Element-wise Mathematical Operations on Tensors

Math operation with scalar

If you have used the NumPy array before, you may already know that we can perform a math operation between a NumPy array and a scalar. PyTorch tensor supports almost the same operations. In PyTorch, we perform it in two ways:

• The operators, such as +, -, and *.
• The functions, such as add, sub, and mul.

a = torch.tensor([1,2,3])
b = a + 3
b = a.add(3)

PyTorch supports most of the math functions under the native Python math module. You could find a complete list from the official sites.

Notice: These functions wouldn’t change the original tensor, the changed tensor would be returned as a new tensor. PyTorch provides another version of these functions, which allows us to change the tensor in-place. These functions’ names remain the same, except that the function names are suffixed with an underscore. Such as add_.