Compatibility of tensor types

1

Good afternoon! I have a question why when executing the following code

import torch
a = torch.tensor([1, 2, 3], dtype = torch.float32)
b = torch.tensor([1, 2, 3], dtype = torch.int)
print(a + b)

I get an error, but when I run other code

import torch
a = torch.tensor([1, 2, 3], dtype = torch.float32)
print(a + int(1))

there is no error and the code works correctly?
I noticed that you can only add tensors of one type. Why is this necessary?

2

It isn’t and indeed PyTorch 1.3 introduced type promotion and make the first work, too.

3

Thank you very much! I’m using PyTorch 0.4 and it’s not there. I will be updated.