如何使用Dataset API读取变量长度列表的TFRecords文件？

Question

我想使用Tensorflow的Dataset API来读取变量长度列表的TFRecords文件。这是我的密码。

def _int64_feature(value):
    # value must be a numpy array.
    return tf.train.Feature(int64_list=tf.train.Int64List(value=value))
def main1():
    # Write an array to TFrecord.
    # a is an array which contains lists of variant length.
    a = np.array([[0, 54, 91, 153, 177],
                 [0, 50, 89, 147, 196],
                 [0, 38, 79, 157],
                 [0, 49, 89, 147, 177],
                 [0, 32, 73, 145]])

    writer = tf.python_io.TFRecordWriter('file')

    for i in range(a.shape[0]): # i = 0 ~ 4
        x_train = a[i]
        feature = {'i': _int64_feature(np.array([i])), 'data': _int64_feature(x_train)}

        # Create an example protocol buffer
        example = tf.train.Example(features=tf.train.Features(feature=feature))

        # Serialize to string and write on the file
        writer.write(example.SerializeToString())

    writer.close()

    # Check TFRocord file.
    record_iterator = tf.python_io.tf_record_iterator(path='file')
    for string_record in record_iterator:
        example = tf.train.Example()
        example.ParseFromString(string_record)

        i = (example.features.feature['i'].int64_list.value)
        data = (example.features.feature['data'].int64_list.value)
        #data = np.fromstring(data_string, dtype=np.int64)
        print(i, data)

    # Use Dataset API to read the TFRecord file.
    def _parse_function(example_proto):
        keys_to_features = {'i'   :tf.FixedLenFeature([], tf.int64),
                            'data':tf.FixedLenFeature([], tf.int64)}
        parsed_features = tf.parse_single_example(example_proto, keys_to_features)
        return parsed_features['i'], parsed_features['data']

    ds = tf.data.TFRecordDataset('file')
    iterator = ds.map(_parse_function).make_one_shot_iterator()
    i, data = iterator.get_next()
    with tf.Session() as sess:
        print(i.eval())
        print(data.eval())

检查TF记录文件

[0] [0, 54, 91, 153, 177]
[1] [0, 50, 89, 147, 196]
[2] [0, 38, 79, 157]
[3] [0, 49, 89, 147, 177]
[4] [0, 32, 73, 145]

但是当我试图使用数据集API读取TFRecords文件时，它显示了以下错误。

张量流。python框架错误。InvalidArgumentError:名称：，键：数据，索引：0。int64值的数目！=预期。值大小：5，但输出形状：[]

多谢各位
更新：我尝试使用以下代码读取带有Dataset API的TFRecord，但都失败了。

def _parse_function(example_proto):
    keys_to_features = {'i'   :tf.FixedLenFeature([], tf.int64),
                        'data':tf.VarLenFeature(tf.int64)}
    parsed_features = tf.parse_single_example(example_proto, keys_to_features)
    return parsed_features['i'], parsed_features['data']

ds = tf.data.TFRecordDataset('file')
iterator = ds.map(_parse_function).make_one_shot_iterator()
i, data = iterator.get_next()
with tf.Session() as sess:
    print(sess.run([i, data]))

或

def _parse_function(example_proto):
    keys_to_features = {'i'   :tf.VarLenFeature(tf.int64),
                        'data':tf.VarLenFeature(tf.int64)}
    parsed_features = tf.parse_single_example(example_proto, keys_to_features)
    return parsed_features['i'], parsed_features['data']

ds = tf.data.TFRecordDataset('file')
iterator = ds.map(_parse_function).make_one_shot_iterator()
i, data = iterator.get_next()
with tf.Session() as sess:
    print(sess.run([i, data]))

错误是：

回溯（最后一次调用）：文件“/usr/local/lib/python3.5/dist packages/tensorflow/python/framework/tensor_util.py”，第468行，make_tensor_proto str_values=[compat.as_bytes（x）for x in proto_values]文件“/usr/local/lib/python3.5/dist packages/tensorflow/python/python/framework/python/framework/framework/tensor/tensor_util.py”，第468行，str_值=[compat对于proto_values]文件“/usr/local/lib/python3.5/dist packages/tensorflow/python/util/compat.py”中的x，第65行，在as_字节（bytes_或_text）中，类型错误：应为二进制或unicode字符串，已获取

