def train(self, X, y, Xtest, ytest, modelname='GRU', batch_size=32):
model = Sequential()
if modelname == 'GRU':
model.add(GRU(output_dim=self.hid_dim, return_sequence=True, input_shape=(maxvalue, self.word_dim)))
if modelname == 'LSTM':
model.add(LSTM(output_dim=self.hid_dim, return_sequence=True, input_shape=(maxvalue, self.word_dim)))
else:
raise Exception("Model name is not found")
model.add(Dropout(0.5))
model.add(Activation('sigmoid'))
model.compile(loss='categorical_crossentropy', optimizer='adadelta')
model.fit(X,y, batch_size=batch_size, nb_epoch=4, validation_data=(Xtest, ytest), show_accuracy=True)
score, acc = model.evaluate(Xtest, ytest, batch_size=batch_size, show_accuracy=True)
print('Score: {0}'.format(score))
print('Accuracy: {0}'.format(acc))
