Efficient Finetuning of Quantized Large Language Models with QLoRA

QLoRA introduces innovations like 4-bit NormalFloat and double quantization to efficiently finetune large language models without sacrificing performance.