Fine-tuning
Fine-tuning continues a pretrained model's training on your own examples so that it behaves better for a specific task, style or domain.
Instead of training from scratch, you start with an open model and adjust it with a much smaller dataset, for example your support tickets or a particular writing style. Methods such as LoRA update only a small set of extra weights, which makes fine-tuning possible on a single GPU.
Whether you may fine-tune a model and share the result depends on its license. Permissive licenses such as Apache 2.0 allow it, while others add conditions such as naming requirements or revenue limits. Fine-tuning also needs more memory than inference, because training stores extra data for every parameter.
Related terms
- LoRA LoRA, short for low-rank adaptation, fine-tunes a model by training a small set of added weights instead of changing all of its parameters.
- Distillation Distillation trains a smaller student model to imitate a larger teacher model, keeping much of its ability at a fraction of the size.
- Open weights An open-weights model is an AI model whose trained parameters, called weights, are published so that anyone can download and run it, usually under a license that sets the terms of use.