PyTorch Tensors Function That Will Save Time

PyTorch is an open-source library for Python. It is a machine learning library. A tensor is scalar, vector, matrix, or n-dimensional data container which is similar to NumPy’s ndarray. Lets see 5 PyTorch tensors. They are as follows -

torch.where(condition, a, b)
torch.unbind(input, dim=0)
torch.roll(input, shifts, dims=None)
torch.tril(input, diagonal=0, *, out=None)
select(dim, index)

Before we begin, let's install and import PyTorch

import torch

Function 1 - torch.where(condition, x, y)

This returns a tensor of elements selected from either x or y, depending on condition. The operation is defined as:

Example - 1

a1 = torch.randn(3, 3)
b1 = torch.ones([3, 3], dtype = torch.float)
print('Tensor a1:\n', a1)
print('Tensor b1:\n', b1)

Tensor a1:
 tensor([[-0.4091,  0.6474, -0.5680],
        [-0.3016,  1.7412,  0.6511],
        [ 0.3040,  0.3834, -0.4173]])
Tensor b1:
 tensor([[1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.]])