This is a minimal implementation of Boris Dayma's DALL·E Mini in flax. It has been stripped to the bare essentials necessary for doing inference. This repository also contains code for converting the flax model to torch.
See min(DALL·E) for the PyTorch version.
Run sh setup.sh to install dependencies and download pretrained models. The flax models can be manually downloaded here:
VQGan,
DALL·E Mini,
DALL·E Mega
Use the python script image_from_text.py to generate images from the command line. Note: the command line script loads the models and parameters each time. To load a model once and generate multiple times, initialize MinDalleFlax, then call generate_image with some text and a seed.