Glossary · Model Training

QLoRA

A fine-tuning technique combining quantization with LoRA, allowing large models to be fine-tuned on much less GPU memory by keeping base weights in a compressed 4-bit format during training.