hyperparameter_hunter package¶
Subpackages¶
- hyperparameter_hunter.callbacks package
- hyperparameter_hunter.data package
- hyperparameter_hunter.io package
- hyperparameter_hunter.keys package
- hyperparameter_hunter.library_helpers package
- hyperparameter_hunter.optimization package
- hyperparameter_hunter.space package
- hyperparameter_hunter.utils package
- Submodules
- hyperparameter_hunter.utils.boltons_utils module
- hyperparameter_hunter.utils.file_utils module
- hyperparameter_hunter.utils.general_utils module
- hyperparameter_hunter.utils.learning_utils module
- hyperparameter_hunter.utils.optimization_utils module
- hyperparameter_hunter.utils.parsing_utils module
- hyperparameter_hunter.utils.result_utils module
- hyperparameter_hunter.utils.version_utils module
- Module contents
Submodules¶
hyperparameter_hunter.algorithm_handlers module¶
hyperparameter_hunter.environment module¶
hyperparameter_hunter.experiment_core module¶
hyperparameter_hunter.experiments module¶
hyperparameter_hunter.feature_engineering module¶
This module organizes and executes feature engineering/preprocessing step functions. The central
components of the module are FeatureEngineer
and EngineerStep
- everything else
is built to support those two classes. This module works with a very broad definition of
“feature engineering”. The following is a non-exhaustive list of transformations that are
considered valid for FeatureEngineer step functions:
Manual feature creation
Input data scaling/normalization/standardization
Target data transformation
Re-sampling
Data imputation
Feature selection/elimination
Encoding (one-hot, label, etc.)
Binarization/binning/discretization
Feature extraction (as for NLP/image recognition tasks)
Feature shuffling
hyperparameter_hunter.importer module¶
This module provides utilities to intercept external imports and load them using custom logic
Related¶
hyperparameter_hunter.__init__
Executes the import hooks to ensure assets are properly imported prior to starting any real work
hyperparameter_hunter.tracers
Defines tracing metaclasses applied by
hyperparameter_hunter.importer
to imports
-
class
hyperparameter_hunter.importer.
Interceptor
(module_name, custom_loader, asset_name=None)¶ Bases:
_frozen_importlib_external.PathFinder
Class to intercept loading of an external module in order to provide custom loading logic
- Parameters
- module_name: String
The path of the module, for which loading should be handled by custom_loader
- custom_loader: Descendant of `importlib.machinery.SourceFileLoader`
Should implement
exec_module()
, which should call its superclass’sexec_module()
, then perform the custom loading logic, and return module
Methods
find_module
(fullname[, path])find the module on sys.path or ‘path’ based on sys.path_hooks and sys.path_importer_cache.
find_spec
(full_name[, path, target])Perform custom loading logic if full_name ==
module_name
invalidate_caches
()Call the invalidate_caches() method on all path entry finders stored in sys.path_importer_caches (where implemented).
-
find_spec
(full_name, path=None, target=None)¶ Perform custom loading logic if full_name ==
module_name
-
class
hyperparameter_hunter.importer.
KerasLayerLoader
(fullname, path)¶ Bases:
_frozen_importlib_external.SourceFileLoader
Cache the module name and the path to the file found by the finder.
Methods
create_module
(spec)Use default semantics for module creation.
exec_module
(module)Set module.Layer to a traced version of itself via
tracers.ArgumentTracer
get_code
(fullname)Concrete implementation of InspectLoader.get_code.
get_data
(path)Return the data from path as raw bytes.
get_filename
([name])Return the path to the source file as found by the finder.
get_source
(fullname)Concrete implementation of InspectLoader.get_source.
is_package
(fullname)Concrete implementation of InspectLoader.is_package by checking if the path returned by get_filename has a filename of ‘__init__.py’.
load_module
([name])Load a module from a file.
path_mtime
(path)Optional method that returns the modification time (an int) for the specified path, where path is a str.
path_stats
(path)Return the metadata for the path.
set_data
(path, data, *[, _mode])Write bytes data to a file.
source_to_code
(data, path, *[, _optimize])Return the code object compiled from source.
