Module: tf.compat.v1.keras.layers

Keras layers API.

Modules

experimental module: Public API for tf.keras.layers.experimental namespace.

Classes

class AbstractRNNCell: Abstract object representing an RNN cell.

class Activation: Applies an activation function to an output.

class ActivityRegularization: Layer that applies an update to the cost function based input activity.

class Add: Layer that adds a list of inputs.

class AdditiveAttention: Additive attention layer, a.k.a. Bahdanau-style attention.

class AlphaDropout: Applies Alpha Dropout to the input.

class Attention: Dot-product attention layer, a.k.a. Luong-style attention.

class Average: Layer that averages a list of inputs element-wise.

class AveragePooling1D: Average pooling for temporal data.

class AveragePooling2D: Average pooling operation for spatial data.

class AveragePooling3D: Average pooling operation for 3D data (spatial or spatio-temporal).

class AvgPool1D: Average pooling for temporal data.

class AvgPool2D: Average pooling operation for spatial data.

class AvgPool3D: Average pooling operation for 3D data (spatial or spatio-temporal).

class BatchNormalization: Layer that normalizes its inputs.

class Bidirectional: Bidirectional wrapper for RNNs.

class CategoryEncoding: A preprocessing layer which encodes integer features.

class CenterCrop: A preprocessing layer which crops images.

class Concatenate: Layer that concatenates a list of inputs.

class Conv1D: 1D convolution layer (e.g. temporal convolution).

class Conv1DTranspose: Transposed convolution layer (sometimes called Deconvolution).

class Conv2D: 2D convolution layer (e.g. spatial convolution over images).

class Conv2DTranspose: Transposed convolution layer (sometimes called Deconvolution).

class Conv3D: 3D convolution layer (e.g. spatial convolution over volumes).

class Conv3DTranspose: Transposed convolution layer (sometimes called Deconvolution).

class ConvLSTM1D: 1D Convolutional LSTM.

class ConvLSTM2D: 2D Convolutional LSTM.

class ConvLSTM3D: 3D Convolutional LSTM.

class Convolution1D: 1D convolution layer (e.g. temporal convolution).

class Convolution1DTranspose: Transposed convolution layer (sometimes called Deconvolution).

class Convolution2D: 2D convolution layer (e.g. spatial convolution over images).

class Convolution2DTranspose: Transposed convolution layer (sometimes called Deconvolution).

class Convolution3D: 3D convolution layer (e.g. spatial convolution over volumes).

class Convolution3DTranspose: Transposed convolution layer (sometimes called Deconvolution).

class Cropping1D: Cropping layer for 1D input (e.g. temporal sequence).

class Cropping2D: Cropping layer for 2D input (e.g. picture).

class Cropping3D: Cropping layer for 3D data (e.g. spatial or spatio-temporal).

class CuDNNGRU: Fast GRU implementation backed by cuDNN.

class CuDNNLSTM: Fast LSTM implementation backed by cuDNN.

class Dense: Just your regular densely-connected NN layer.

class DenseFeatures: A layer that produces a dense Tensor based on given feature_columns.

class DepthwiseConv1D: Depthwise 1D convolution.

class DepthwiseConv2D: Depthwise 2D convolution.

class Discretization: A preprocessing layer which buckets continuous features by ranges.

class Dot: Layer that computes a dot product between samples in two tensors.

class Dropout: Applies Dropout to the input.

class ELU: Exponential Linear Unit.

class EinsumDense: A layer that uses tf.einsum as the backing computation.

class Embedding: Turns positive integers (indexes) into dense vectors of fixed size.

class Flatten: Flattens the input. Does not affect the batch size.

class GRU: Gated Recurrent Unit - Cho et al. 2014.

class GRUCell: Cell class for the GRU layer.

class GaussianDropout: Apply multiplicative 1-centered Gaussian noise.

class GaussianNoise: Apply additive zero-centered Gaussian noise.

class GlobalAveragePooling1D: Global average pooling operation for temporal data.

class GlobalAveragePooling2D: Global average pooling operation for spatial data.

class GlobalAveragePooling3D: Global Average pooling operation for 3D data.

class GlobalAvgPool1D: Global average pooling operation for temporal data.

class GlobalAvgPool2D: Global average pooling operation for spatial data.

class GlobalAvgPool3D: Global Average pooling operation for 3D data.

class GlobalMaxPool1D: Global max pooling operation for 1D temporal data.

class GlobalMaxPool2D: Global max pooling operation for spatial data.

class GlobalMaxPool3D: Global Max pooling operation for 3D data.

