[BEAM-10996] AssertionError: (10, <class 'int'>) when writing TF Records

• Type: Bug
• Status: Resolved
• Priority: P1
• Resolution: Unresolved
• Affects Version/s: 2.22.0
• Component/s: sdk-py-core
• Assignee: Unassigned
• Reporter: Valliappa Lakshmanan
• Created: 01/Oct/20 00:11
• Updated: 06/Oct/20 04:13

Description

This code snippet:

 

def create_tfrecord:
  size = np.array([2.0, 3.0])
  tfexample = tf.train.Example( 
     features=tf.train.Features(
        feature=

{            'size': tf.train.Feature(float_list=tf.train.FloatList(value=size))    }

))
  return tfexample.SerializeToString()

 

...

beam.FlatMap(lambda x: create_tfrecord)

...

 

throws this error:

 

Traceback (most recent call last): File "apache_beam/runners/common.py", line 961, in apache_beam.runners.common.DoFnRunner.process File "apache_beam/runners/common.py", line 726, in apache_beam.runners.common.PerWindowInvoker.invoke_process File "apache_beam/runners/common.py", line 814, in apache_beam.runners.common.PerWindowInvoker._invoke_process_per_window File "/opt/conda/lib/python3.7/site-packages/apache_beam/io/iobase.py", line 1061, in process self.writer.write(element) File "/opt/conda/lib/python3.7/site-packages/apache_beam/io/filebasedsink.py", line 420, in write self.sink.write_record(self.temp_handle, value) File "/opt/conda/lib/python3.7/site-packages/apache_beam/io/filebasedsink.py", line 146, in write_record self.write_encoded_record(file_handle, self.coder.encode(value)) File "/opt/conda/lib/python3.7/site-packages/apache_beam/coders/coders.py", line 463, in encode return self.get_impl().encode(value) File "apache_beam/coders/coder_impl.py", line 494, in apache_beam.coders.coder_impl.BytesCoderImpl.encode File "apache_beam/coders/coder_impl.py", line 495, in apache_beam.coders.coder_impl.BytesCoderImpl.encode AssertionError: (10, <class 'int'>)