import numpy import theano import theano.tensor as T from seq_to_seq.layers_core import Layer from seq_to_seq import activations softmax = activations.get('softmax') class Softmax(Layer): """ Softmax class. :param n_in: int The size of the input to the layer (i.e., the number of rows in the weight matrix). :param n_out: int The size of layer's output (i.e., the number of columns of the weight matrix and the bias vector). This is the size of the vector that will represent each of the inputs. :param previous_layer: Layer object The previous layer in the computational path. :param layer_number: int The layer position in the computational path. :param seed: int The seed to feed the random number generator. :param auto_setup: boolean
import numpy import theano import theano.tensor as T from seq_to_seq.layers_core import Layer from seq_to_seq import activations softmax = activations.get('softmax') relu = activations.get('relu') class Softmax(Layer): """ Softmax class. :param n_in: int The size of the input to the layer (i.e., the number of rows in the weight matrix). :param n_out: int The size of layer's output (i.e., the number of columns of the weight matrix and the bias vector). This is the size of the vector that will represent each of the inputs. :param previous_layer: Layer object The previous layer in the computational path. :param layer_number: int The layer position in the computational path. :param seed: int The seed to feed the random number generator.
import numpy import theano import theano.tensor as T from seq_to_seq import activations from seq_to_seq.layers_core import Layer sigmoid = activations.get('sigmoid') tanh = activations.get('tanh') class RecurrentLayer(Layer): """ Base class for recurrent layers. :param n_in: int The size of the input to the layer (i.e., the number of rows in the weight matrix). :param dim_proj: int The size of layer's output (i.e., the number of columns of the weight matrix and the bias vector). This is the size of the vector that will represent each of the inputs. :param previous_layer: Layer object The previous layer in the computational path. :param return_sequences: boolean Flag indicating whether or not to the layer should output the previous hidden states. :param layer_number: int The layer position in the computational path.
import numpy import theano import theano.tensor as T from seq_to_seq import activations from seq_to_seq.layers_core import Layer sigmoid = activations.get('sigmoid') tanh = activations.get('tanh') relu = activations.get('relu') class RecurrentLayer(Layer): """ Base class for recurrent layers. :param n_in: int The size of the input to the layer (i.e., the number of rows in the weight matrix). :param n_out: int The size of layer's output (i.e., the number of columns of the weight matrix and the bias vector). This is the size of the vector that will represent each of the inputs. :param previous_layer: Layer object The previous layer in the computational path. :param return_sequences: boolean Flag indicating whether or not to the layer should output the previous hidden states. :param layer_number: int The layer position in the computational path.