Daniel Voigt Godoy - Deep Learning with PyTorch Step-by-Step A Beginner’s Guide-leanpub

Notebook Cell 1.3 - Loading data: turning Numpy arrays into PyTorch tensors
1 device = 'cuda' if torch.cuda.is_available() else 'cpu'
2
3 # Our data was in Numpy arrays, but we need to transform them
4 # into PyTorch tensors and then send them to the
5 # chosen device
6 x_train_tensor = torch.as_tensor(x_train).float().to(device)
7 y_train_tensor = torch.as_tensor(y_train).float().to(device)
So, we defined a device, converted both Numpy arrays into PyTorch tensors, cast
them to floats, and sent them to the device. Let’s take a look at the types:
# Here we can see the difference - notice that .type() is more
# useful since it also tells us WHERE the tensor is (device)
print(type(x_train), type(x_train_tensor), x_train_tensor.type())
Output - GPU
<class 'numpy.ndarray'> <class 'torch.Tensor'>
torch.cuda.FloatTensor
Output - CPU
<class 'numpy.ndarray'> <class 'torch.Tensor'>
torch.FloatTensor
If you compare the types of both variables, you’ll get what you’d expect:
numpy.ndarray for the first one and torch.Tensor for the second one.
But where does the x_train_tensor "live"? Is it a CPU or a GPU tensor? You can’t
say, but if you use PyTorch’s type(), it will reveal its location
—torch.cuda.FloatTensor—a GPU tensor in this case (assuming the output using a
GPU, of course).
There is one more thing to be aware of when using GPU tensors. Remember
numpy()? What if we want to turn a GPU tensor back into a Numpy array? We’ll get
an error:
80 | Chapter 1: A Simple Regression Problem