[Feature Request] Add Rosa Training

The code isn't out yet but thought I would link here Rosa is a modification of lora to increase the quality would be interesting to see how it applies to image models https://twitter.com/marktenenholtz/status/1745875618732925086

Start by reviewing the linked Rosa description and confirming its implementation status, since the issue says the code is not yet available. Then identify the project entry points for LoRA training and image models. Done should include a defined Rosa training approach and evidence that it applies to image models.

Written by the indexing model from the issue text.