EleutherAI/DALLE-mtf
Implementation of OpenAI's DALL-E text-to-image model for large-scale distributed training using Mesh-TensorFlow on TPUs.
Not currently ranked — collecting fresh signals.
star history
This repository is an open implementation of OpenAI’s DALL-E model, designed for large-scale training across distributed TPU infrastructure. The system uses a variational autoencoder to compress images into discrete tokens, combined with an autoregressive transformer for text-to-image generation. It leverages Mesh-TensorFlow to enable training at scales comparable to or larger than the original 12-billion parameter DALL-E model.
Frequently asked
- What is EleutherAI/DALLE-mtf?
- Implementation of OpenAI's DALL-E text-to-image model for large-scale distributed training using Mesh-TensorFlow on TPUs.
- Is DALLE-mtf open source?
- Yes — EleutherAI/DALLE-mtf is open source, released under the MIT license.
- What language is DALLE-mtf written in?
- EleutherAI/DALLE-mtf is primarily written in Python.
- How popular is DALLE-mtf?
- EleutherAI/DALLE-mtf has 431 stars on GitHub.
- Where can I find DALLE-mtf?
- EleutherAI/DALLE-mtf is on GitHub at https://github.com/EleutherAI/DALLE-mtf.