pyopmnearwell.ml.scaler_layers module

Provide MinMax scaler layers for tensorflow.keras.

Warning: Tensorflow 2.17 and Keras 3.0 introduce many pylint errors, hence we disable linting completely. It is possible that the module is not functional at the moment.

class pyopmnearwell.ml.scaler_layers.MinMaxScalerLayer(*args, **kwargs)

Bases: ScalerLayer, Layer

Scales the input according to MinMaxScaling.

See https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.MinMaxScaler.html for an explanation of the transform.

Attributes:

compute_dtype: The dtype of the computations performed by the layer.
data_max
data_min
dtype: Alias of layer.variable_dtype.
dtype_policy
feature_range
input: Retrieves the input tensor(s) of a symbolic operation.
input_dtype: The dtype layer inputs should be converted to.
input_spec
is_adapted
losses: List of scalar losses from add_loss, regularizers and sublayers.
metrics: List of all metrics.
metrics_variables: List of all metric variables.
non_trainable_variables: List of all non-trainable layer state.
non_trainable_weights: List of all non-trainable weight variables of the layer.
output: Retrieves the output tensor(s) of a layer.
path: The path of the layer.
quantization_mode: The quantization mode of this layer, None if not quantized.
supports_masking: Whether this layer supports computing a mask using compute_mask.
trainable: Settable boolean, whether this layer should be trainable or not.
trainable_variables: List of all trainable layer state.
trainable_weights: List of all trainable weight variables of the layer.
variable_dtype: The dtype of the state (weights) of the layer.
variables: List of all layer state, including random seeds.
weights: List of all weight variables of the layer.

Methods

`__call__`(args, *kwargs)	Call self as a function.
`adapt`(data)	Fit the layer to the min and max of the data.
`add_loss`(loss)	Can be called inside of the call() method to add a scalar loss.
`add_variable`(shape, initializer[, dtype, ...])	Add a weight variable to the layer.
`add_weight`([shape, initializer, dtype, ...])	Add a weight variable to the layer.
`build`(input_shape)	Initialize `data_min` and `data_max` with the default values if they have not been initialized yet.
`build_from_config`(config)	Builds the layer's states with the supplied config dict.
`call`(inputs)
`compute_output_shape`(input_shape)	Calculate the output shape.
`count_params`()	Count the total number of scalars composing the weights.
`from_config`(config)	Reconstruct the layer from its config.
`get_build_config`()	Returns a dictionary with the layer's input shape.
`get_config`()	Return the config for serialization.
`get_weights`()	Return parameters of the scaling.
`load_own_variables`(store)	Loads the state of the layer.
`rematerialized_call`(layer_call, args, *kwargs)	Enable rematerialization dynamically for layer's call method.
`save_own_variables`(store)	Saves the state of the layer.
`set_weights`(weights)	Set parameters of the scaling.
`stateless_call`(trainable_variables, ...[, ...])	Call the layer without any side effects.

add_metric
compute_mask
compute_output_spec
quantize
quantized_build
quantized_call
symbolic_call

call(inputs: Tensor) → Tensor

compute_output_shape(input_shape): Calculate the output shape.

classmethod from_config(config): Reconstruct the layer from its config.

get_config(): Return the config for serialization.

class pyopmnearwell.ml.scaler_layers.MinMaxUnScalerLayer(*args, **kwargs)

Bases: ScalerLayer, Layer

Unscales the input by applying the inverse transform of MinMaxScalerLayer.

See https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.MinMaxScaler.html for an explanation of the transformation.

Attributes:

compute_dtype: The dtype of the computations performed by the layer.
data_max
data_min
dtype: Alias of layer.variable_dtype.
dtype_policy
feature_range
input: Retrieves the input tensor(s) of a symbolic operation.
input_dtype: The dtype layer inputs should be converted to.
input_spec
is_adapted
losses: List of scalar losses from add_loss, regularizers and sublayers.
metrics: List of all metrics.
metrics_variables: List of all metric variables.
non_trainable_variables: List of all non-trainable layer state.
non_trainable_weights: List of all non-trainable weight variables of the layer.
output: Retrieves the output tensor(s) of a layer.
path: The path of the layer.
quantization_mode: The quantization mode of this layer, None if not quantized.
supports_masking: Whether this layer supports computing a mask using compute_mask.
trainable: Settable boolean, whether this layer should be trainable or not.
trainable_variables: List of all trainable layer state.
trainable_weights: List of all trainable weight variables of the layer.
variable_dtype: The dtype of the state (weights) of the layer.
variables: List of all layer state, including random seeds.
weights: List of all weight variables of the layer.

