LoRA vs QLoRA: which should you use to fine-tune an LLM?

Use LoRA when the 16-bit model fits your GPU and QLoRA when it does not. Memory math, a cost break-even rule and the Nodus config for both.