GPT2-Large For Writing Stable Diffusion Prompts

This is a GPT2-Large model fine-tuned for rewriting prompts for Stable Diffusion v1.5 through RLHF. The model takes an image description and writes a prompt resulting in aestheticcally pleasing images.

How To Use a Model

Write an image description and add a special SEP token: </s>, the model writes the prompt after it. Example:

a cat</s>