Methods

`__call__`(args, *kwargs)	Call self as a function.
`adapt`(data)	Fit the layer to the min and max of the data.
`add_loss`(loss)	Can be called inside of the call() method to add a scalar loss.
`add_variable`(shape, initializer[, dtype, ...])	Add a weight variable to the layer.
`add_weight`([shape, initializer, dtype, ...])	Add a weight variable to the layer.
`build`(input_shape)	Initialize `data_min` and `data_max` with the default values if they have not been initialized yet.
`build_from_config`(config)	Builds the layer's states with the supplied config dict.
`compute_output_shape`(input_shape)	Calculate the output shape.
`count_params`()	Count the total number of scalars composing the weights.
`from_config`(config)	Reconstruct the layer from its config.
`get_build_config`()	Returns a dictionary with the layer's input shape.
`get_config`()	Return the config for serialization.
`get_weights`()	Return parameters of the scaling.
`load_own_variables`(store)	Loads the state of the layer.
`rematerialized_call`(layer_call, args, *kwargs)	Enable rematerialization dynamically for layer's call method.
`save_own_variables`(store)	Saves the state of the layer.
`set_weights`(weights)	Set parameters of the scaling.
`stateless_call`(trainable_variables, ...[, ...])	Call the layer without any side effects.

add_metric
call
compute_mask
compute_output_spec
quantize
quantized_build
quantized_call
symbolic_call

call(inputs: Tensor) → Tensor

compute_output_shape(input_shape): Calculate the output shape.

classmethod from_config(config): Reconstruct the layer from its config.

get_config(): Return the config for serialization.

class pyopmnearwell.ml.scaler_layers.ScalerLayer(*args, **kwargs)

Bases: Layer

MixIn to provide functionality for the Scaler Layer.

Attributes:

compute_dtype: The dtype of the computations performed by the layer.
data_max
data_min
dtype: Alias of layer.variable_dtype.
dtype_policy
feature_range
input: Retrieves the input tensor(s) of a symbolic operation.
input_dtype: The dtype layer inputs should be converted to.
input_spec
is_adapted
losses: List of scalar losses from add_loss, regularizers and sublayers.
metrics: List of all metrics.
metrics_variables: List of all metric variables.
non_trainable_variables: List of all non-trainable layer state.
non_trainable_weights: List of all non-trainable weight variables of the layer.
output: Retrieves the output tensor(s) of a layer.
path: The path of the layer.
quantization_mode: The quantization mode of this layer, None if not quantized.
supports_masking: Whether this layer supports computing a mask using compute_mask.
trainable: Settable boolean, whether this layer should be trainable or not.
trainable_variables: List of all trainable layer state.
trainable_weights: List of all trainable weight variables of the layer.
variable_dtype: The dtype of the state (weights) of the layer.
variables: List of all layer state, including random seeds.
weights: List of all weight variables of the layer.

Methods

`__call__`(args, *kwargs)	Call self as a function.
`adapt`(data)	Fit the layer to the min and max of the data.
`add_loss`(loss)	Can be called inside of the call() method to add a scalar loss.
`add_variable`(shape, initializer[, dtype, ...])	Add a weight variable to the layer.
`add_weight`([shape, initializer, dtype, ...])	Add a weight variable to the layer.
`build`(input_shape)	Initialize `data_min` and `data_max` with the default values if they have not been initialized yet.
`build_from_config`(config)	Builds the layer's states with the supplied config dict.
`count_params`()	Count the total number of scalars composing the weights.
`from_config`(config)	Creates an operation from its config.
`get_build_config`()	Returns a dictionary with the layer's input shape.
`get_config`()	Returns the config of the object.
`get_weights`()	Return parameters of the scaling.
`load_own_variables`(store)	Loads the state of the layer.
`rematerialized_call`(layer_call, args, *kwargs)	Enable rematerialization dynamically for layer's call method.
`save_own_variables`(store)	Saves the state of the layer.
`set_weights`(weights)	Set parameters of the scaling.
`stateless_call`(trainable_variables, ...[, ...])	Call the layer without any side effects.

add_metric
call
compute_mask
compute_output_shape
compute_output_spec
quantize
quantized_build
quantized_call
symbolic_call

_adapt() → None

Fit the layer to the min and max of the data. This is done individually for each input feature.

Note:: So far, this is only tested for 1 dimensional input and output. For higher dimensional input and output some functionality might need to be added.
Args:: data: _description_

build(input_shape: tuple[int, ...]) → None

Initialize data_min and data_max with the default values if they have not been initialized yet.

Args:: input_shape (tuple[int, …]): _description_

property data_max

property data_min

property feature_range

Return parameters of the scaling.

Returns:: list[ArrayLike]: List with three elements in the following order: self.data_min_, self.data_max_, self.feature_range_

property is_adapted

min: Tensor

scalar: Tensor

Set parameters of the scaling.

Args:: weights (list[ArrayLike]): List with three elements in the following order: data_min, data_max, feature_range
Raises:: ValueError: If feature_range[0] >= feature_range[1].