tensorflow 數據預處理

import  tensorflow as tf

from  tensorflow import  keras

def preprocess(x,y):
    x = tf.cast(x, dtype = tf.float32) /255.
    y = tf.cast(y, dtype = tf.int64)

    y = tf.one_hot(y,depth = 10)
    print('y shape :',y.shape)
    return  x,y

(x,y),(x_test,y_test) = keras.datasets.fashion_mnist.load_data()
db = tf.data.Dataset.from_tensor_slices((x,y))

db2 = db.map(preprocess).shuffle(60000).batch(100)

res = next(iter(db2))

print('res[0] shape',res[0].shape)
print('res[1] shape',res[1].shape)