-
exec_module
(module)¶ Set module.Layer to a traced version of itself via
tracers.ArgumentTracer
-
-
hyperparameter_hunter.importer.
hook_keras_layer
()¶ If Keras has yet to be imported, modify the inheritance structure of its base Layer class to inject attributes that keep track of the parameters provided to each layer
-
class
hyperparameter_hunter.importer.
KerasMultiInitializerLoader
(fullname, path)¶ Bases:
_frozen_importlib_external.SourceFileLoader
Cache the module name and the path to the file found by the finder.
Methods
create_module
(spec)Use default semantics for module creation.
exec_module
(module)Execute the module.
get_code
(fullname)Concrete implementation of InspectLoader.get_code.
get_data
(path)Return the data from path as raw bytes.
get_filename
([name])Return the path to the source file as found by the finder.
get_source
(fullname)Concrete implementation of InspectLoader.get_source.
is_package
(fullname)Concrete implementation of InspectLoader.is_package by checking if the path returned by get_filename has a filename of ‘__init__.py’.
load_module
([name])Load a module from a file.
path_mtime
(path)Optional method that returns the modification time (an int) for the specified path, where path is a str.
path_stats
(path)Return the metadata for the path.
set_data
(path, data, *[, _mode])Write bytes data to a file.
source_to_code
(data, path, *[, _optimize])Return the code object compiled from source.
-
exec_module
(module)¶ Execute the module.
-
-
hyperparameter_hunter.importer.
hook_keras_initializers
()¶
hyperparameter_hunter.metrics module¶
hyperparameter_hunter.models module¶
hyperparameter_hunter.sentinels module¶
hyperparameter_hunter.settings module¶
This module is the doorway for other modules to access the information set by the active
hyperparameter_hunter.environment.Environment
, and to access the appropriate logging
methods. Specifically, other modules will most often use hyperparameter_hunter.settings.G
to access the aforementioned information. Additionally, this module defines several variables to
assist in navigating the ‘HyperparameterHunterAssets’ directory structure
Related¶
hyperparameter_hunter.environment
This module sets
hyperparameter_hunter.settings.G.Env
to itself, creating the primary gateway used by other modules to access the active Environment’s information
-
class
hyperparameter_hunter.settings.
G
¶ Bases:
object
This class defines global attributes that are set upon instantiation of
environment.Environment
. All attributes contained herein are class variables (not instance variables) because the expectation is for the attributes of this class to be set only once, then referenced by operations that may be executed after instantiating aenvironment.Environment
. This allows functions to be called or classes to be initiated without passing a reference to the currently active Environment, because they check the attributes of this class, instead- Attributes
- Env: None
This is set to “self” in
environment.Environment.__init__()
. This fact allows other modules to check ifsettings.G.Env
is None. If None, aenvironment.Environment
has not yet been instantiated. If not None, any attributes or methods of the instantiated Env may be called- save_transformed_predictions: False
Declares format in which a model’s predictions should be saved, with regard to
feature_engineering.FeatureEngineer
transformations. If no transformation of the target variable takes place (either throughfeature_engineering.FeatureEngineer
,feature_engineering.EngineerStep
, or otherwise), then this setting can be ignored.If save_transformed_predictions is True, and target transformation does occur, then experiment predictions are saved in the same form as the transformed target, which is the form returned directly by a fitted model’s predict method. For example, if target data is label-encoded, and an
feature_engineering.EngineerStep
is used to one-hot encode the target, then one-hot-encoded predictions will be saved.Conversely, if save_transformed_predictions is False (default), and target transformation does occur, then experiment predictions are saved in the inverted form of the transformed target, which is the same form as the original target data. Continuing the example of label-encoded target data, and an
feature_engineering.EngineerStep
to one-hot encode the target, in this case, label-encoded predictions will be saved.- priority_callbacks: Tuple
Intended for internal use only. The contents of this tuple are inserted at the front of an Experiment’s list of callback bases via
experiment_core.ExperimentMeta
, ahead of even the Experiment’s original base classes. This is used primarily for testing callbacks, but it can also be used if you absolutely need a callback to be placed before the Experiment’s other ancestors in its MRO- log_: print
…
- debug_: print
…
- warn_: print
…
- import_hooks: List
…
- sentinel_registry: List
…
Methods
debug
(content, *args, **kwargs)Set in
environment.Environment.initialize_reporting()
to the updated version ofreporting.ReportingHandler.debug()
debug_
(value, …[, sep, end, file, flush])Prints the values to a stream, or to sys.stdout by default.
