tensorforce.core.preprocessing package

Submodules

tensorforce.core.preprocessing.clip module

class tensorforce.core.preprocessing.clip.Clip(min_value, max_value, scope='clip', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Clip by min/max.

tf_process(tensor)

tensorforce.core.preprocessing.divide module

class tensorforce.core.preprocessing.divide.Divide(scale, scope='divide', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Divide state by scale.

tf_process(tensor)

tensorforce.core.preprocessing.grayscale module

class tensorforce.core.preprocessing.grayscale.Grayscale(weights=(0.299, 0.587, 0.114), scope='grayscale', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Turn 3D color state into grayscale.

processed_shape(shape)

tf_process(tensor)

tensorforce.core.preprocessing.image_resize module

class tensorforce.core.preprocessing.image_resize.ImageResize(width, height, scope='image_resize', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Resize image to width x height.

processed_shape(shape)

tf_process(tensor)

tensorforce.core.preprocessing.normalize module

class tensorforce.core.preprocessing.normalize.Normalize(scope='normalize', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Normalize state. Subtract minimal value and divide by range.

tf_process(tensor)

tensorforce.core.preprocessing.preprocessor module

class tensorforce.core.preprocessing.preprocessor.Preprocessor(scope='preprocessor', summary_labels=None)

Bases: object

get_variables()

Returns the TensorFlow variables used by the preprocessor.

Returns:List of variables.

processed_shape(shape)

Shape of preprocessed state given original shape.

Parameters:shape – original shape.

Returns: processed tensor shape

reset()

tf_process(tensor)

Process state.

Parameters:tensor – tensor to process.

Returns: processed tensor.

tensorforce.core.preprocessing.preprocessor_stack module

class tensorforce.core.preprocessing.preprocessor_stack.PreprocessorStack

Bases: object

static from_spec(spec)

Creates a preprocessing stack from a specification dict.

get_variables()

process(tensor)

Process state.

Parameters:tensor – tensor to process

Returns: processed state

processed_shape(shape)

Shape of preprocessed state given original shape.

Parameters:shape – original state shape

Returns: processed state shape

reset()

tensorforce.core.preprocessing.running_standardize module

class tensorforce.core.preprocessing.running_standardize.RunningStandardize(axis=None, reset_after_batch=True, scope='running_standardize', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Standardize state w.r.t past states. Subtract mean and divide by standard deviation of sequence of past states.

reset()

tf_process(tensor)

tensorforce.core.preprocessing.sequence module

class tensorforce.core.preprocessing.sequence.Sequence(length=2, scope='sequence', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Concatenate length state vectors. Example: Used in Atari problems to create the Markov property.

processed_shape(shape)

reset()

tf_process(tensor)

tensorforce.core.preprocessing.standardize module

class tensorforce.core.preprocessing.standardize.Standardize(across_batch=False, scope='standardize', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Standardize state. Subtract mean and divide by standard deviation.

tf_process(tensor)

Module contents

class tensorforce.core.preprocessing.Preprocessor(scope='preprocessor', summary_labels=None)

Bases: object

get_variables()

Returns the TensorFlow variables used by the preprocessor.

Returns:List of variables.

processed_shape(shape)

Shape of preprocessed state given original shape.

Parameters:shape – original shape.

Returns: processed tensor shape

reset()

tf_process(tensor)

Process state.

Parameters:tensor – tensor to process.

Returns: processed tensor.

class tensorforce.core.preprocessing.Sequence(length=2, scope='sequence', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Concatenate length state vectors. Example: Used in Atari problems to create the Markov property.

processed_shape(shape)

reset()

tf_process(tensor)

class tensorforce.core.preprocessing.Standardize(across_batch=False, scope='standardize', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Standardize state. Subtract mean and divide by standard deviation.

tf_process(tensor)

class tensorforce.core.preprocessing.RunningStandardize(axis=None, reset_after_batch=True, scope='running_standardize', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Standardize state w.r.t past states. Subtract mean and divide by standard deviation of sequence of past states.

reset()

tf_process(tensor)

class tensorforce.core.preprocessing.Normalize(scope='normalize', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Normalize state. Subtract minimal value and divide by range.

tf_process(tensor)

class tensorforce.core.preprocessing.Grayscale(weights=(0.299, 0.587, 0.114), scope='grayscale', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Turn 3D color state into grayscale.

processed_shape(shape)

tf_process(tensor)

class tensorforce.core.preprocessing.ImageResize(width, height, scope='image_resize', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Resize image to width x height.

processed_shape(shape)

tf_process(tensor)

class tensorforce.core.preprocessing.PreprocessorStack

Bases: object

static from_spec(spec)

Creates a preprocessing stack from a specification dict.

get_variables()

process(tensor)

Process state.

Parameters:tensor – tensor to process

Returns: processed state

processed_shape(shape)

Shape of preprocessed state given original shape.

Parameters:shape – original state shape

Returns: processed state shape

reset()

class tensorforce.core.preprocessing.Divide(scale, scope='divide', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Divide state by scale.

tf_process(tensor)

class tensorforce.core.preprocessing.Clip(min_value, max_value, scope='clip', summary_labels=())

Bases: tensorforce.core.preprocessing.preprocessor.Preprocessor

Clip by min/max.

tf_process(tensor)