Torch.float vs torch.FloatTensor

1

What is the difference between

a) torch.from_numpy(a).type(torch.FloatTensor)
b) torch.from_numpy(a).type(torch.float) ?

When I installed PyTorch via the command conda install pytorch torchvision -c pytorch, (b) works. When I installed PyTorch via the command conda install pytorch=0.1.12 cuda75 -c pytorch on another PC, (b) does not work but (a) works.

2

Hi,

torch.float has been added recently, it was not in old releases like 0.1.xx, this is why it does not work.