A guide explains low-rank adaptation (LoRA), which fine-tunes a frozen model by learning small low-rank update matrices instead of changing every parameter, covering trade-offs, evaluation and controls.
Read it at Unite.AI →
← Front page