BaseTrainer

class neuralhydrology.training.basetrainer.BaseTrainer(cfg: Config)

Bases: object

Default class to train a model.

Parameters:

cfg (Config) – The run configuration.

initialize_training()

Initialize the training class.

This method will load the model, initialize loss, regularization, optimizer, dataset and dataloader, tensorboard logging, and Tester class. If called in a continue_training context, this model will also restore the model and optimizer state.

train_and_validate()

Train and validate the model.

Train the model for the number of epochs specified in the run configuration, and perform validation after every validate_every epochs. Model and optimizer state are saved after every save_weights_every epochs.