spark 读取 tfrecord文件，解析tf.train.Feature(bytes_list=tf.train.BytesList(np.array(feature, dtype=np.float32).tobytes())) 文件字节码异常

Nobody has claimed this yet.

1. 如下图所示：用spark直接去读取tfrecord数据，string类型的字段可以直接读取，但是adc_traj等numpy.array的数据会是乱码的形式。

2. 原因是spark加载tfrecord数据的时候，会默认把bytes类型的数据转成str，目前不清楚字符编码。而adc_traj存储的时候是用tf.train.Feature(bytes_list=tf.train.BytesList(np.array(feature, dtype=np.float32).tobytes()))方式存储的。
3. 用 np.frombuffer(str_y., dtype=np.float32)去解析数据
数据还是会报错：
ValueError: buffer size must be a multiple of element size

No source file or test is identified. Reproduce the Spark TFRecord read using the bytes_list value created from a float32 NumPy array, then trace where bytes become a string and compare the original byte length with np.frombuffer's input. Done means binary feature data can be decoded without the buffer-size error, with a regression test covering this case.

Written by the indexing model from the issue text.