train

train().

cneuromax.fitting.deeplearning.train.train(trainer, datamodule, litmodule, logger, config)

Trains a Deep Neural Network.

Note that this function will be executed by num_nodes * gpus_per_node processes/tasks. Those variables are set in the Hydra launcher configuration.

Trains (or resumes training) the model, saves a checkpoint and returns the final validation loss.

Parameters:

• trainer (partial[Trainer])

• datamodule (BaseDataModule)

• litmodule (BaseLitModule)

• logger (partial[WandbLogger])

• config (DeepLearningSubtaskConfig)

Return type:

float

Returns:

The final validation loss.