/min-dalle

Primary LanguageJupyter NotebookMIT LicenseMIT

min(DALL·E)

Open In Colab

This is a minimal implementation of DALL·E Mini. It has been stripped to the bare essentials necessary for doing inference, and converted to PyTorch. The only third party dependencies are torch for the torch model and flax for the flax model.

Setup

Run sh setup.sh to install dependencies and download pretrained models. In the bash script, Git LFS is used to download the VQGan detokenizer from Hugging Face and the Weight & Biases python package is used to download the DALL·E Mini and DALL·E Mega transformer models. These models can also be downloaded manually: VQGan, DALL·E Mini, DALL·E Mega

Usage

Use the command line python script image_from_text.py to generate images. Here are some examples:

python image_from_text.py --text='alien life' --seed=7

Alien

python image_from_text.py --text='a comfy chair that looks like an avocado' --mega --seed=4

Avocado Armchair

python image_from_text.py --text='court sketch of godzilla on trial' --mega --seed=100

Godzilla Trial