"Error when checking model input: you are passing a list as input to your model, but the model expects a a list of 2 Numpy arrays instead."

[mose.w...@gmail.com]

Hi everyone!

I recently got the above bewildering error message upon attempting to train my model. Here is the complete exception:

Exception: Error when checking model input: you are passing a list as input to your model, but the model expects a a list of 2 Numpy arrays instead. The list you passed was: [[array([  1.11022302e-16,  -3.72734042e-08,  -7.45453583e-08,
        -1.11815772e-07,  -1.11022302e-16,  -2.88746119e-08,
        -5.77451024e-08,  -8.66114245e-08]), array([  1.11022302e-16,  -3.72

As one can see, my input is a list of 2 Numpy arrays. Can someone give a clue as to what might be going on here and/or how to fix it? Thanks very much!

Mose

[Bruno Guberfain do Amaral]

It seems that you are passing an 'array of array' as input. Notice the double '[' at the begining: `[[array ...`.

Try to inspect the `len` of your input (it should return 2). 

Em sexta-feira, 17 de junho de 2016 15:36:51 UTC-3, Mose Win
tner escreveu:

[Mose Wintner]

Hi Bruno!

The length of each input is indeed 2 (there are two inputs to my network). The exception itself recognizes the input as a list and not an array, but I was able to verify it as a list manually as well. Any other ideas? Thanks!

[Bruno Guberfain do Amaral]

Can you show us your model.summary() (at least the lines regarding the input layers) and the shape of each np.array you are passing to the fit method?

--
You received this message because you are subscribed to the Google Groups "Keras-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to keras-users...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/keras-users/e12a648e-57ef-452a-8d2e-ccabf22455ce%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[François Chollet]

The model expects a list of 2 arrays. You are *not* passing a list of two arrays (looks at what gets printed). Hence the error. It seems to me that the error message if perfectly clear.

To view this discussion on the web, visit https://groups.google.com/d/msgid/keras-users/5B353C7D-1D60-4CC2-B373-51B317372EA8%40gmail.com.

[Mose Wintner]

    from keras.models import Sequential
    from keras.layers import Merge, Dense, Activation
    from keras.optimizers import SGD
   
    nu=4
    ny=4
    nhidden=4


    thetaHat = Sequential([
        Dense(nhidden, input_dim=nu+ny),
        Activation('softmax'),
        Dense(nhidden),
        Activation('relu'),
        Dense(nu)
        ])
       
    thetaHat.summary()
   
    inputVector=Sequential()
    inputVector.add(Dense(nu, input_dim=nu))
   
    inputVector.summary()
   
    dotted=Merge([thetaHat, inputVector] , mode='dot', dot_axes=1)
   
    out = Sequential()
    out.add(dotted)
   
    sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)

    out.compile(loss='mse', optimizer=sgd, metrics=['accuracy'])    

Running the above yields:

___________________________________________________________________________________________________
Layer (type)                       Output Shape        Param #     Connected to                    
====================================================================================================
dense_1 (Dense)                    (None, 4)           36          dense_input_1[0][0]              
____________________________________________________________________________________________________
activation_1 (Activation)          (None, 4)           0           dense_1[0][0]                    
____________________________________________________________________________________________________
dense_2 (Dense)                    (None, 4)           20          activation_1[0][0]              
____________________________________________________________________________________________________
activation_2 (Activation)          (None, 4)           0           dense_2[0][0]                    
____________________________________________________________________________________________________
dense_3 (Dense)                    (None, 4)           20          activation_2[0][0]              
====================================================================================================
Total params: 76
____________________________________________________________________________________________________
____________________________________________________________________________________________________
Layer (type)                       Output Shape        Param #     Connected to                    
====================================================================================================
dense_4 (Dense)                    (None, 4)           20          dense_input_2[0][0]              
====================================================================================================
Total params: 20
____________________________________________________________________________________________________

Additionally,

    print type(i_train)
    print type(i_train[0])
    print type(i_train[0][0])
    print type(i_train[0][1])
    print len(i_train)
    print len(i_train[0])
    print len(i_train[0][0])
    print len(i_train[0][1])
    print len(i_train[1])
    print i_train[0][0].shape
   
#    out.fit(i_train, o_train, batch_size='128', verbose=1)

gives me

<type 'list'>
<type 'list'>
<type 'numpy.ndarray'>
<type 'numpy.ndarray'>
2144970
2
8
4
2
(8L,)

Please let me know if I can clarify anything more. Thanks!

[Carlos Bentes]

Hi,

maybe this helps:

array_1 = np.array(range(100*(nu+ny))).reshape(100,nu+ny) # thetaHat

array_2 = np.array(range(100*nu)).reshape(100,nu) # inputVector

list_of_2_arrays = [array_1, array_2]

i_train = list_of_2_arrays

cheers,

cb

[Bruno Guberfain do Amaral]

Hi Mose,

I think Carlos has the answer. But to make thinks clear, you should pass a list of two numpy arrays: the first one is the input of thetaHat and the second the input of inputVector.

These inputs must have dimensions: (None, nu+ny) and (None, nu). Here 'None' may be any dimension (the length of your input).

As a final thought, I'd suggest you to refactor your code using the Functional API: http://keras.io/getting-started/functional-api-guide/

Take a look at the "Multi-input and multi-output models" section for a clear example on how to use multiple inputs.

Regards,

Bruno

To view this discussion on the web, visit https://groups.google.com/d/msgid/keras-users/899a8d4a-f5ca-4108-8c94-104a97ac2279%40googlegroups.com.

[Mose Wintner]

Thanks all!

I think I've worked out the issue. I am essentially working with one-dimensional time series; I suspect I may have intended for the length of the input to be nu+ny (resp. nu) and for the input dimension to be 1 (resp. 1).

If I do understand correctly, then my i_train should be a list of two numpy arrays. Should, say, i_train[0] then be an array, where

i_train[0][any_index].shape = (nu+ny, 1) ?

Was my assessment of the initial issue correct?

Thanks again for your patient guidance.

[Bruno Guberfain do Amaral]

Actually:

i_train[0][any_index].shape = (nu+ny,)