V2.2 640 API Documentation

Improved the dataset further, reinforced many of the rarer characters and concepts, now at just under 50k images

The "Training data" file contains a folder of examples of each usable artist tag, and a list of all tags in the dataset with their frequencies