Using `Evaluate` with multiple inputs

[cynthia166]

Hello, I was wondering if you could please help me,

Confirm the below I have looked for an answer in the Docs. Yes, I have done it My Python version is 3.5 or higher: Python 3.8.5 and TensorFlow 2.4.1 I have searched through the issues Issues for a duplicate. I've tested that my Keras model works as a stand-alone. : Yes, I have been running the keras.

My python is 3.8 and .I am trying to hypertune a deep autoencoder, in my function I need too pass thE X value, I have get the following error: ValueError: Layer model_4 expects 2 input(s), but it received 1 input tensors. Inputs received: [<tf.Tensor 'IteratorGetNext:0' shape=(None, 470) dtype=float32>] Here ir my code: def autoEncoder(X_train, y_train, x_val, y_val,params): ''' Autoencoder for Collaborative Filter Model '''

model = Sequential()

users_items_matrix, content_info = X

content_info = X[:,420:X.shape[1]]

users_items_matrix = X[:,0:420]

Input

input_layer = Input(shape=(users_items_matrix.shape[1],), name='UserScore') input_content = Input(shape=(content_info.shape[1],), name='Itemcontent')

Encoder

-----------------------------

enc = Dense(512, activation=params["activation"], name='EncLayer1')(input_layer)

Content Information

embbeding Turns positive integers (indexes) into dense vectors of fixed size.

x_content = Embedding(100, 256, input_length=content_info.shape[1])(input_content) x_content = Flatten()(x_content) x_content = Dense(256, activation=params["activation"], name='ItemLatentSpace')(x_content)

Latent Space

-----------------------------

Dropout layer randomly sets input units to 0 with a frequency of rate at each step during training time, which helps prevent overfitting. Inputs not set to 0 are scaled up by 1/(1 - rate) such that the sum over all inputs is unchanged.

lat_space = Dense(256, activation=params["activation"], name='UserLatentSpace')(enc) lat_space= add([lat_space, x_content], name='LatentSpace') lat_space = Dropout(params["dropout"], name='Dropout')(lat_space) # Dropout

Decoder

-----------------------------

dec = Dense(512, activation=params["activation"], name='DecLayer1')(lat_space)

Output

output_layer = Dense(users_items_matrix.shape[1], activation='linear', name='UserScorePred')(dec)

this model maps an input to its reconstruction

model = Model([input_layer, input_content], output_layer)

model.compile(optimizer = SGD(lr=0.0001), loss='mse')

model.compile( optimizer=params'optimizer', loss='mean_squared_error',) model.summary() history = model.fit(X,y , validation_data=(X_test, y_test), batch_size=params['batch_size'], epochs=params['epochs'], verbose=0) return history,model p = {'lr': (0.5, 5, 10), 'hidden_layers':[0, 1, 2], 'batch_size': (20, 30, 50), 'epochs': [50],

'times':[216,300,600],

'neurons':[416,600,1200],

'dropout': (0, 0.5, 5), 'weight_regulizer':[None], 'emb_output_dims': [None], 'optimizer': [Adam, "Nadam", "RMSprop"], 'activation':["relu", "selu"], } t = ta.Scan(x=X, y=y, model=autoEncoder,

grid_downsample=1,

    params=p,
    val_split = 0,
    experiment_name='im' )

Than you very much

[mikkokotila]

@cynthia166 could you clean up the formatting of the above so it becomes more readable.

[cynthia166]

Hello Miko,

I cleaned up the format: My question is: How can i make the talos.Evaluate(scan_object) work I get an error becaus my x is multiple inputs, and the error says that list has no shape.

My code:

def autoEncoder(x_train, y_train, x_val, y_val, params):
    '''
    Autoencoder for Collaborative Filter Model
    '''
    #model = Sequential()
    #users_items_matrix, content_info = X
    #content_info = X[:,420:X.shape[1]]
    #users_items_matrix = X[:,0:420]
    # Input
    input_layer   = Input(shape=(420,), name='UserScore')
    input_content = Input(shape=(50,), name='Itemcontent')

    # Encoder
    # -----------------------------
    enc = Dense(512, activation=params["activation"], name='EncLayer1')(input_layer)

    # Content Information
    #embbeding Turns positive integers (indexes) into dense vectors of fixed size.
    x_content = Embedding(100, params['firtr_neurons'], input_length=50)(input_content)
    x_content = Flatten()(x_content)
    x_content = Dense(params['firtr_neurons'], activation=params["activation"], 
                                name='ItemLatentSpace')(x_content)
    # Latent Space
    # -----------------------------
    # Dropout layer randomly sets input units to 0 with a frequency of rate at each step during training time, which helps prevent overfitting. Inputs not set to 0 are scaled up by 1/(1 - rate) such that the sum over all inputs is unchanged.
    lat_space = Dense(params['firtr_neurons'], activation=params["activation"], name='UserLatentSpace')(enc)

    lat_space= add([lat_space, x_content], name='LatentSpace')
    lat_space = Dropout(params["dropout"], name='Dropout')(lat_space) # Dropout

    # Decoder
    # -----------------------------
    dec = Dense(params['firtr_neurons']*2, activation=params["activation"], name='DecLayer1')(lat_space)

    # Output
    output_layer = Dense(420, activation='linear', name='UserScorePred')(dec)

    # this model maps an input to its reconstruction
    model = Model([input_layer, input_content], output_layer)    

#model.compile(optimizer = SGD(lr=0.0001), loss='mse')
    model.compile(
     optimizer="Adam",
    loss='mean_squared_error',)
    model.summary()
    out = model.fit(x=x_train,
                    y=y_train,
                    validation_data=(x_val, y_val),
                    epochs=50,
                    batch_size=params['batch_size'],
                    verbose=0)
    return  out,model

p = {#'lr': (0.5, 5, 10),
     #'hidden_layers':[0, 1, 2],
     'batch_size': [50,100,150],
     'epochs': [50,60],
     #'times':[216,300,600],
     'firtr_neurons':[216,316],
     'dropout': [  .8,.9],
     #'weight_regulizer':[None],
     #'emb_output_dims': [None],
     #'optimizer': ["Adam", "Nadam", "RMSprop"],
     'activation':[ "selu","relu"],
     }

  scan_object = ta.Scan(x=[x_train, x1_train],
                          y=y_train,
                          x_val=[x_val, x1_val],
                          y_val=y_val,
                          params=p,
                          model=autoEncoder,
                        experiment_name="1",

                        )

talos.Evaluate(scan_object)

[cynthia166]

Thank you so much :)

[cynthia166]

[mikkokotila]

Sorry for not replying earlier. Have you looked at this example for multiple inputs: https://autonomio.github.io/talos/#/Examples_Multiple_Inputs

[mikkokotila]

This will be handled in #582 so merging with that.