Easy to put into a container and the first training with the defined data is already running (on a 16core, 128GB, Nvidia 3090 with 14GB available).