“NoneType”对象在keras中没有带有hyperas的属性“计算”错误

Question

我一直得到一个'NoneType‘对象没有属性’计算‘错误，下面的代码使用了hyperas和keras。任何帮助都将不胜感激！

错误是AttributeError：'NoneType‘对象没有属性’计算‘

这是我的第一个keras和hyperas项目。

#First keras program
from keras.datasets import mnist
from keras import models
from keras import layers
from keras import optimizers
from keras.utils import to_categorical
from hyperas import optim
from hyperas.distributions import choice
from hyperopt import Trials, STATUS_OK, tpe
#Loading Data
#Train_images has shape (60000,28,28)
#Train_labels has shape (60000)
def data():
    (train_images, train_labels), (test_images, test_labels) = mnist.load_data()
    train_images = train_images.reshape((60000, 28 * 28))
    train_images = train_images.astype('float32') / 255
    test_images = test_images.reshape((10000, 28 * 28))
    test_images = test_images.astype('float32') / 255
    train_labels = to_categorical(train_labels)
    test_labels = to_categorical(test_labels)
    return train_images,train_labels,test_images,test_labels

#Defining Network and adding Dense Layers
#Compiling Network
def create_model(train_images,train_labels,test_images,test_labels):

    network = models.Sequential()
    network.add(layers.Dense({{choice([256,512,1024])}}, activation='relu', input_shape=(28 * 28,)))
    network.add(layers.Dense(10, activation='softmax'))
    network.compile(optimizer='rmsprop',loss='categorical_crossentropy',metrics=['accuracy'])
    network.fit(train_images,
                train_labels,
                validation_split=0.33,
                epochs=5,
                batch_size=128)

    score,acc = network.evaluate(train_images,train_labels,verbose=0)
    print('Test accuracy:',acc)
    out={'loss':-acc,'score':score,'status':STATUS_OK}
    return out

if __name__ == '__main__':
    best_run, best_model = optim.minimize(model=create_model,
                                          data=data,
                                          algo=tpe.suggest,
                                          max_evals=1,
                                          trials=Trials())
    x_train, y_train, x_test, y_test = data()
    # mnist_model=create_model(x_train,y_train,x_test,y_test)
    print("Evaluation of best performing model:")
    print(best_model.evaluate(x_test, y_test,verbose=0))
    print("Best performing model chosen hyper-parameters:")
    print(best_run)

这是完整的回溯

Traceback (most recent call last):
  File "C:/Users/anubhav/Desktop/Projects/chollet/keras_mnist_dense_hyperas.py", line 51, in <module>
    print(best_model.evaluate(x_test, y_test,verbose=0))
AttributeError: 'NoneType' object has no attribute 'evaluate'

Answer 1

好的，您的optim.minimize函数似乎没有返回模型。

查看图书馆，我发现默认情况下best_model是= None，如果您没有输入有效的trials，那么它将一直保持到最后。我在Keras之上没有太多的知识，所以hyperas在Trials()中所做的事情超出了我的知识范围。

检查这个函数，看看它会输出什么，或者它是否需要任何输入来生成输出等等。

祝好运。

Answer 2

看起来，您正在create_model()中运行实验，然后丢弃模型。如果您返回模型，这可能解决您的问题。尝试：

out={'loss':-acc,'score':score,'status':STATUS_OK, 'model': network}