def get_pipeline_config_options(self):
options = [
ConfigOption(name="over_sampling_methods", default=list(self.over_sampling_methods.keys()), type=str, list=True, choices=list(self.over_sampling_methods.keys())),
ConfigOption(name="under_sampling_methods", default=list(self.under_sampling_methods.keys()), type=str, list=True, choices=list(self.under_sampling_methods.keys())),
ConfigOption(name="target_size_strategies", default=list(self.target_size_strategies.keys()), type=str, list=True, choices=list(self.target_size_strategies.keys())),
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption("ensemble_size", default=50, type=int, info="Build a ensemble of well performing autonet configurations. 0 to disable."),
ConfigOption("ensemble_only_consider_n_best", default=30, type=int, info="Only consider the n best models for ensemble building."),
ConfigOption("ensemble_sorted_initialization_n_best", default=0, type=int, info="Initialize ensemble with n best models.")
]
return options
def get_pipeline_config_options(self):
return [
ConfigOption("result_logger_dir",
default=".",
type="directory"),
ConfigOption("optimize_metric", default="a", type=str),
]
def get_pipeline_config_options(self):
options = [
ConfigOption("refit_config", default=None, type='directory'),
ConfigOption("refit_budget", default=None, type=int),
ConfigOption("confirmation_gmail_user", default=None, type=str),
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption(
name="batch_loss_computation_techniques",
default=list(self.batch_loss_computation_techniques.keys()),
type=str,
list=True,
choices=list(self.batch_loss_computation_techniques.keys())),
ConfigOption("cuda",
default=True,
type=to_bool,
choices=[True, False]),
ConfigOption("torch_num_threads", default=1, type=int),
ConfigOption(
"full_eval_each_epoch",
default=False,
type=to_bool,
choices=[True, False],
info=
"Whether to evaluate everything every epoch. Results in more useful output"
),
ConfigOption(
"best_over_epochs",
default=False,
type=to_bool,
choices=[True, False],
info="Whether to report the best performance occurred to BOHB")
]
for name, technique in self.training_techniques.items():
options += technique.get_pipeline_config_options()
return options
def get_pipeline_config_options(self):
options = [
ConfigOption('ensemble_size', default=0, type=int),
ConfigOption('ensemble_only_consider_n_best', default=0, type=int),
ConfigOption('ensemble_sorted_initialization_n_best', default=0, type=int)
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption("test_split", default=0.0, type=float),
ConfigOption("problem_type", default='feature_classification', type=str, choices=['feature_classification', 'feature_multilabel', 'feature_regression', 'image_classification']),
ConfigOption("data_manager_verbose", default=False, type=to_bool),
ConfigOption("test_instances", default=None, type=str)
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption('plot_logs', default=None, type='str', list=True),
ConfigOption('output_folder', default=None, type='directory'),
ConfigOption('agglomeration', default='mean', choices=['mean', 'median']),
ConfigOption('scale_uncertainty', default=1, type=float),
ConfigOption('font_size', default=12, type=int)
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption('result_dir',
default=None,
type='directory',
required=True),
ConfigOption('name', default=None, type=str, required=True)
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption('only_finished_runs', default=True, type=to_bool),
ConfigOption('result_dir',
default=None,
type='directory',
required=True),
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption("autonet_configs",
default=None,
type='directory',
list=True,
required=True),
ConfigOption("autonet_config_slice", default=None, type=str)
]
return options
def get_pipeline_config_options():
options = [
ConfigOption("early_stopping_patience",
default=float("inf"),
type=float),
ConfigOption("early_stopping_reset_parameters",
default=False,
type=to_bool)
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption("use_dataset_metric", default=False, type=to_bool),
ConfigOption("use_dataset_max_runtime",
default=False,
type=to_bool),
ConfigOption("working_dir", default=None, type='directory'),
ConfigOption("network_interface_name", default=None, type=str)
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption("task_id", default=-1, type=int),
ConfigOption("run_id", default="0", type=str),
ConfigOption("log_level",
default="info",
type=str,
choices=list(self.logger_settings.keys()))
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption(
"default_dataset_download_dir",
default=ConfigFileParser.get_autonet_home(),
type='directory',
info="Directory default datasets will be downloaded to."),
ConfigOption("dataloader_worker", default=1, type=int),
ConfigOption("dataloader_cache_size_mb", default=0, type=int)
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption("validation_split", default=0.0, type=float, choices=[0, 1],
info='In range [0, 1). Part of train dataset used for validation. Ignored in fit if cv_splits > 1 or valid data given.'),
ConfigOption("cv_splits", default=1, type=int, info='The number of CV splits.'),
ConfigOption("use_stratified_cv_split", default=self.use_stratified_cv_split_default, type=to_bool, choices=[True, False]),
ConfigOption("min_budget_for_cv", default=0, type=float,
info='Specify minimum budget for cv. If budget is smaller only evaluate a single fold.'),
ConfigOption("half_num_cv_splits_below_budget", default=0, type=float,
info='Incorporate number of cv splits in budget: Use half the number of specified cv splits below given budget.')
