tf.data.Dataset.from_tensor_slices(array)獲取分片資料，將輸入的array按照第0維度分片

dataset.make_initializable_iterator() 建立dataset迭代器，需要進行初始化

one_element = iterator.get_next() 從迭代器中獲取一個iter的資料

iterator.initializer 迭代器初始化

import tensorflow as tf
import numpy as np
dataset = tf.data.Dataset.from_tensor_slices(np.random.uniform(size=(5, 2)))
iterator = dataset.make_initializable_iterator()
one_element = iterator.get_next()
with tf.Session() as sess:
    sess.run(iterator.initializer)
    for i in range(5):
        print(sess.run(one_element))
 

[0.89118383 0.24899264]
[0.31330564 0.38645845]
[0.92109775 0.89482123]
[0.3567731 0.9308349]
[0.72154475 0.12027287]

arr = np.arange(100)
dataset = tf.data.Dataset.from_tensor_slices(arr)
dataset = dataset.batch(20).repeat(5)
iterator = dataset.make_initializable_iterator()
one_element = iterator.get_next()
with tf.Session() as sess:
    sess.run(iterator.initializer)
    try:
        while True:
            print(sess.run(one_element))
    except:
        print("end")
 

arr = np.arange(100)
dataset = tf.data.Dataset.from_tensor_slices(arr)
dataset = dataset.shuffle(buffer_size=10000)
dataset = dataset.map(lambda x: x+1).batch(20)
iterator = dataset.make_initializable_iterator()
one_element = iterator.get_next()
with tf.Session() as sess:
    sess.run(iterator.initializer)
    for i in range(5):
        print(sess.run(one_element))
 

dataset中常見的三個transform方法為map，batch，shuffle

dataset.map()用來對dataset的元素進行操作

dataset.batch(num)用來改變dataset的結構，使資料變為以num為一個batch

dataset.shuffle()用來打亂dataset

file_path = "E:\\tf_project\\NMT\\train.txt.zh"
dataset = tf.data.TextLineDataset(file_path)
dataset = dataset.batch(10)
iterator = dataset.make_initializable_iterator()
one_element = iterator.get_next()
with tf.Session() as sess:
    sess.run(iterator.initializer)
    for i in range(3):
        print(sess.run(one_element))

tf.data.TextLineDataset(file_path) 是讀入文字操作