以十为单位创建估计器时出现无效语法错误

Question

我试图在Colab笔记本中使用Tensorflow来训练结构相同的神经网络，从MNIST数据集中随机选择不同大小的训练示例。我以前已经成功地用类似的方式训练神经网络好几次了，但这次我收到了一个无效的语法错误，我无法纠正。以下代码行出现错误：

mnist_classifier = tf.estimator.Estimator(model_fn=cnn_model_fn)

我已经在另一个笔记本上成功地使用了这一行代码，但是我无法识别语法上的任何问题（它来自于Tensorflow文档：https://www.tensorflow.org/tutorials/layers）

我试图执行的代码可以在下面找到（但为了简洁起见，我省略了cnn_model_fn的定义）。在

#Initialize list to hold all of the accuracy data
all_accuracies = []

#For each training set group
for group_number in range(len(train_set_sizes)):

  #For each ANN in training set group
  for i in range(num_ANNs):

    #Initialize group_accuracies list
    group_accuracies = []

    #--------------Generate a training set--------------------------------

    #Initialize train_set lists
    train_set_examples = []
    train_set_labels = []

    #randomly select indices to pull train set examples
    example_select_indices = np.random.randint(0,55000,size=train_set_sizes[group_number])

    #For number of images in the appropriate training set group
    for j in range(train_set_sizes[group_number]):

      random_index = example_select_indices[j]
      train_set_examples = np.asarray(train_set_examples.append(train_images[random_index]))
      train_set_labels = np.asarray(train_set_labels.append(train_labels[random_index])

    # Create the MNIST Estimator
    mnist_classifier = tf.estimator.Estimator(model_fn=cnn_model_fn)

    # Train the CNN model
    mnist_classifier.train(input_fn=train_input_fn,steps=train_steps)

    # Evaluate the models and print results
    eval_result = mnist_classifier.evaluate(input_fn=eval_input_fn)
    print("Group Number ",group_number+1,", ANN #",i+1," Accuracy: ",eval_result," %")

    #Record ANN accuracy
    group_accuracies.append(eval_result)

#record group accuracies in all_accuracies
all_accuracies.append(group_accuracies)

有人对为什么会发生这种错误有什么建议吗？在

谢谢你！在