Creative idea. Any data on how much it takes to load the LoRAs and how much latency it adds to the generation speed?