>>94857342Low Rank Adaptation, it's a way of training a model without recalculating billions of parameters
There are many methods of training, but you can think of it as the AI guessing random solutions to problems and writing down which solutions were more correct. In the case of AI image generation, it's like guessing the original image given a blurry image as input. Eventually, the AI becomes so good at guessing, it can "guess" from a fake blurry image that has no original. This is how it makes images from just a description - it's sort of unblurring an image of random noise in a way that looks like images it has seen before related to the words in the prompt.
