How do the text2image Models Work?

You probably have heard about Dalle-2, Midjourney, and Stable Diffusion. These are models that generate images from text. Recently, they have advanced to the point where artists are protesting by putting pictures with calls to ban AI in stock images, and recently even a lawsuit was filed. In this article, we will be discussing how these models work. We will start with the basics and gradually add details and techniques for a generation....

January 17, 2023 Â· 8 min Â· Anton Lebedev

How the PyTorch convolutions work or how to collapse two convolutions into one

This article has two goals. The first one is to show how two PyTorch convolutions can be combined into one. If you want the realisation, scroll through the article to the end. The second, the main goal is to take a closer look into the convolution realisation by the convolution merge example. We will learn how they are stored and used particularly in PyTorch without hardcore details like im2col. Before we start looking at the implementation, let’s remember what we are working with....

June 12, 2022 Â· 10 min Â· Anton Lebedev