Datasets update

[stassius]

New:
SD Dataset Preparation node. Streamlines a dataset preparation task. It copies images from the source folder, crops and resizes them, adds text files with captions. Optionally you can add a prefix to the output folder (for example for Kohya_ss it would be the number of iterations). You can dive inside the node and insert your own COP nodes between Input and Output to process your images in an unique way.

bandicam.2023-05-30.15-57-18-804.mp4

Improvements:
SD Interrogate node: Ability to turn off artist-flavor-style keywords to create captions for training datasets (BLIP-like). Ability to save the caption to a text file next to the image on disk. This node is used in the SD Dataset Preparation node.

SD Set Option node: Now it includes a set of predefined options with their own types and ranges. It also reads the current option value. Note that the changed value will be saved to Automatic config file and will remain until the next change. Let me know if you want to add a particular Automatic1111 option here.

bandicam.2023-05-30.15-59-11-989.mp4

Fixes:

• Renamed settings.py to avoid possible conflicts
• Fixed a possible bug in Cop processor
• Minor bugfixes