LoRA

AI

Low-Rank Adaptation is an efficient fine-tuning technique that adapts large pre-trained models for specific tasks by updating only low-rank weight matrices rather than all parameters. LoRA significantly reduces memory requirements and training time while maintaining model quality.