def create_model(embeddings_file: str, input_size, input_length, hidden_size): """ Create simple regression model with a single embedding layer. :param embeddings_file: embeddings file to load :param input_size: size of input layer :param hidden_size: size of embeddings :return: Keras model """ model = Sequential() model.add( Embedding(input_size, hidden_size, input_length=input_length, name='embedding')) model.add(keras.layers.Lambda(lambda x: keras.backend.sum(x, axis=1))) #model.add(Flatten()) model.add(Dense(92, activation="sigmoid")) if embeddings_file is not None: embeddings = np.loadtxt(embeddings_file) model.get_layer("embedding").set_weights([embeddings]) #model.summary() return model
def create_model(embeddings_file: str, input_size, hidden_size, output_size): """ Create simple regression model with a single embedding layer. :param embeddings_file: embeddings file to load :param input_size: size of input layer :param hidden_size: size of embeddings :param output_size: size of output layer :return: Keras model """ model = Sequential() model.add( Embedding(input_size, hidden_size, input_length=1, name='embedding')) if embeddings_file is not None: embeddings = np.loadtxt(embeddings_file) model.get_layer("embedding").set_weights([embeddings]) model.add(Flatten()) model.add(Dense(output_size, activation="sigmoid")) return model