kumoai.trainer.TrainingJobPlan

Contents

kumoai.trainer.TrainingJobPlan#

class kumoai.trainer.TrainingJobPlan[source]#

Configuration parameters that define the general execution of a Kumo AutoML search. Please see the Kumo documentation for more information.

Variables:

  • num_experiments – (int) The number of experiments to run (default: run_mode-dependent). Supported Task Types: All

  • metrics – (list[str]) The metrics to compute for the run (default: task_type-dependent). Supported Task Types: All

  • tune_metric – (str) The metric to judge performance on (default: task_type-dependent). Supported Task Types: All

  • pruning – (PruningConfig | None) A pruning strategy to early stop unpromising AutoML trials based on the performance of previous trials to speed-up the AutoML search process. (default: {min_epochs=2, k=3, min_delta=0.0, patience=1}). Supported Task Types: All

  • refit_trainval – (bool) Whether to refit the model after training on the training and validation splits (default: True). Supported Task Types: All

  • refit_full – (bool) Whether to refit the model after training on the training, validation and test splits (default: False). Supported Task Types: All

Contents