class GlobalMaxPooling1D: Global max pooling operation for 1D temporal data.

class GlobalMaxPooling2D: Global max pooling operation for spatial data.

class GlobalMaxPooling3D: Global Max pooling operation for 3D data.

class Hashing: A preprocessing layer which hashes and bins categorical features.

class InputLayer: Layer to be used as an entry point into a Network (a graph of layers).

class InputSpec: Specifies the rank, dtype and shape of every input to a layer.

class LSTM: Long Short-Term Memory layer - Hochreiter 1997.

class LSTMCell: Cell class for the LSTM layer.

class Lambda: Wraps arbitrary expressions as a Layer object.

class Layer: This is the class from which all layers inherit.

class LayerNormalization: Layer normalization layer (Ba et al., 2016).

class LeakyReLU: Leaky version of a Rectified Linear Unit.

class LocallyConnected1D: Locally-connected layer for 1D inputs.

class LocallyConnected2D: Locally-connected layer for 2D inputs.

class Masking: Masks a sequence by using a mask value to skip timesteps.

class MaxPool1D: Max pooling operation for 1D temporal data.

class MaxPool2D: Max pooling operation for 2D spatial data.

class MaxPool3D: Max pooling operation for 3D data (spatial or spatio-temporal).

class MaxPooling1D: Max pooling operation for 1D temporal data.

class MaxPooling2D: Max pooling operation for 2D spatial data.

class MaxPooling3D: Max pooling operation for 3D data (spatial or spatio-temporal).

class Maximum: Layer that computes the maximum (element-wise) a list of inputs.

class Minimum: Layer that computes the minimum (element-wise) a list of inputs.

class MultiHeadAttention: MultiHeadAttention layer.

class Multiply: Layer that multiplies (element-wise) a list of inputs.

class Normalization: A preprocessing layer which normalizes continuous features.

class PReLU: Parametric Rectified Linear Unit.

class Permute: Permutes the dimensions of the input according to a given pattern.

class RNN: Base class for recurrent layers.

class ReLU: Rectified Linear Unit activation function.

class RepeatVector: Repeats the input n times.

class Rescaling: A preprocessing layer which rescales input values to a new range.

class Reshape: Layer that reshapes inputs into the given shape.

class Resizing: A preprocessing layer which resizes images.

class SeparableConv1D: Depthwise separable 1D convolution.

class SeparableConv2D: Depthwise separable 2D convolution.

class SeparableConvolution1D: Depthwise separable 1D convolution.

class SeparableConvolution2D: Depthwise separable 2D convolution.

class SimpleRNN: Fully-connected RNN where the output is to be fed back to input.

class SimpleRNNCell: Cell class for SimpleRNN.

class Softmax: Softmax activation function.

class SpatialDropout1D: Spatial 1D version of Dropout.

class SpatialDropout2D: Spatial 2D version of Dropout.

class SpatialDropout3D: Spatial 3D version of Dropout.

class StackedRNNCells: Wrapper allowing a stack of RNN cells to behave as a single cell.

class Subtract: Layer that subtracts two inputs.

class ThresholdedReLU: Thresholded Rectified Linear Unit.

class TimeDistributed: This wrapper allows to apply a layer to every temporal slice of an input.

class UpSampling1D: Upsampling layer for 1D inputs.

class UpSampling2D: Upsampling layer for 2D inputs.

class UpSampling3D: Upsampling layer for 3D inputs.

class Wrapper: Abstract wrapper base class.

class ZeroPadding1D: Zero-padding layer for 1D input (e.g. temporal sequence).

class ZeroPadding2D: Zero-padding layer for 2D input (e.g. picture).

class ZeroPadding3D: Zero-padding layer for 3D data (spatial or spatio-temporal).

Functions

Input(...): Input() is used to instantiate a Keras tensor.

add(...): Functional interface to the tf.keras.layers.Add layer.

average(...): Functional interface to the tf.keras.layers.Average layer.

concatenate(...): Functional interface to the Concatenate layer.

deserialize(...): Instantiates a layer from a config dictionary.

disable_v2_dtype_behavior(...): Disables the V2 dtype behavior for Keras layers.

dot(...): Functional interface to the Dot layer.

enable_v2_dtype_behavior(...): Enable the V2 dtype behavior for Keras layers.

maximum(...): Functional interface to compute maximum (element-wise) list of inputs.

minimum(...): Functional interface to the Minimum layer.

multiply(...): Functional interface to the Multiply layer.

serialize(...): Serializes a Layer object into a JSON-compatible representation.

subtract(...): Functional interface to the Subtract layer.