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption("instances",
default=None,
type='directory',
required=True),
ConfigOption("instance_slice", default=None, type=str),
ConfigOption("dataset_root",
default=ConfigFileParser.get_autonet_home(),
type='directory'),
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption("run_id_range", type=str, default=None),
ConfigOption("log_level",
default="info",
type=str,
choices=list(self.logger_settings.keys())),
ConfigOption("benchmark_name",
default=None,
type=str,
required=True)
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption("autonet_configs",
default=None,
type='directory',
list=True,
required=True),
ConfigOption("autonet_config_root",
default=ConfigFileParser.get_autonet_home(),
type='directory'),
ConfigOption("autonet_config_slice", default=None, type=str)
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption(
name='categorical_features',
default=None,
type=to_bool,
list=True,
info=
'List of booleans that specifies for each feature whether it is categorical.'
),
ConfigOption(name='dataset_name', default=None, type=str)
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption(name="networks",
default=list(self.networks.keys()),
type=str,
list=True,
choices=list(self.networks.keys())),
ConfigOption(name="final_activation",
default=self.default_final_activation,
type=str,
choices=list(self.final_activations.keys()))
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption(name="initialization_methods",
default=list(self.initialization_methods.keys()),
type=str,
list=True,
choices=list(self.initialization_methods.keys())),
ConfigOption(name="initializer",
default=self.default_initializer,
type=str,
choices=list(self.initializers.keys()))
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption('dataset_order',
default=None,
type=int,
list=True,
info="Only used for multiple datasets."),
#autonet.refit sets this to false to avoid refit budget issues
ConfigOption('increase_number_of_trained_datasets',
default=False,
type=to_bool,
info="Only used for multiple datasets.")
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption(
name="train_metric",
default=self.default_train_metric,
type=str,
choices=list(self.metrics.keys()),
info="This is the meta train metric BOHB will try to optimize."
),
ConfigOption(name="additional_metrics",
default=[],
type=str,
list=True,
choices=list(self.metrics.keys()))
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption("additional_trajectories",
default=list(),
type="directory",
list=True)
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption(
name="batch_loss_computation_techniques",
default=list(self.batch_loss_computation_techniques.keys()),
type=str,
list=True,
choices=list(self.batch_loss_computation_techniques.keys())),
ConfigOption("minimize",
default=self.default_minimize_value,
type=to_bool,
choices=[True, False]),
ConfigOption("cuda",
default=True,
type=to_bool,
choices=[True, False]),
ConfigOption("save_checkpoints",
default=False,
type=to_bool,
choices=[True, False],
info="Wether to save state dicts as checkpoints."),
ConfigOption("tensorboard_min_log_interval", default=30, type=int),
ConfigOption("tensorboard_images_count", default=0, type=int),
ConfigOption("evaluate_on_train_data", default=True, type=to_bool),
]
for name, technique in self.training_techniques.items():
options += technique.get_pipeline_config_options()
for name, technique in self.batch_loss_computation_techniques.items():
options += technique.get_pipeline_config_options()
return options
def get_pipeline_config_options(self):
options = [
ConfigOption(name="additional_logs",
default=[],
type=str,
list=True,
choices=list(self.log_functions.keys())),
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption(name="embeddings",
default=list(self.embedding_modules.keys()),
type=str,
list=True,
choices=list(self.embedding_modules.keys())),
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption(name="optimizer",
default=list(self.optimizer.keys()),
type=str,
list=True,
choices=list(self.optimizer.keys())),
]
return options
def get_pipeline_config_options(self):
options = [
ConfigOption('dataset_order',
default=None,
type=int,
list=True,
info="Order in which datasets are considered."),
#autonet.refit sets this to false to avoid refit budget issues
ConfigOption(
'increase_number_of_trained_datasets',
default=True,
type=to_bool,
info=
"Wether to increase the number of considered datasets with each successive halfing iteration."
)
]
return options
