clinicadl.train.TrainerState

class clinicadl.train.TrainerState(**data: Any) → None

Represents the state of a Trainer.

Attributes:

• called (Optional[TrainerCall]) – The method of the Trainer that has been called. One of "train", "validate" or "test". None if no method has been called so far.

• stage (Optional[TrainerStage]) – Current action performed by the Trainer. One of "training" or "evaluation".

• should_stop (bool) – Whether the training should be stopped at the end of the current epoch during Trainer.train.

• current_train_batch (int) – Index of the current training batch in Trainer.train.

• num_train_batches (int) – Total number of training batches in Trainer.train.

• current_val_batch (int) – Index of the current validation batch in Trainer.validate or Trainer.train.

• num_val_batches (int) – Total number of validation batches in Trainer.validate or Trainer.train.

• current_test_batch (int) – Index of the current test batch Trainer.test.

• num_test_batches (int) – Total number of test batches in Trainer.test.

• current_epoch (int) – Index of the current epoch in Trainer.train.

• num_epochs (int) – Total number of epochs in Trainer.train.

• optim_step (int) – The total number of optimization steps performed so far in Trainer.train.

• split_idx (Optional[int]) – Index of the split on which training/validation is currently performed in Trainer.train or Trainer.validate.