Skip to content

builder

build_callback function

dlf.core.builder.build_callback(callback, args)

Builds and initializes a callback object.

Valid names are all framework callbacks located at frameworks/callback

Args

  • callback: str. Name of the registered callback
  • args: dict[str, dict[str, Any]]. Arguments to initialize the callback

Raises

  • ValueError: If callback not exists
  • ValueError: If initialization of callback went wrong

Returns

An instances of dlf.core.callback.Callback

build_data_generator function

dlf.core.builder.build_data_generator(name, args)

Builds and initializes a data generator.

Valid names are all registered data generators which are located at dlf/data_generators

Args

  • name: str. Name of the registered data generator
  • args: dict[str, dict[str, Any]]. Keyword arguments for the data generator

Raises

  • FileNotFoundError: If there is no data reader for the given name
  • ValueError: If the initialization of the data reader went wrong

Returns

A tf.data.Dataset object

build_evaluator function

dlf.core.builder.build_evaluator(evaluator, args)

Builds and initializes an evaluator object.

Valid names are all framework evaluators located at frameworks/evaluator

Args

  • evaluator: str. Name of the registered evaluator
  • args: dict[str, dict[str, Any]]. Arguments to initialize the evaluator

Raises

  • ValueError: If evaluator not exists
  • ValueError: If initialization of evaluator went wrong

Returns

An instances of dlf.core.evaluator.evaluator

build_loss function

dlf.core.builder.build_loss(name, args)

Builds and initializes a loss function.

Valid names are all in Keras available Losses and all losses in module dlf/losses (if registered)

Args

  • name: str. Name of the loss function
  • args: dict[str, dict[str, Any]]. Arguments to initialize the loss function

Raises

  • ValueError: If loss function not exists

Returns

A tf.keras.losses.Loss object

build_metric function

dlf.core.builder.build_metric(name, kwargs)

Builds and initializes a metric based on a given name for this metric and with given arguments.

Valid names are all in Keras available Metrics and all metrics in module dlf/metrics (if registered)

Args

name (str): Name of the registered metric kwargs (dict): Arguments for the metric to initialize

Raises

  • ValueError: If the metric is not valid or not registered

Returns

A metric with tf.keras.metrics.Metric as base class

build_model_wrapper function

dlf.core.builder.build_model_wrapper(name, kwargs)

Builds and initializes a model based on the name parameter with kwargs arguments.

Valid names are all in Keras available Applications and all models in module dlf/models (if registered)

Args

  • name: str. Name of the model to initialize (Keras model or a framework model)
  • kwargs: dict. Required argument to initialize a model

Raises

  • FileNotFoundError: If no model exists for a given name
  • ValueError: If the registered model is not subclass of ModelWrapper or the preprocessing function is not callable
  • ValueError: If something went wrong during the initialization with the kwargs

Returns

A ModelWrapper object containing a the specified model, training and validation step

build_optimizer function

dlf.core.builder.build_optimizer(optimizer, args)

Builds and initializes an optimizer.

Valid names are all in Keras available Optimizers

Args

  • optimizer: str. Name of the optimizer to initialize
  • args: dict[str, dict[str, Any]]. Intialization arguments for the selected optimizier

Raises

  • ValueError: If optimizer name does not exists in tf.keras.optimizers
  • ValueError: If an invalid decay learning object is used as learning_rate

Returns

An instance of tf.keras.optimizers.Optimiizer

build_preprocessing_exectutor function

dlf.core.builder.build_preprocessing_exectutor(pipeline)

Builds and initializes a PreprocessingExecutor which contains a list of preprocessing methods.

Valid names are all registered methods in dlf/preprocessing (if registered)

Args

  • pipeline: dict[str, dict[str, Any]. dictionary containing the name of a method as key and keyword arguments as values

Raises

  • ValueError: If the requested preprocessing method not exists

Returns

A dlf.core.preprocessing.PreprocessingExecutor objects