tdd = U.TimeDistributedDense(output_length, hidden_dim, output_dim, 'TDD') #Activation activation = U.Activation('softmax') #Output layer output = U.Output() #Output layer mask_out = U.Output() '''Define Relations''' rnn.set_input('input_sequence', data, 'output') rnn.set_input('input_mask', mask_in, 'output') tdd.set_input('input', rnn, 'output_sequence') activation.set_input('input', tdd, 'output') output.set_input('input', activation, 'output') mask_out.set_input('input', mask_in, 'output') '''Build Model''' model = Model() model.add_input(data, 'X') model.add_output(output, 'Y') model.add_input(mask_in, 'MASK_IN') model.add_output(mask_out, 'MASK_OUT') model.add_hidden(rnn) model.add_hidden(tdd) model.add_hidden(activation) model.compile(optimizer='adam', loss_configures=[ ('Y', 'categorical_crossentropy', 'MASK_OUT', False, "categorical"), ], verbose=0)
activation = U.Activation('softmax') #Output layer output_y = U.Output() output_alpha = U.Output() '''Define Relations''' add1.set_input('input', dataA, 'output') decoder.set_input('input_sequence', add1, 'output') decoder.set_input('context', dataB, 'output') one2many.set_input('input', decoder, 'output_sequence_with_alpha') remove1.set_input('input', one2many, 'y') tdd.set_input('input', remove1, 'output') activation.set_input('input', tdd, 'output') output_y.set_input('input', activation, 'output') output_alpha.set_input('input', one2many, 'alpha') '''Build Model''' model = Model() model.add_input(dataA, 'A') model.add_input(dataB, 'B') model.add_output(output_y, ['y']) model.add_output(output_alpha, ['alpha']) model.add_hidden(decoder) model.add_hidden(one2many) model.add_hidden(add1) model.add_hidden(remove1) model.add_hidden(tdd) model.add_hidden(activation) model.compile(optimizer='adam', loss_configures=[ ('y', 'categorical_crossentropy', None, False, "categorical"),
decoder = R.RNN(output_length, hidden_dim, hidden_dim, name='DECODER') #decoder = R.LSTM(output_length, hidden_dim, hidden_dim, name='DECODER') #Time Distributed Dense tdd = U.TimeDistributedDense(output_length, hidden_dim, output_dim, 'TDD') #Activation activation = U.Activation('softmax') #Output layer output = U.Output() '''Define Relations''' encoder.set_input('input_sequence', data, 'output') decoder.set_input('input_single', encoder, 'output_last') tdd.set_input('input', decoder, 'output_sequence') activation.set_input('input', tdd, 'output') output.set_input('input', activation, 'output') '''Build Model''' model = Model() model.add_input(data, 'X') model.add_output(output, 'y') model.add_hidden(encoder) model.add_hidden(decoder) model.add_hidden(tdd) model.add_hidden(activation) model.compile(optimizer='adam', loss_configures=[ ('y', 'categorical_crossentropy', None, False, "categorical"), ], verbose=0) score = model.evaluate(data={
#Time Distributed Dense tdd = U.TimeDistributedDense(output_length, hidden_dim, output_dim, 'TDD') #Activation activation = U.Activation('softmax') #Output layer output = U.Output() '''Define Relations''' encoder.set_input('input_sequence', data, 'output') decoder.set_input('input_single', encoder, 'output_last') tdd.set_input('input', decoder, 'output_sequence') activation.set_input('input', tdd, 'output') output.set_input('input', activation, 'output') '''Build Model''' model = Model() model.add_input(data, 'X') model.add_output(output, 'y') model.add_hidden(encoder) model.add_hidden(decoder) model.add_hidden(tdd) model.add_hidden(activation) model.compile(optimizer='adam', loss_configures = [('y', 'categorical_crossentropy', None, False, "categorical"),], verbose=0) score = model.evaluate(data = {'X': D_X_val, 'y': D_y_val}, show_accuracy=True, verbose=0) print('Test score:', score[0]) print('Test accuracy:', score[1])
