LoRA (low-rank adaptation) fine-tunes a large model by freezing its original weights and training small low-rank matrices added to selected layers. QLoRA does the same on a 4-bit quantised base model, cutting memory further so large models can be tuned on a single GPU.
Because only the small adapter is trained and stored, one base model can serve many task-specific adapters.
Read more about LoRA and QLoRA
How Marquedata helps
Data Science & MLData science, machine learning, model training and LLM fine-tuning — from experiment to monitored production model.
Related terms
Have a Data, AI or Customer Intelligence Challenge?
Let's explore what you can build with your data.
Have an idea for an intelligent product?