Why do I have to transform every data to cuda???

[apollo000104]

Hello, Everybody.
I am just studying pytorch with https://www.learnpytorch.io/02_pytorch_classification/.
And Here, I have met a problem. https://www.learnpytorch.io/02_pytorch_classification/#31-going-from-raw-model-outputs-to-predicted-labels-logits-prediction-probabilities-prediction-labels here, I have to move each data to avoid such error:
{ValueError: Target size (torch.Size([800, 1, 1])) must be the same as input size (torch.Size([800, 1]))}

If I delete any <.to(device)> part, I meet above error.
I have avoid it by writing like this
""
model_0.train()
y_pred=model_0(X_train.to(device))
X_test=X_test.unsqueeze(dim=1)
loss=loss_fn(y_pred.to(device), X_test.to(device))
optimizer.zero_grad()
loss.backward()
optimizer.step()
""""

But I don't know why I can't remove any of "to(device)" method.

I wonder if anybody explain me step by step.
Why I can't do that.
Thanks for reaching at here and Thanks again to try to help me.

[mrdbourke]

Hi @apollo000104 ,

Great question!

The reason is because you have to make computations all on the same device.

For example, if your model is on the CPU (default) but your data is on the GPU ("cuda"), you will get a device error.

Data and model must be on the same device (CPU or GPU).

By default, PyTorch saves everything to CPU, so in order to use a GPU, we use the .to(device) method to move them there.

However, in PyTorch 2.0, you can set a default device.

This helps make sure everything is on the same device (e.g. "cuda").

import torch

# Set the device
device = "cuda" if torch.cuda.is_available() else "cpu"

# Set the device globally
torch.set_default_device(device)

# All tensors created will be on the global device by default
layer = torch.nn.Linear(20, 30)
print(f"Layer weights are on device: {layer.weight.device}")
print(f"Layer creating data on device: {layer(torch.randn(128, 20)).device}")

Output:

Layer weights are on device: cuda:0
Layer creating data on device: cuda:0

See:

• https://www.learnpytorch.io/pytorch_2_intro/#11-globally-set-devices
• https://pytorch.org/docs/stable/generated/torch.set_default_device.html