Multiple inputs with Keras Functional API

Question

It seems that Keras lacks documentation regarding functional API but I might be getting it all wrong.

I have multiple independent inputs and I want to predict an output for each input. Here's my code so far:

 hour = Input(shape=(1,1), name='hour')
 hour_output = LSTM(1, name='hour_output')(hour)

 port = Input(shape=(1,1), name='port')
 port_output = LSTM(1, name='port_output')(port)

 model = Model(inputs=[hour, port], outputs = [hour_output, port_output])

 model.compile(loss="mean_squared_error", optimizer="adam", metrics=['accuracy'])
 model.fit(trainX, trainY, epochs=10 batch_size=1, verbose=2, shuffle=False)

Error I get for this:

ValueError: No data provided for "hour_output". Need data for each key in: ['hour_output', 'port_output']

I also had a hard time getting the right input for this, so I ended up using dictionary with example structure:{'hour': array([[[0]], [[1]], [[3]]]) }. I don't like that (using dict) either.

Note that I have more inputs to use and for this to make sense, but now I am just trying to make the model work.

adhg · Accepted Answer · 2019-09-27 20:57:58Z

1

In your model.fit you need to provide a list of inputs with length = 2, since you define two inputs in your Model.

Split your training data in train_hour and train_port and call fit like:

model.fit([train_X_hour, train_X_port], [train_Y_hour, train_Y_port] epochs=10, batch_size=1, verbose=2, shuffle=False)

edited Sep 27, 2019 at 20:57

adhg

11k13 gold badges63 silver badges99 bronze badges

answered Jun 19, 2017 at 8:01

Michele Tonutti

4,3582 gold badges23 silver badges23 bronze badges

Sign up to request clarification or add additional context in comments.

6 Comments

mortusdetus Over a year ago

According to the docs input can be:

Numpy array of training data, or list of Numpy arrays if the model has multiple inputs. If all inputs in the model are named, you can also pass a dictionary mapping input names to Numpy arrays.

meaning that your approach is the same as my dictionary. However, I've tried that and it really doesn't make a change.

mortusdetus Over a year ago

But yes, I guess that answers my other question/complaint about not using dictionary :)

Michele Tonutti Over a year ago

Well, you said you did not like the dictionary option, so I gave you the other option. You have to use one of the two - maybe I'm not really understanding what you are asking?

mortusdetus Over a year ago

Yes, that's fine, thank you. The main question here is about running the training, I still can't pass the error.

Michele Tonutti Over a year ago

You mean that even by passing the inputs as a list it throws you that ValueError?

|

Collectives™ on Stack Overflow

Multiple inputs with Keras Functional API

1 Answer 1

6 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

6 Comments

Your Answer

Sign up or log in

Post as a guest

Related