activation = U.Activation('softmax') #Output layer output = U.Output() #Output layer mask_out = U.Output() '''Define Relations''' rnn.set_input('input_sequence', data, 'output') rnn.set_input('input_mask', mask_in, 'output') tdd.set_input('input', rnn, 'output_sequence') activation.set_input('input', tdd, 'output') output.set_input('input', activation, 'output') mask_out.set_input('input', mask_in, 'output') '''Build Model''' model = Model() model.add_input(data, 'X') model.add_output(output, 'Y') model.add_input(mask_in, 'MASK_IN') model.add_output(mask_out, 'MASK_OUT') model.add_hidden(rnn) model.add_hidden(tdd) model.add_hidden(activation) model.compile(optimizer='adam', loss_configures = [('Y', 'categorical_crossentropy', 'MASK_OUT', False, "categorical"),], verbose=0) score = model.evaluate(data = {'X': D_X_val, 'MASK_IN': D_mask_val, 'Y': D_Y_val}, batch_size=BATCH_SIZE, show_accuracy=True, verbose=0)
#Output layer output_y = U.Output() output_alpha = U.Output() '''Define Relations''' encoder.set_input('input_sequence', data, 'output') decoder.set_input('input_single', encoder, 'output_last') decoder.set_input('context', data, 'output') one2many.set_input('input', decoder, 'output_sequence_with_alpha') tdd.set_input('input', one2many, 'y') activation.set_input('input', tdd, 'output') output_y.set_input('input', activation, 'output') output_alpha.set_input('input', one2many, 'alpha') '''Build Model''' model = Model() model.add_input(data, 'X') model.add_output(output_y, ['y']) model.add_output(output_alpha, ['alpha']) model.add_hidden(encoder) model.add_hidden(decoder) model.add_hidden(one2many) model.add_hidden(tdd) model.add_hidden(activation) model.compile(optimizer='adam', loss_configures = [('y', 'categorical_crossentropy', None, False, "categorical"),], verbose=2) score = model.evaluate(data = {'X': D_X_val, 'y': D_y_val}, show_accuracy=True, verbose=0)
#Dense unit 3 dense_3 = U.Dense(hidden_size, output_size, 'Dense3') #Activation unit 3 activation_3 = U.Activation('softmax') #Output unit output = U.Output() '''Define Relations''' dense_1.set_input('input', data, 'output') activation_1.set_input('input', dense_1, 'output') dense_2.set_input('input', activation_1, 'output') activation_2.set_input('input', dense_2, 'output') dense_3.set_input('input', activation_2, 'output') activation_3.set_input('input', dense_3, 'output') output.set_input('input', activation_3, 'output') '''Build Model''' model = Model() model.add_input(data, 'X') model.add_output(output, 'y') model.add_hidden(dense_1) model.add_hidden(dense_2) model.add_hidden(dense_3) model.add_hidden(activation_1) model.add_hidden(activation_2) model.add_hidden(activation_3) model.compile(optimizer='rmsprop', loss_configures=[ ('y', 'categorical_crossentropy', None, False, "categorical"), ], verbose=0)
#Activation unit 3 activation_3 = U.Activation('softmax') #Output unit output = U.Output() '''Define Relations''' dense_1.set_input('input', data, 'output') activation_1.set_input('input', dense_1, 'output') dense_2.set_input('input', activation_1, 'output') activation_2.set_input('input', dense_2, 'output') dense_3.set_input('input', activation_2, 'output') activation_3.set_input('input', dense_3, 'output') output.set_input('input', activation_3, 'output') '''Build Model''' model = Model() model.add_input(data, 'X') model.add_output(output, 'y') model.add_hidden(dense_1) model.add_hidden(dense_2) model.add_hidden(dense_3) model.add_hidden(activation_1) model.add_hidden(activation_2) model.add_hidden(activation_3) model.compile(optimizer='rmsprop', loss_configures = [('y', 'categorical_crossentropy', None, False, "categorical"),], verbose=0) model.fit(data = {'X': D_X_train, 'y': D_Y_train}, batch_size=batch_size, nb_epoch=nb_epoch,
