Pytorch Tutorials

This Pytorch tutorial will teach you about Pytorch Tensor, functions, and concepts about broadcasting, reshaping and generating random numbers

Pytorch Introduction. Pytorch is an open-source Python library that is used for building and training deep learning models by Facebook. This tutorial will help you go from the very scratch to a sufficiently advanced level. The only pre-requisite for this course is the knowledge of python. However, any prior knowledge about Pytorch and Machine Learning or Deep Learning is quite helpful. This chapter will talk about how to check if you already have pytorch installed or how to install pytorch if you do not have it installed already.

Pytorch Tensor. A Pytorch Tensor is the most basic data structures in the Pytorch library. Tensor. Creating a Tensor. Tensor Types. Tensor is nothing but a n-dimesional matrix that stores data. The tensor object of pytorch can have multiple types. For example, the tensor can store integers of floating point numbers. These various types of data also require different amounts of memory, which creates different tensors in memory. This chapter discusses about tensor, and the various types of tensor that can be created in pytorch.

Pytorch Tensor Indexing. Indexing in Pytorch is similar to that of numpy. The indexing is 0 based, i.e, the first element has 0 index. Indexing allows us to access elements from a tensor. Using indexing we can change the elements in the tensor. Indexing in pytorch is similar to numpy.

Pytorch Tensor Arithmetic Operations. Pytorch allows you to perform element-wise arithmetic operations using functions and operands. The interesting thing is that Pytorch provides both functions and operands for performing arithmetic operations. For example, to perform an addition operation, you can use the + operand as well as the add() function of pytorch. Similarly we have both operands and functions for subtraction, multiplication, division, remainder, exponentiation, etc.

Pytorch Broadcasting. Broadcasting allows the performance certain operations on tensors that are not of the same size. Broadcasting functionality of pytorch is borrowed from numpy. Broadcasting allows the performing of arithmetic operations on tensors which are not of the same size. This is done by stretching the smaller tensor to the size of the larger of larger tensor. Although this operation of stretching the smaller tensor to the size of the larger tensor is not always possible. Therefore, broadcasting should follow some rules.

Pytorch Math Functions. The Pytorch library has many inbuilt functions which can be used to perform specific Mathematical Operations.

Pytorch Reduction Functions. Pytorch library has many inbuilt functions for performing specific reduction operations on tensors. Reduction is the process in which a series of values is reduced to one value. This one value is descriptive of the entire series of values. Reduction functions include, but are not limited to mean, median, mode, standard deviation, variance, summation, product, maximum and minimum. This chapter discusses the various reduction functions that are available in pytorch.

Pytorch Matrix Functions. Pytorch has many inbuilt functions for operations on matrix. A handful of those will be covered in this chapter. A matrix is a special case of tensor which has a rank 2. The pytorch library provides function to perform operations on matrices. These include operations such as finding the transpose of a matrix, determinant, finding the vector product and dot product of matrices, finding the inverse of a matrix, etc.

Pytorch Random. Pytorch allows us to generate tensors with random values. The Random module is used for generating random numbers.
This chapter of the Pytorch tutorial will take you through the creation of random numbers in pytorch. It will also discuss in brief about the Pytorch Random generator and how to change the seed.