Deprecated MNIST data loader

[zhanghuimeng]

In the MLP example, this line

mnist = tf.contrib.learn.datasets.load_dataset("mnist")

prints tons of deprecation error:

WARNING:tensorflow:From /home/zhanghuimeng/Documents/learnTensorFlow/simple_introduction/multilayer_perceptron.py:9: load_dataset (from tensorflow.contrib.learn.python.learn.datasets) is deprecated and will be removed in a future version.
Instructions for updating:
Please use tf.data.
...

And it actually can't download anything. (The reason might be...) In the end, you might have to download MNIST by hand. (see this)

A better (not deprecated) alternative is:

(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()

But it cannot download anything either. Finally I had to download from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz and load it by NumPy.

There might be a better alternative, but I still suggest using something not deprecated.

[Ubpa]

You can put mnist.npz to user-root/.keras/datasets/. tf.keras.datasets.mnist.load_data() will try to load it.

[Ubpa]

class DataLoader():
    def __init__(self):
        (train_images, train_labels), (test_images, test_labels) = tf.keras.datasets.mnist.load_data()
        train_images = train_images.reshape(-1, 28 * 28) / 255.0
        test_images = test_images.reshape(-1, 28 * 28) / 255.0
        self.train_data = train_images                     # np.array [55000, 784]
        self.train_labels = train_labels.astype(int)       # np.array [55000] of int32
        self.eval_data = test_images                       # np.array [10000, 784]
        self.eval_labels = test_labels.astype(int)         # np.array [10000] of int32

    def get_batch(self, batch_size):
        index = np.random.randint(0, np.shape(self.train_data)[0], batch_size)
        return self.train_data[index, :], self.train_labels[index]