PyTorch is a Python library that is an open source machine learning framework. It provides automatic differentiation for building and training neural networks, as well as an n-dimensional array data structure like NumPy, except that it can be run on GPUs to produce dynamic computation graphs and gradients, as well as a generic tool for scientific computing. This is what's known as a Tensor.

To install the PyTorch library, you must have a version of Python 3.x or above to run it. To install onto your version of Python from command line using the following command:

                        pip3 install pytorch torchvision torchaudio

To import the PyTorch library into your scripts, use the following code to have access to it as a reference:


Source: TutorialsPoint - PyTorch Quick Guide