baofff/U-ViT

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1110 stars · 77 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

U-ViT is the official PyTorch implementation of a ViT-based backbone architecture for diffusion models from the CVPR 2023 paper 'All are Worth Words: A ViT Backbone for Diffusion Models'. It provides model code, pretrained checkpoints, and training/evaluation scripts for pixel-space, latent-space, and text-to-image diffusion models.

Use cases

• train a diffusion model with a vision transformer backbone instead of a CNN U-Net
• generate images unconditionally or class-conditionally on CIFAR10, CelebA, or ImageNet
• build a text-to-image diffusion model
• run latent diffusion models with a transformer backbone
• research long skip connections in transformer architectures for generative modeling
• fine-tune pretrained U-ViT checkpoints on custom image generation tasks

When to choose

• you need a transformer-based backbone for diffusion models with pretrained weights
• you want to reproduce or extend CVPR 2023 diffusion research
• you need training and evaluation scripts for pixel-space, latent, or text-to-image diffusion in PyTorch

When to avoid

• you need a production-ready image generation service rather than research code
• you require a CNN-based U-Net backbone for diffusion models
• you lack GPU resources, since training diffusion models is compute-intensive

Facets

library · maturity stable

deep-learning machine-learning image-processing deep-learning machine-learning image-processing artificial-intelligence python vision-transformer diffusion-models image-generation pytorch research-code text-to-image cvpr-2023

Member repositories

For agents

markdown · JSON · MCP: product_card(name="baofff/U-ViT")