decoder = R.RNN(output_length, hidden_dim, hidden_dim, name='DECODER') #decoder = R.LSTM(output_length, hidden_dim, hidden_dim, name='DECODER') #Time Distributed Dense tdd = U.TimeDistributedDense(output_length, hidden_dim, output_dim, 'TDD') #Activation activation = U.Activation('softmax') #Output layer output = U.Output() '''Define Relations''' encoder.set_input('input_sequence', data, 'output') decoder.set_input('input_single', encoder, 'output_last') tdd.set_input('input', decoder, 'output_sequence') activation.set_input('input', tdd, 'output') output.set_input('input', activation, 'output') '''Build Model''' model = Model() model.add_input(data, 'X') model.add_output(output, 'y') model.add_hidden(encoder) model.add_hidden(decoder) model.add_hidden(tdd) model.add_hidden(activation) model.compile(optimizer='adam', loss_configures=[ ('y', 'categorical_crossentropy', None, False, "categorical"), ], verbose=0) import os
#Time Distributed Dense tdd = U.TimeDistributedDense(output_length, hidden_dim, output_dim, 'TDD') #Activation activation = U.Activation('softmax') #Output layer output = U.Output() '''Define Relations''' encoder.set_input('input_sequence', data, 'output') decoder.set_input('input_single', encoder, 'output_last') tdd.set_input('input', decoder, 'output_sequence') activation.set_input('input', tdd, 'output') output.set_input('input', activation, 'output') '''Build Model''' model = Model() model.add_input(data, 'X') model.add_output(output, 'y') model.add_hidden(encoder) model.add_hidden(decoder) model.add_hidden(tdd) model.add_hidden(activation) model.compile(optimizer='adam', loss_configures = [('y', 'categorical_crossentropy', None, False, "categorical"),], verbose=0) import os usr_home = os.path.expanduser('~') save_dir = os.path.join(usr_home, ".dlx/saved_model/") model_path = os.path.join(save_dir, "addition_rnn_h128_i200.dlx")
tdd = U.TimeDistributedDense(length, n_hidden, n_class, 'TDD') tdm = U.TimeDistributedMerge('ave') #Activation activation = U.Activation('softmax') #Output layer output = U.Output() '''Define Relations''' rnn.set_input('input_sequence', data, 'output') tdd.set_input('input', rnn, 'output_sequence') tdm.set_input('input', tdd, 'output') activation.set_input('input', tdm, 'output') output.set_input('input', activation, 'output') '''Build Model''' model = Model() model.add_input(data, 'X') model.add_output(output, 'y') model.add_hidden(rnn) model.add_hidden(tdd) model.add_hidden(tdm) model.add_hidden(activation) model.compile(optimizer='rmsprop', loss_configures = [('y', 'categorical_crossentropy', None, False, "categorical"),], verbose=0) score = model.evaluate(data = {'X': test.feature, 'y': test.label}, show_accuracy=True, verbose=0) print('Test score:', score[0]) print('Test accuracy:', score[1])
output_y = U.Output() output_alpha = U.Output() """Define Relations""" add1.set_input("input", dataA, "output") decoder.set_input("input_sequence", add1, "output") decoder.set_input("context", dataB, "output") one2many.set_input("input", decoder, "output_sequence_with_alpha") remove1.set_input("input", one2many, "y") tdd.set_input("input", remove1, "output") activation.set_input("input", tdd, "output") output_y.set_input("input", activation, "output") output_alpha.set_input("input", one2many, "alpha") """Build Model""" model = Model() model.add_input(dataA, "A") model.add_input(dataB, "B") model.add_output(output_y, ["y"]) model.add_output(output_alpha, ["alpha"]) model.add_hidden(decoder) model.add_hidden(one2many) model.add_hidden(add1) model.add_hidden(remove1) model.add_hidden(tdd) model.add_hidden(activation) model.compile( optimizer="adam", loss_configures=[("y", "categorical_crossentropy", None, False, "categorical")], verbose=0 )