El ajuste es el proceso de modificar parámetros, hiperparámetros, prompts o comportamiento del modelo para mejorar el rendimiento en una tarea deseada. Puede implicar seleccionar valores de configuración, fine-tuning de un modelo o refinar entradas sin cambiar los pesos del modelo base.
The process of adapting a foundation model to perform specific tasks with greater precision and accuracy. Tuning is achieved by adjusting some or all of the model's parameters or training a model on a dataset that contains examples that replicate the tasks and results that you want. Tuning techniques include full fine-tuning (updates all parameters), parameter-efficient tuning / adapter tuning (updates some parameters), and supervised fine-tuning (trains on labeled input-output pairs).