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.

How much mini-er can it get from here? 🤔