Question

我正在使用生成器（即tf.train.dataset）生成一些from_generator()，然后我对其进行序列化并将其存储在文件中，当我再次加载并通过parse_example对其进行解析时，它可以工作，但是如果我关闭了Jupiter服务器，然后读取了TFRecord文件，它失败了，我不明白为什么。有人看到此错误了吗？：

InvalidArgumentError: Key: wing4.  Can't parse serialized Example.
[[{{node ParseSingleExample/ParseSingleExample}}]]
[[IteratorGetNext]]

During handling of the above exception, another exception occurred:

我抱怨钥匙不存在，看起来像它无法解析序列化的样本。

这是我的一些功能：

def create_example(row):
    features = {
        'wing1': _float_feature(values=row['wing1']),
        'wing2': _float_feature(values=row['wing2']),
        'wing3': _float_feature(values=row['wing3']),
        'wing4': _float_feature(values=row['wing4']),
    }

    return tf.train.Example(features=tf.train.Features(feature=features))


def create_records(gene, annotations, lds, record_path, train=True):
    with tf.python_io.TFRecordWriter(record_path) as writer:
        generator = data_generator(gene, annotations, lds, train)()    

        for row in generator:
            example = create_example(row)
            writer.write(example.SerializeToString())


def parse_function(example_proto):
    features = {
        'wing1': tf.FixedLenFeature([num_features], tf.float32),
        'wing2': tf.FixedLenFeature([num_features], tf.float32),
        'wing3': tf.FixedLenFeature([num_features], tf.float32),
        'wing4': tf.FixedLenFeature([num_features], tf.float32),
    }


    return  tf.parse_single_example(example_proto, features)

如果重新启动Jupyter笔记本，则读取TensorFlow TFRecordDataset失败

0 个答案: