caikit.core.augmentors

Augmentors apply different types of data augmentation to data model objects, or data streams of data model objects.

Submodules

Classes

`AugmentorBase`
`MergedAugmentor`
`AlwaysSelectionScheme`
`SchemeBase`
`RandomMultiSelectionScheme`
`RandomSingleSelectionScheme`

Package Contents

class caikit.core.augmentors.AugmentorBase(random_seed, produces_none=False)[source]

produces_none = False

augment(inp_obj)[source]

Take an object in, give an object back. Calls ._augment in the subclass.

Args:

inp_obj (str | caikit.core.data_model.DataBase): Object to be: augmented.

Returns:

str | caikit.core.data_model.DataBase: Augmented object of same type: as input inp_obj.

reset()[source]: Reset random number generation for the current augmentor. Note that this currently assumes the augmentor is using the builtin random generator leveraged by Python; if you end up using something else, you may want to override this or restructure this base class to allow resetting of random states based on seed type.

class caikit.core.augmentors.MergedAugmentor(scheme)[source]

Bases: caikit.core.augmentors.base.AugmentorBase

_scheme

augmentor_type

_augment(obj)[source]

Apply a merged augmentor whose behavior is controlled by the encapsulated scheme.

Args:

obj (str | caikit.core.data_model.DataBase): Object to be augmented.

Returns:

str | caikit.core.data_model.DataBase: Augmented object of same type: as input obj.

reset()[source]: Reset the merge augmentor’s random number generation. In this case we actually don’t need to care about the augmentor random state; what matters for this is the augmentor. scheme.

class caikit.core.augmentors.AlwaysSelectionScheme(preserve_order, augmentors, random_seed=1001)[source]

Bases: caikit.core.augmentors.schemes.base.SchemeBase

_execute(obj)[source]

Execute the merged scheme by always applying every contained augmentor (in a potentially shuffled ordering, based on the value of self.preserve_order).

Args:

obj (str | caikit.core.data_model.DataBase): Object to be augmented.

Returns:

str | caikit.core.data_model.DataBase: Augmented object of same type: as input obj.

class caikit.core.augmentors.SchemeBase(preserve_order, augmentors, random_seed)[source]

_preserve_order

_current_order

_augmentors

_init_state

execute(obj)[source]

Execute the merged scheme, i.e., augment the object by leveraging the encapsulated augmentors.

Args:

obj (str | caikit.core.data_model.DataBase): Object to be augmented.

Returns:

str | caikit.core.data_model.DataBase: Augmented object of same type: as input obj.

reset()[source]: Reset the random state of all encapsulated augmentors and the scheme itself.

class caikit.core.augmentors.RandomMultiSelectionScheme(preserve_order, selection_probs, augmentors, random_seed=1001)[source]

Bases: caikit.core.augmentors.schemes.base.SchemeBase

_selection_probs

_execute(obj)[source]

Execute the merged scheme by sequentially applying augmentors (in a potentially shuffled ordering, based on the value of self.preserve_order).

Args:

obj (str | caikit.core.data_model.DataBase): Object to be augmented.

Returns:

str | caikit.core.data_model.DataBase: Augmented object of same type: as input obj.

class caikit.core.augmentors.RandomSingleSelectionScheme(selection_probs, augmentors, random_seed=1001)[source]

Bases: caikit.core.augmentors.schemes.base.SchemeBase

_selection_probs

_execute(obj)[source]

Execute the merged scheme by picking one random augmentor and applying it.

Args:

obj (str | caikit.core.data_model.DataBase): Object to be augmented.

Returns:

str | caikit.core.data_model.DataBase: Augmented object of same type: as input obj.