CODEWITHSUNDEEP
pythonmachine-learningkerasdeep-learningkeras-tuner
pysaundary
pysaundary

Reputation: 306

Model-building function in Keras Tuner did not return a valid Keras Model instance

I am trying to learn CNN following Krish naik Youtube tutorial in one program i am getting this error:

RuntimeError: Model-building function did not return a valid Keras Model instance, found <keras.engine.sequential.Sequential object at 0x7fd882393b38>

my code is given below 

import keras.datasets
fashion=keras.datasets.fashion_mnist
(x_train,y_train),(x_test,y_test)=fashion.load_data()

x_train=x_train/255.00
x_test=x_test/255.00
x_train=x_train.reshape(x_train.shape[0],x_train.shape[1],x_train.shape[2],1)
x_test=x_test.reshape(x_test.shape[0],x_test.shape[1],x_test.shape[2],1)
from keras.models import Sequential
from keras.layers import Conv2D,Flatten,Dropout,Dense
from keras.optimizers import Adam
def build_knn(hp):
    models=Sequential()
    models.add(Conv2D(filters=hp.Int('conv2d_1',min_value=32,max_value=128,step=16),
                      kernel_size=hp.Choice('conv1_kernal',values=[3,5]),
                      activation='relu',
                      input_shape=(28,28,1)
                                    ))
    models.add(Conv2D(filters=hp.Int('conv2d_1',min_value=32,max_value=128,step=16),
                      kernel_size=hp.Choice('conv1_kernal',values=[3,5]),
                      activation='relu'
                                    ))
    models.add(Flatten())
    models.add(Dense(hp.Int('neural1',min_value=32,max_value=128,step=16),activation='relu'))
    models.add(Dense(10,activation='softmax'))
    models.compile(optimizer=Adam(hp.Choice('learning_rate',values=[1e-2,1e-3])),
                  loss='sparse_categorical_crossentropy',
                  metrics=['accuracy']
                 )
    return models
from kerastuner import RandomSearch
from kerastuner.engine.hyperparameters import HyperParameters
tuner_search=RandomSearch(build_knn,objective='val_accuracy',max_trials=5,directory='jupyterfiles',project_name='krish_naik_fashion_mnist')

Upvotes: 1

Views: 1219

Answers (1)

desertnaut
desertnaut

Reputation: 60318

It is not very clear in the project site, but is indeed mentioned that Keras Tuner is (emphasis added):

A hyperparameter tuner for Keras, specifically for tf.keras with TensorFlow 2.0.

In other words, it works for tf.keras, but not for the stand-alone version of Keras, which you seem to be using here.

Indeed, adapting the basic example from the project site linked above, but changing the imports to use stand-alone Keras, replicates the error you report:

!pip install -U keras-tuner

import keras
from keras import layers
from kerastuner.tuners import RandomSearch

def build_model(hp):
    model = keras.Sequential()
    model.add(layers.Dense(units=hp.Int('units',
                                        min_value=32,
                                        max_value=512,
                                        step=32),
                           activation='relu'))
    model.add(layers.Dense(10, activation='softmax'))
    model.compile(
        optimizer=keras.optimizers.Adam(
            hp.Choice('learning_rate',
                      values=[1e-2, 1e-3, 1e-4])),
        loss='sparse_categorical_crossentropy',
        metrics=['accuracy'])
    return model

tuner = RandomSearch(
    build_model,
    objective='val_accuracy',
    max_trials=5,
    executions_per_trial=3,
    directory='my_dir',
    project_name='helloworld')

Result:

RuntimeError: Model-building function did not return a valid Keras Model instance, found <keras.engine.sequential.Sequential object at 0x7fd882393b38>

But just by changing the imports to use tf.keras instead, i.e.:

from tensorflow import keras
from tensorflow.keras import layers
from kerastuner.tuners import RandomSearch

the same code above runs OK without error.

So, you should change the imports shown here to:

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D,Flatten,Dropout,Dense
from tensorflow.keras.optimizers import Adam

and you should be fine (provided of course that you use Tensorflow 2.0 indeed).

I have just opened a relevant Github issue, advising that they include this clarification explicitly in a visible position.

Upvotes: 3

Related Questions