log
(content, *args, **kwargs)Set in
environment.Environment.initialize_reporting()
to the updated version ofreporting.ReportingHandler.log()
log_
(value, …[, sep, end, file, flush])Prints the values to a stream, or to sys.stdout by default.
Return the attributes of
settings.G
to their original valueswarn
(content, *args, **kwargs)Set in
environment.Environment.initialize_reporting()
to the updated version ofreporting.ReportingHandler.warn()
Issue a warning, or maybe ignore it or raise an exception.
-
Env
= None¶
-
save_transformed_predictions
= False¶
-
priority_callbacks
= ()¶
-
static
log
(content, *args, **kwargs)¶ Set in
environment.Environment.initialize_reporting()
to the updated version ofreporting.ReportingHandler.log()
-
static
debug
(content, *args, **kwargs)¶ Set in
environment.Environment.initialize_reporting()
to the updated version ofreporting.ReportingHandler.debug()
-
static
warn
(content, *args, **kwargs)¶ Set in
environment.Environment.initialize_reporting()
to the updated version ofreporting.ReportingHandler.warn()
-
log_
(value, ..., sep=' ', end='\n', file=sys.stdout, flush=False)¶ Prints the values to a stream, or to sys.stdout by default. Optional keyword arguments: file: a file-like object (stream); defaults to the current sys.stdout. sep: string inserted between values, default a space. end: string appended after the last value, default a newline. flush: whether to forcibly flush the stream.
-
debug_
(value, ..., sep=' ', end='\n', file=sys.stdout, flush=False)¶ Prints the values to a stream, or to sys.stdout by default. Optional keyword arguments: file: a file-like object (stream); defaults to the current sys.stdout. sep: string inserted between values, default a space. end: string appended after the last value, default a newline. flush: whether to forcibly flush the stream.
-
warn_
()¶ Issue a warning, or maybe ignore it or raise an exception.
-
import_hooks
= ['keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling', 'keras_layer', 'keras_initializer', 'keras_variance_scaling']¶
-
sentinel_registry
= []¶
-
classmethod
reset_attributes
()¶ Return the attributes of
settings.G
to their original values
hyperparameter_hunter.tracers module¶
This module defines metaclasses used to trace the parameters passed through operation-critical classes that are members of other libraries. These are only used in cases where it is impractical or impossible to effectively retrieve the arguments explicitly provided by a user, as well as the default arguments for the classes being traced. Generally, tracer metaclasses will aim to add some attributes to the class, that will collect default values, and provided arguments on the class’s creation, and an instance’s call
Related¶
hyperparameter_hunter.importer
This module handles the interception of certain imports in order to inject the tracer metaclasses defined in
hyperparameter_hunter.tracers
into the inheritance structure of objects that need to be traced
-
class
hyperparameter_hunter.tracers.
ArgumentTracer
(name, bases, namespace, **kwargs)¶ Bases:
type
Metaclass to trace the default arguments and explicitly provided arguments of its descendants. It also has special provisions for instantiating dummy models if directed to
Methods
__call__
(*args, **kwargs)Call self as a function.
mro
()return a type’s method resolution order
-
class
hyperparameter_hunter.tracers.
LocationTracer
(name, bases, namespace, **kwargs)¶ Bases:
hyperparameter_hunter.tracers.ArgumentTracer
Metaclass to trace the origin of the call to initialize the descending class
Methods
__call__
(*args, **kwargs)Call self as a function.
mro
()return a type’s method resolution order