Skip to content
ai

QLoRA

Quantized Low-Rank Adaptation

Definition

QLoRA combines 4-bit quantization of the base model with LoRA fine-tuning, enabling fine-tuning of very large models (65B+ parameters) on a single consumer GPU. The base model is loaded in NF4 (normalized float 4-bit) format and kept frozen, while LoRA adapters are trained in 16-bit precision.

QLoRA demonstrated that high-quality fine-tuning of 65B models is feasible on a single 48GB GPU.

Can this ship? Are we exposed? Can you prove it?

Every CISO is asking these three questions. We answer them with facts, not guesses, from repository to runtime.