Data Structure
Networking
RDBMS
Operating System
Java
MS Excel
iOS
HTML
CSS
Android
Python
C Programming
C++
C#
MongoDB
MySQL
Javascript
PHP
The backward() method is used to compute the gradient during the backward pass in a neural network.The gradients are computed when this method is executed.These gradients are stored in the respective variables.The gradients are computed with respect to these variables, and the gradients are accessed using .grad.If we do not call the backward() method for computing the gradient, the gradients are not computed.And, if we access the gradients using .grad, the result is None.Let's have a couple of examples to demonstrate how it works.Example 1In this example, we attempt to access the gradients without calling the backward() method. We notice ... Read More
A contiguous tensor is a tensor whose elements are stored in a contiguous order without leaving any empty space between them. A tensor created originally is always a contiguous tensor. A tensor can be viewed with different dimensions in contiguous manner.A transpose of a tensor creates a view of the original tensor which follows non-contiguous order. The transpose of a tensor is non-contiguous.SyntaxTensor.is_contiguous()It returns True if the Tensor is contiguous; False otherwise.Let's take a couple of example to demonstrate how to use this function to check if a tensor is contiguous or non-contiguous.Example 1# import torch library import torch ... Read More
To transpose a tensor, we need two dimensions to be transposed. If a tensor is 0-D or 1-D tensor, the transpose of the tensor is same as is. For a 2-D tensor, the transpose is computed using the two dimensions 0 and 1 as transpose(input, 0, 1).SyntaxTo find the transpose of a scalar, a vector or a matrix, we can apply the first syntax defined below.And for any dimensional tensor, we can apply the second syntax.For
The rank of a matrix can be obtained using torch.linalg.matrix_rank(). It takes a matrix or a batch of matrices as the input and returns a tensor with rank value(s) of the matrices. torch.linalg module provides us many linear algebra operations.Syntaxtorch.linalg.matrix_rank(input)where input is the 2D tensor/matrix or batch of matrices.StepsWe could use the following steps to get the rank of a matrix or batch of matrices −Import the torch library. Make sure you have it already installed.import torch Create a 2D tensor/matrix or a batch of matrices and print it.t = torch.tensor([[1., 2., 3.], [4., 5., 6.]]) print("Tensor:", t)Compute the rank ... Read More
To find the exponential of the elements of an input tensor, we can apply Tensor.exp() or torch.exp(input). Here, input is the input tensor for which the exponentials are computed. Both the methods return a new tensor with the exponential values of the elements of the input tensor.SyntaxTensor.exp()ortorch.exp(input) StepsWe could use the following steps to compute the exponentials of the elements of an input tensor −Import the torch library. Make sure you have it already installed.import torchCreate a tensor and print it.t1 = torch.rand(4, 3) print("Tensor:", t1)Compute the exponential of the elements of the tensor. For this, use torch.exp(input) and optionally ... Read More
We use the torch.log2() method to compute logarithm to the base 2 of the elements of a tensor. It returns a new tensor with the logarithm values of the elements of the original input tensor. It takes a tensor as the input parameter and outputs a tensor.Syntaxtorch.log2(input)where input is a PyTorch tensor.It returns a new tensor with logarithm base 2 values.StepsImport the torch library. Make sure you have it already installed.import torch Create a tensor and print it.tensor1 = torch.rand(5, 3) print("Tensor:", tensor1)Compute torch.log2(input) and optionally assign this value to a new variable. Here, input is the created tensor.logb2 = ... Read More
Tensor.detach() is used to detach a tensor from the current computational graph. It returns a new tensor that doesn't require a gradient.When we don't need a tensor to be traced for the gradient computation, we detach the tensor from the current computational graph.We also need to detach a tensor when we need to move the tensor from GPU to CPU.SyntaxTensor.detach()It returns a new tensor without requires_grad = True. The gradient with respect to this tensor will no longer be computed.StepsImport the torch library. Make sure you have it already installed.import torch Create a PyTorch tensor with requires_grad = True and ... Read More
To compute the gradients, a tensor must have its parameter requires_grad = true. The gradients are same as the partial derivatives.For example, in the function y = 2*x + 1, x is a tensor with requires_grad = True. We can compute the gradients using y.backward() function and the gradient can be accessed using x.grad.Here, the value of x.gad is same as the partial derivative of y with respect to x. If the tensor x is without requires_grad, then the gradient is None. We can define a function of multiple variables. Here the variables are the PyTorch tensors.StepsWe can use the ... Read More
torch.logical_xor() computes the element-wise logical XOR of the given two input tensors. In a tensor, the elements with zero values are treated as False and non-zero elements are treated as True. It takes two tensors as input parameters and returns a tensor with values after computing the logical XOR.Syntaxtorch.logical_xor(tensor1, tensor2)where tensor1 and tensor2 are the two input tensors.StepsTo compute element-wise logical XOR of given input tensors, one could follow the steps given below −Import the torch library. Make sure you have it already installed.Create two tensors, tensor1 and tensor2, and print the tensors.Compute torch.logical_xor(tesnor1, tesnor2) and assign the value to ... Read More
torch.narrow() method is used to perform narrow operation on a PyTorch tensor. It returns a new tensor that is a narrowed version of the original input tensor.For example, a tensor of [4, 3] can be narrowed to a tensor of size [2, 3] or [4, 2]. We can narrow down a tensor along a single dimension at a time. Here, we cannot narrow down both dimensions to a size of [2, 2]. We can also use Tensor.narrow() to narrow down a tensor.Syntaxtorch.narrow(input, dim, start, length) Tensor.narrow(dim, start, length)Parametersinput – It's the PyTorch tensor to narrow.dim – It's the dimension along ... Read More