Initializers¶
Initializers for runner
-
class
tflibs.runner.initializer.DatasetInitializer(dataset_pkg='datasets.dataset')[source]¶ Bases:
tflibs.runner.initializer.BaseInitializer-
add_arguments(argparser)[source]¶ Adds arguments
- –dataset-dir
- –dataset-name
Parameters: argparser (argparse.ArgumentParser) – Argument parser used to add arguments
-
handle(parse_args, unknown)[source]¶ Handles arguments.
Exhausts –dataset-name and –dataset-dir, and generates dataset.
Parses dataset-specific arguments.
See: tflib.datasets.BaseDataset.add_arguments()
Parameters: - parse_args (argparse.Namespace) – Parsed arguments.
- unknown (list) – A list of unknown arguments. Exhaust these list.
Returns: A tuple of a dict of handled arguments and unknown arguments.
Return type: tuple
-
-
class
tflibs.runner.initializer.EvalInitializer(model_pkg='models')[source]¶ Bases:
tflibs.runner.initializer.ModelInitializer-
add_arguments(argparser)[source]¶ Adds arguments.
Adds arguments of ModelInitializer.
- –model-name
Parameters: argparser (argparse.ArgumentParser) – Argument parser used to add arguments.
-
handle(parse_args, unknown)[source]¶ Handles arguments.
Parses model-specific arguments.
- model_args: Common arguments. See: tflib.model.Model.add_model_args()
- eval_args: Evaluation arguments. See: tflib.model.Model.add_eval_args()
Parameters: - parse_args (argparse.Namespace) – Parsed arguments.
- unknown (list) – A list of unknown arguments. Exhaust these list.
Returns: A tuple of a dict of handled arguments and unknown arguments.
Return type: tuple
-
-
class
tflibs.runner.initializer.ModelInitializer(model_pkg='models')[source]¶ Bases:
tflibs.runner.initializer.BaseInitializer-
add_arguments(argparser)[source]¶ Adds arguments.
- –model-name
Parameters: argparser (argparse.ArgumentParser) – Argument parser used to add arguments.
-
handle(parse_args, unknown)[source]¶ Handles arguments.
Exhausts –model-name and generates model_cls.
Parameters: - parse_args (argparse.Namespace) – Parsed arguments.
- unknown (list) – A list of unknown arguments. Exhaust these list.
Returns: A tuple of a dict of handled arguments and unknown arguments.
Return type: tuple
-
-
class
tflibs.runner.initializer.TrainInitializer(model_pkg='models')[source]¶ Bases:
tflibs.runner.initializer.ModelInitializer-
add_arguments(argparser)[source]¶ Adds arguments.
Adds arguments of ModelInitializer.
- –model-name
and
- –save-steps
- –keep-checkpoint-max
- –log-steps
- –random-seed
- –train-iters
Parameters: argparser (argparse.ArgumentParser) – Argument parser used to add arguments.
-
handle(parse_args, unknown)[source]¶ Handles arguments.
Parses model-specific arguments.
- model_args: Common arguments. See: tflib.model.Model.add_model_args()
- train_args: Train arguments. See: tflib.model.Model.add_train_args()
- eval_args: Evaluation arguments. See: tflib.model.Model.add_eval_args()
Parameters: - parse_args (argparse.Namespace) – Parsed arguments.
- unknown (list) – A list of unknown arguments. Exhaust these list.
Returns: A tuple of a dict of handled arguments and unknown arguments.
Return type: tuple
-