Search code examples
pythonarrayskerasneural-networkreinforcement-learning

Error: Setting an Array Element with a Sequence. Keras Neural Network

I'm want to build DQNAgent but I have a problem whit the data or with the NN(I'm not sure). I tried to solve it by changing the shape of the array but I'm always getting the same error:ValueError: setting an array element with a sequence.

This is the data:

state(array,shape:(2,2)):

       [4499.74073719,  121.58564876],
       [4669.91329184,   42.37631835]])
  array([[-2000.        ,   290.01270128]])]
 [array([[-4370.,   800.],
       [ -635.,   800.]]) -12.0]]

Label(Q value):

array([0.23,1,3,0.1234])

The programe need to prodict the q_value baice on the data. This is the code:

    model=Sequential()
    model.add(Flatten(input_shape=(2,2)))
    model.add(Dense(24,  activation='relu'))
    model.add(Dense(24, activation='relu'))
    model.add(Dense(acton_size, activation='linear'))
    model.compile(loss='mse',optimizer=Adam(lr=learning_rate))
    return model

    r_locs, i_locs, c_locs, ang, score=Game_step(random.randint(0,4))
    state=np.array([r_locs, i_locs, c_locs, ang])
    state=np.reshape(state,[2,2])

ERROR:

return array(a, dtype, copy=False, order=order)
ValueError: setting an array element with a sequence.

Any idea how to solve this? THANKS😁

Solution

  • Not entirely sure about this, but it may be that the state matrix you are trying to process through the network is not being interpreted as you expect. I believe, the network is expecting each sample to be a 2x2 matrix of numbers, and your state is a combination of numbers and arrays from what it appears. This might point you in the right direction. I hope it helps.