This is a fast, minimal port of Boris Dayma's [DALL·E Mini](https://github.com/borisdayma/dalle-mini) (with mega weights). It has been stripped down for inference and converted to PyTorch. The only third party dependencies are numpy, requests, pillow and torch.
The required models will be downloaded to `models_root` if they are not already there. Set the `dtype` to `torch.float16` to save GPU memory. If you have an Ampere architecture GPU you can use `torch.bfloat16`. Once everything has finished initializing, call `generate_image` with some text as many times as you want. Use a positive `seed` for reproducible results. Higher values for `log2_supercondition_factor` result in better agreement with the text but a narrower variety of generated images. Every image token is sampled from the top-$k$ most probable tokens.
If the model is being used interactively (e.g. in a notebook) `generate_image_stream` can be used to generate a stream of images as the model is decoding. The detokenizer adds a slight delay for each image. Setting `log2_mid_count` to 3 results in a total of `2 ** 3 = 8` generated images. The only valid values for `log2_mid_count` are 0, 1, 2, 3, and 4. This is implemented in the colab.