tensorflow-compression

Data compression in TensorFlow

View on GitHub

Module: tfc

View source on GitHub

Data compression tools.

Modules

python module

Classes

class EntropyBottleneck: Entropy bottleneck layer.

class EntropyModel: Entropy model (base class).

class GDN: Generalized divisive normalization layer.

class GaussianConditional: Conditional Gaussian entropy model.

class IdentityInitializer: Initialize to the identity kernel with the given shape.

class LaplacianConditional: Conditional Laplacian entropy model.

class LogisticConditional: Conditional logistic entropy model.

class NonnegativeParameterizer: Object encapsulating nonnegative parameterization as needed for GDN.

class PackedTensors: Packed representation of compressed tensors.

class Parameterizer: Parameterization object (abstract base class).

class RDFTParameterizer: Object encapsulating RDFT reparameterization.

class SignalConv1D: 1D convolution layer.

class SignalConv2D: 2D convolution layer.

class SignalConv3D: 3D convolution layer.

class StaticParameterizer: A parameterizer that returns a non-variable.

class SymmetricConditional: Symmetric conditional entropy model (base class).

Functions

irdft_matrix(...): Matrix for implementing kernel reparameterization with tf.matmul.

lower_bound(...): Same as tf.maximum, but with helpful gradient for inputs < bound.

pmf_to_quantized_cdf(...): Converts PMF to quantized CDF. This op uses floating-point operations

range_decode(...): Decodes a range-coded code into an int32 tensor of shape shape.

range_encode(...): Using the provided cumulative distribution functions (CDF) inside cdf, returns

same_padding_for_kernel(...): Determine correct amount of padding for same convolution.

unbounded_index_range_decode(...): This is the reverse op of UnboundedIndexRangeEncode, and decodes the range

unbounded_index_range_encode(...): Range encodes unbounded integer data using an indexed probability table.

upper_bound(...): Same as tf.minimum, but with helpful gradient for inputs > bound.