deepscribe.pipeline.training

class deepscribe.pipeline.training.RunTalosScanTask(*args, **kwargs)

Runs a Talos scan from the model_definition parameter (a dictionary of lists instead of single values)

load_def()

Loads and preprocesses the model definition JSON file.

output()

Pickled Talos Scan object.

Returns:luigi.LocalTarget

run_training(model_params: dict)

Executes model training and saves result to disk.

Parameters:model_params – dictionary containing model parameter information.

class deepscribe.pipeline.training.TrainKerasModelFromDefinitionTask(*args, **kwargs)

Trains a Keras model from the model_definition parameter and saves it to disk.

load_def()

Loads and preprocesses the model definition JSON file.

output()

Output location of trained Keras model in HDF5 format.

Returns:luigi.LocalTarget

run_training(model_params: dict)

Selects model class from params dictionary and runs training.

Parameters:model_params – dict Returns:None

class deepscribe.pipeline.training.TrainModelFromDefinitionTask(*args, **kwargs)

Luigi task skeleton for a task that loads parameters from a JSON file, trains a model based on those parameters, and saves the model (or other results) to disk.

load_def()

Loads and preprocesses the model definition JSON file.

requires()

Returns:SelectDatasetTask

run()

Creates output directories, load model definition, run training

Returns:

run_training(model_params: dict)

Executes model training and saves result to disk.

Parameters:model_params – dictionary containing model parameter information.