Searching for the training settings, such as learning rate and batch size, that give a model its best performance.
Hyperparameters are the settings engineers choose before training begins, such as the learning rate or the depth of the network; the training run itself only adjusts the model's weights. Hyperparameter Tuning searches for the configuration that yields the most accurate and efficient Private AI model, often through grid searches that are computationally expensive.
See also Private AI, Large Language Model, Fine-Tuning and Model Inference.