Skip to content

Stable Diffusion from Scratch | A PyTorch-based implementation of the full Stable Diffusion pipeline

Notifications You must be signed in to change notification settings

torinriley/Diffusion

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Diffusion from Sratch

This project implements a text-to-image generation pipeline inspired by the Stable Diffusion architecture. The pipeline was built entirely from scratch in PyTorch. It integrates a Variational Autoencoder (VAE) for latent space compression, Denoising Diffusion Probabilistic Models (DDPM) for iterative denoising, and CLIP-based text embeddings for aligning text and images effectively.

• Custom Variational Autoencoder (VAE): Compresses images into latent representations for efficient generation.

• DDPM Sampling: Implements iterative denoising to generate high-quality images from noise.

• Text Embedding with CLIP: Ensures precise alignment of text and generated images.

Thanks to the following resources

Resource Description
Tokenizer Tokenizer files for Stable Diffusion.
Model Repository Main Hugging Face repository for Stable Diffusion v1.5.
Research Paper Original Stable Diffusion research paper.

About

Stable Diffusion from Scratch | A PyTorch-based implementation of the full Stable Diffusion pipeline

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages