本文共 3736 字，大约阅读时间需要 12 分钟。

本文将介绍：

• 构建一个tf.train.Example对象
• 将tf.train.Example对象存入文件中,生成的tf_record文件.
• 使用tf.data的API读取tf_record文件，并实现反序列化
• 将tf.train.Example对象存入压缩文件中,生成的tf_record压缩文件.
• 使用tf.data的API读取tf_record压缩文件.

一，构建一个ByteList,FloatList,Int64List的对象

#!/usr/bin/env python3# -*- coding: utf-8 -*-import matplotlib as mplimport matplotlib.pyplot as pltimport numpy as npimport sklearnimport pandas as pdimport osimport sysimport timeimport tensorflow as tffrom tensorflow import keras# 打印使用的python库的版本信息print(tf.__version__)print(sys.version_info)for module in mpl, np, pd, sklearn, tf, keras:    print(module.__name__, module.__version__)    # tfrecord 是一种文件格式# -> tf.train.Example可以是一个样本或者一组#    每个Example-> tf.train.Features -> {"key": tf.train.Feature}#       每个Feature有不同的格式-> tf.train.Feature -> tf.train.ByteList/FloatList/Int64List# 1,构建一个ByteList,FloatList,Int64List的对象# tf.train.ByteListfavorite_books = [name.encode('utf-8') for name in ["machine learning", "cc150"]]favorite_books_bytelist = tf.train.BytesList(value = favorite_books)print(favorite_books_bytelist,type(favorite_books_bytelist))# tf.train.FloatListhours_floatlist = tf.train.FloatList(value = [15.5, 9.5, 7.0, 8.0])print(hours_floatlist,type(hours_floatlist))# tf.train.Int64Listage_int64list = tf.train.Int64List(value = [42])print(age_int64list,type(age_int64list))

二，构建一个tf.train.Features对象

# tf.train.Featuresfeatures = tf.train.Features(    feature = {
           "favorite_books": tf.train.Feature(bytes_list = favorite_books_bytelist),        "hours": tf.train.Feature(float_list = hours_floatlist),        "age": tf.train.Feature(int64_list = age_int64list),    }) print(features)

三，构建一个tf.train.Example对象

# tf.train.Example(使用features构建Example)example = tf.train.Example(features=features)print(example)# 将tf.train.Example对象序列化.serialized_example = example.SerializeToString()print(serialized_example)

四，将tf.train.Example对象存入文件中,生成的tf_record文件

output_dir = 'tfrecord_basic'if not os.path.exists(output_dir):    os.mkdir(output_dir)filename = "test.tfrecords"filename_fullpath = os.path.join(output_dir, filename)# 将序列化后的tf.train.Example对象存入文件中.with tf.io.TFRecordWriter(filename_fullpath) as writer:    for i in range(3):# 写入三次        writer.write(serialized_example)

五，使用tf.data的API读取tf_record文件

dataset = tf.data.TFRecordDataset([filename_fullpath])for serialized_example_tensor in dataset:    print(serialized_example_tensor)

六，将tf.train.Example对象反序列化

# 定义每个feature的数据类型expected_features = {
       "favorite_books": tf.io.VarLenFeature(dtype = tf.string), # VarLenFeature代表变长的数据类型    "hours": tf.io.VarLenFeature(dtype = tf.float32),     "age": tf.io.FixedLenFeature([], dtype = tf.int64), # FixedLenFeature代表定长的数据类型}# 解析TFRecord数据集dataset = tf.data.TFRecordDataset([filename_fullpath])for serialized_example_tensor in dataset:    example = tf.io.parse_single_example(        serialized_example_tensor,        expected_features)    # print(example)    books = tf.sparse.to_dense(example["favorite_books"],default_value=b"")    for book in books:        print(book.numpy().decode("UTF-8"))

七，将tf.train.Example对象存入压缩文件中,生成的tf_record压缩文件

filename_fullpath_zip = filename_fullpath + '.zip'options = tf.io.TFRecordOptions(compression_type = "GZIP")with tf.io.TFRecordWriter(filename_fullpath_zip, options) as writer:    for i in range(3):        writer.write(serialized_example)

八，使用tf.data的API读取tf_record压缩文件

dataset_zip = tf.data.TFRecordDataset([filename_fullpath_zip],compression_type= "GZIP")for serialized_example_tensor in dataset_zip:    example = tf.io.parse_single_example(        serialized_example_tensor,        expected_features)    books = tf.sparse.to_dense(example["favorite_books"],default_value=b"")    for book in books:        print(book.numpy().decode("UTF-8"))

转载地址：http://nvili.baihongyu.com/