gistfile1.txt

maxlen = 75
output_size = y_train.shape[1]
max_features= output_size
embed_size = 50
input_size = (maxlen, 1,)
def get_model():
    global input_size, output_size
    inp = Input(shape=input_size)
    x = Embedding(max_features, embed_size)(inp)
    x = CuDNNGRU(50, return_sequences=True)(inp)
    x = LeakyReLU()(x)
    x = Dropout(0.2)(x)
    x = GlobalMaxPooling1D()(x)
    x = Dense(40)(x)
    x = LeakyReLU()(x)
    x = Dropout(0.2)(x)
    x = Dense(output_size, activation="sigmoid")(x)
    model = Model(inputs=inp, outputs=x)
    model.compile(loss='categorical_crossentropy',
                  optimizer='rmsprop')
    model.layers[1].set_weights([embedding_matrix])
    model.layers[1].trainable = False
    return model