Broadcasting in PyTorch #shorts

PyTorch is a popular open-source machine learning library used for various tasks like deep learning, neural networks, and more. One important feature of PyTorch is broadcasting, which allows for efficient element-wise operations on tensors of different shapes.

When performing operations on tensors of different shapes, PyTorch automatically broadcasts the smaller tensor to match the shape of the larger tensor. This allows for easier and more concise coding when working with tensors in PyTorch.

For example, let’s say we have two tensors, A and B, with shapes (3, 1) and (1, 3) respectively. When we perform an operation like addition between these two tensors, PyTorch will automatically broadcast the smaller tensor to match the shape of the larger tensor before performing the operation.

import torch

A = torch.tensor([[1], [2], [3]])
B = torch.tensor([[4, 5, 6]])

result = A + B
print(result)

In the above example, PyTorch will broadcast the tensor A to have shape (3, 3) before performing the addition operation with tensor B. This makes it easier to perform element-wise operations on tensors of different shapes without having to worry about manually aligning them.

Overall, broadcasting in PyTorch is a powerful feature that simplifies working with tensors of different shapes, making it easier to perform complex operations in an efficient manner.