"Training day" update

[stassius]

New:
SD Trainer Kohya node.
Train LoRAs right from Houdini with wedging possibilities. Use your own dataset of images to recreate their style, object or person. This node is a connector for the most powerful Stable Diffusion training script called Kohya_ss. It bundled with a couple of presets (easy and medium) for different scenarios. The easy one is filled with common values for most of the cases and ready to train LoRA right out of the box, just tweak some optimization parameters if you got Cuda Memory errors. The medium preset has more parameters to explore.
Please, read Wiki in order to understand how to install and use it:
https://github.com/stassius/StableHoudini/wiki/How-to-train-LoRA

Improvements:
SD Dataset Preparation: A lot of tweaks for seamless workflow with Kohya.

Fixes:

• Minor bugfixes