Almost every deep learning model has a large number of hyperparameters. Choosing the proper hyperparameters is one of the most common problems in AutoML. A small change in one of the model’s hyperparameters can significantly change its performance. Hyperparameter Optimization (HPO) is the first and most effective step in deep learning model tuning. Due to its ubiquity, Hyperparameter Optimization is sometimes regarded as synonymous with AutoML.
NNI provides a broad and flexible set ...