Can not send dataset to GPU

data
1

Hello,

I am a newbies for ML. I started training MNIST dataset. I can send my model to GPU by using this code.

network.to(device) # Returned… true

When I sent the dataset to GPU by this testing code.

for batch_idx, (data, target) in enumerate(train_loader):
data.to(device)
target.to(device)
if batch_idx == 100:
print(data.device)
print(type(data))
pass

#Returned … cpu
#Returned … <class ‘torch.Tensor’>

Question: Why does “data.to(device)” still be on CPU?
Note: I use RTX30xx

2

The to() operation is not applied inplace on tensors (unlike on modules as it’s recursively applied on all submodules, parameters, and buffers internally), so you need to reassign the tensor:

x = x.to(...)
1 Like
3

Thank you very much , I got it :slight_smile: solved