GM all. Today we’re back with another [virtual] arXiv Dive where we’ll learn

How to Train an LLM with Diffusion

. We spent a hunk of time implementing the paper and putting together a great review. Join us for an exciting practical dive into Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution! Today we will be: 1. Running pre-trained diffusion models for text 2. Training our own character level diffusion model 3. Discussing practical benefits and limitations within the code https://lu.ma/oxen