DiT Diffusion Transformer for Face Generation This project implements a Diffusion Transformer (DiT) model trained on Celeba latents Generated Samples The model was trained for 15 epochs using a DiT architecture with latent diffusion on compressed VAE representations of Celeba images.
