution is kind of like brute force by comment out all if-condition that involves prog_args.gpu/self.gpu, which enforces the program to send all data into GPU.

I wonder if the authors could provide some hint on how to fix the weird prog_args.gpu value assignment.