[metanonsense]

Those fully trained networks are usually considered full-rank. At least that is what they say in the LoRA paper.