在处理上述异常期间，发生了另一个异常：

Traceback（最近一次调用）：文件"2tfrecord.py"，第126行，在main 1（）文件"2tfrecord.py"，第72行，在main 1迭代器=ds.map（_parse_function）。make_one_shot_iterator（）文件"/usr/本地/lib/python3.5/dist-包/Tensorflow/python/data/ops/dataset_ops.py"，第712行，在地图中返回MapDataset（自，map_func）File"/usr/loc/lib/python3.5/dist-包/Tensorflow/python/data/ops/dataset_ops.py"，第1385行，在初始化自。_map_func.add_to_graph（ops.get_default_graph（））file"/usr/loce/lib/python3.5/dist-pack/tensorflow/python/框架/function.py"，第486行，add_to_graph。_create_definition_if_needed（）File"/usr/loce/lib/python3.5/dist-pack/tensorflow/python/框架/function.py"，第321行，在_create_definition_if_needed._create_definition_if_needed_impl（）File"/usr/loce/lib/python3.5/dist-包/tensorflow/python/框架/function.py"，第338行，_create_definition_if_needed_impl输出=自己。_func（*输入）File"/usr/loce/lib/python3.5/dist-包/tensorflow/python/data/ops/dataset_ops.py"，第1376行，tf_map_funcflattened_ret=[ops.convert_to_tensor（t）用于nest.flatten中的t]File"/usr/local/lib/python3.5/dist-包/tensorflow/python/data/ops/dataset_ops.py"，第1376行，flattened_ret = [ops.convert_to_tensor（t）用于nest.flatten中的t（ret）]File"/usr/local/lib/python3.5/dist-pack/tenorflow/python/框架/ops.py"，第836行，convert_to_tensoras_ref=False）File"/usr/local/lib/python3.5/dist-pack/tenorflow/python/框架/ops.py"，第926行，在internal_convert_to_tensorret=conversion_func（value，dtype=dtype，name=name，as_ref=as_ref）File"/usr/local/lib/python3.5/dist-包/tenstorflow/python/框架/constant_op.py"，第229行，在_constant_tensor_conversion_function返回常量（v，dtype=dtype，name=name）File"/usr/local/lib/python3.5/dist-包/tenorflow/python/框架/constant_op.py"，第208行，常量值，dtype=dtype，形状=形状，verify_shape=verify_shape））File"/usr/local/lib/python3.5/dist-包/tenorflow/python/框架/tensor_util. py"，第472行，在make_tensor_proto"支持的类型."%（类型（值），值））TypeError：无法将类型的对象转换为张量。内容：稀疏张量（索引=张量（ParseSingle示例/Slice_Indices_i： 0，形状=（？，1），dtype=int64），值=张量（ParseSingle示例/Parse示例/Parse示例： 3，形状=(?,), dtype=int64），dense_shape=张量（ParseSingle示例/Squeeze_Shape_i： 0，形状=（1，），dtype=int64））。考虑将元素转换为支持的类型。

Python版本：3.5.2
Tensorflow版本：1.4.1

Answer 1

错误很简单。您的数据不是FixedLenFeature而是VarLenFeature。替换您的线路：

 'data':tf.FixedLenFeature([], tf.int64)}

与

 'data':tf.VarLenFeature(tf.int64)}

另外，当您调用print（i.eval（））和print（data.eval（））时，您将调用迭代器两次。第一个print将打印0，但第二个将打印第二行的值[0,50,89,147,196]。您可以执行print（sess.run（[i，data]））从同一行获取i和数据。

Answer 2

经过数小时的搜索和尝试，我相信答案终于出现了。下面是我的代码。

def _int64_feature(value):
    # value must be a numpy array.
    return tf.train.Feature(int64_list=tf.train.Int64List(value=value.flatten()))

# Write an array to TFrecord.
# a is an array which contains lists of variant length.
a = np.array([[0, 54, 91, 153, 177],
              [0, 50, 89, 147, 196],
              [0, 38, 79, 157],
              [0, 49, 89, 147, 177],
              [0, 32, 73, 145]])

writer = tf.python_io.TFRecordWriter('file')

for i in range(a.shape[0]): # i = 0 ~ 4
    x_train = np.array(a[i])
    feature = {'i'   : _int64_feature(np.array([i])), 
               'data': _int64_feature(x_train)}

    # Create an example protocol buffer
    example = tf.train.Example(features=tf.train.Features(feature=feature))

    # Serialize to string and write on the file
    writer.write(example.SerializeToString())

writer.close()

# Check TFRocord file.
record_iterator = tf.python_io.tf_record_iterator(path='file')
for string_record in record_iterator:
    example = tf.train.Example()
    example.ParseFromString(string_record)

    i = (example.features.feature['i'].int64_list.value)
    data = (example.features.feature['data'].int64_list.value)
    print(i, data)

# Use Dataset API to read the TFRecord file.
filenames = ["file"]
dataset = tf.data.TFRecordDataset(filenames)
def _parse_function(example_proto):
    keys_to_features = {'i':tf.VarLenFeature(tf.int64),
                        'data':tf.VarLenFeature(tf.int64)}
    parsed_features = tf.parse_single_example(example_proto, keys_to_features)
    return tf.sparse_tensor_to_dense(parsed_features['i']), \
           tf.sparse_tensor_to_dense(parsed_features['data'])
# Parse the record into tensors.
dataset = dataset.map(_parse_function)
# Shuffle the dataset
dataset = dataset.shuffle(buffer_size=1)
# Repeat the input indefinitly
dataset = dataset.repeat()  
# Generate batches
dataset = dataset.batch(1)
# Create a one-shot iterator
iterator = dataset.make_one_shot_iterator()
i, data = iterator.get_next()
with tf.Session() as sess:
    print(sess.run([i, data]))
    print(sess.run([i, data]))
    print(sess.run([i, data]))

有一些事情需要注意。
1.这个SO问题帮助很大。
2.tf.VarLenFeature将返回SparseTensor，因此，使用tf.sparse_tensor_to_dense转换为密集张量是必要的。在我的代码中，parse_single_example（）不能用parse_example（）替换，它困扰了我一天。我不知道为什么parse_example（）不起作用。如果有人知道原因，请指教。