DALL-E  by openai

PyTorch package for DALL-E's discrete VAE

created 4 years ago
10,874 stars

Top 4.8% on sourcepulse

GitHubView on GitHub
Project Summary

This repository provides the PyTorch implementation of the discrete Variational Autoencoder (VAE) component used in OpenAI's DALL-E model. It enables researchers and developers to leverage the VAE for generating image tokens, a crucial step in the DALL-E pipeline, facilitating experimentation with image generation architectures.

How It Works

The package implements a discrete VAE, a generative model that learns a compressed, discrete latent representation of images. This VAE is trained to reconstruct images from these discrete latent codes, effectively learning a "visual vocabulary" of image tokens. This approach allows for efficient and high-quality generation of image components that can then be sequenced by a transformer model.

Quick Start & Requirements

  • Install via pip: pip install DALL-E
  • Requires PyTorch.
  • See [Usage] for examples.

Highlighted Details

  • Official PyTorch package for the DALL-E VAE.
  • Focuses on the discrete VAE component, not the transformer.

Maintenance & Community

  • Maintained by OpenAI.
  • No community links (Discord/Slack) or roadmap provided in the README.

Licensing & Compatibility

  • License not specified in the README.
  • Compatibility for commercial use or closed-source linking is undetermined.

Limitations & Caveats

This release only includes the VAE component; the transformer model for text-to-image generation is not provided. The README does not specify the license, which may impact commercial use.

Health Check
Last commit

1 year ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
0
Star History
50 stars in the last 90 days

Explore Similar Projects

Starred by Tobi Lutke Tobi Lutke(Cofounder of Shopify) and Patrick von Platen Patrick von Platen(Core Contributor to Hugging Face Transformers and Diffusers).

taesd by madebyollin

0.5%
758
Tiny AutoEncoder for Stable Diffusion latents
created 2 years ago
updated 3 months ago
Feedback? Help us improve.