kuprel d4af693133 updated colab to load model once and generate multiple times		2 years ago
examples	simplified	2 years ago
min_dalle	refactored to load models once and run multiple times	2 years ago
.gitignore	examples	2 years ago
LICENSE	license and cleanup	2 years ago
README.md	refactored to load models once and run multiple times	2 years ago
image_from_text.py	refactored to load models once and run multiple times	2 years ago
min_dalle.ipynb	updated colab to load model once and generate multiple times	2 years ago
requirements.txt	Simplified requirements:	2 years ago
setup.sh	Update setup.sh	2 years ago

README.md

min(DALL·E)

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 numpy, torch, and flax.

Setup

Run sh setup.sh to install dependencies and download pretrained models. The wandb python package is installed to download DALL·E mini and DALL·E mega. Alternatively, the models can be downloaded manually here: VQGan, DALL·E Mini, DALL·E Mega

Usage

The simplest way to get started is the command line python script image_from_text.py provided. Here are some examples runs:

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

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

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

Load once run multiple times

The command line script loads the models and parameters each time. The colab notebook demonstrates how to load the models once and run multiple times.