Fundamental PyTorch operations

A very brief exploration of some of the basic but important PyTorch operations

PyTorch is a machine learning framework that is very intuitive to grasp once one has a strong understanding of the basic concepts. Although PyTorch has a plethora of operations, this notebook will strive to clarify the very basics.

• torch.tensor
• empty
• normal_
• permute
• stack

Function 1 - torch.tensor

At first sight this might seem to be a very simple operation, but in practise an incomplete understanding of the operation can lead to wastage of memory when dealing with very deep neural networks.

torch.tensor will create a tensor with any data that is passed to it. But remember that this operation will always create a copy of the data that is passed to it. Therefore:

1. If your data is a tensor and you want to avoid making a copy of it, use torch.as_tensor
2. If your data is a numpy array, use either torch.as_tensor or torch.from_numpy to prevent a copy from being created.