将 TFDS 数据集转换为 tensorflow.examples.tutorials.mnist

Question

Previously I got an error on

from tensorflow.examples.tutorials.mnist import input_data because its no longer part of tensorflow 2.0

I was suggested in this post to replace it with

import tensorflow_datasets as tfds
input_data = tfds.load(name="mnist", split=tfds.Split.TRAIN)

But now Im getting the error:

Traceback (most recent call last):
  File "GAN.py", line 79, in <module>
    mnist = input_data.read_data_sets('../../MNIST_data', one_hot=True)
AttributeError: '_OptionsDataset' object has no attribute 'read_data_sets'

How can I convert input_data to the right format to be able to use read_data_sets from it?

Answer 1

As the library has changed for TF2.0 so are its functions. So, read_data_sets is not required and functioning. input_data is already your dataset consisting dictionaries of image and label . Which you can access like

image, label = input_data["image"], input_data["label"]

Read here for more.