TensorFlow 2.0 Quick Start Guide

TFRecord example 2

In this example, we look at a more complicated record structure given by this dictionary:

filename = './students.tfrecords'data = {            'ID': 61553,            'Name': ['Jones', 'Felicity'],            'Scores': [45.6, 97.2]         }

Using this, we can construct a tf.train.Example class, again using the Feature() method. Note how we have to encode our string:

ID = tf.train.Feature(int64_list=tf.train.Int64List(value=[data['ID']]))Name = tf.train.Feature(bytes_list=tf.train.BytesList(value=[n.encode('utf-8') for n in data['Name']]))Scores = tf.train.Feature(float_list=tf.train.FloatList(value=data['Scores']))example = tf.train.Example(features=tf.train.Features(feature={'ID': ID, 'Name': Name, 'Scores': Scores }))

Serializing and writing this record to ...