decode_flac not working in tfio-nightly 0.18.0 and tf-nightly 2.5.0

[lminer]

I'm trying to get the nightly version of tensorflow 2.5.0-dev to work with the nightly version of tensorflow-io.

However, when running a simply tf.data dataset that incorporates tfio.audio.decode_flac, I get a NotImplementedError.

Versions are:

• tb-nightly-2.5.0a20210322
• tensorflow-io-nightly-0.18.0.dev20210318195435
• tf-estimator-nightly-2.5.0.dev2021032201
• tf-nightly-2.5.0.dev20210322

Error is below:

../lib/python3.7/site-packages/tensorflow/python/data/ops/dataset_ops.py:1931: in map
    preserve_cardinality=True)

../lib/python3.7/site-packages/tensorflow/python/data/ops/dataset_ops.py:4520: in __init__
    use_legacy_function=use_legacy_function)

../lib/python3.7/site-packages/tensorflow/python/data/ops/dataset_ops.py:3706: in __init__
    self._function = fn_factory()

../lib/python3.7/site-packages/tensorflow/python/eager/function.py:3130: in get_concrete_function
    *args, **kwargs)

../lib/python3.7/site-packages/tensorflow/python/eager/function.py:3095: in _get_concrete_function_garbage_collected
    graph_function, _ = self._maybe_define_function(args, kwargs)

../lib/python3.7/site-packages/tensorflow/python/eager/function.py:3439: in _maybe_define_function
    graph_function = self._create_graph_function(args, kwargs)

../lib/python3.7/site-packages/tensorflow/python/eager/function.py:3284: in _create_graph_function
    capture_by_value=self._capture_by_value),

../lib/python3.7/site-packages/tensorflow/python/framework/func_graph.py:998: in func_graph_from_py_func
    func_outputs = python_func(*func_args, **func_kwargs)

../lib/python3.7/site-packages/tensorflow/python/data/ops/dataset_ops.py:3681: in wrapped_fn
    ret = wrapper_helper(*args)

../lib/python3.7/site-packages/tensorflow/python/data/ops/dataset_ops.py:3611: in wrapper_helper
    ret = autograph.tf_convert(self._func, ag_ctx)(*nested_args)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

args = (<tf.Tensor 'args_0:0' shape=() dtype=string>,), kwargs = {}
options = <tensorflow.python.autograph.core.converter.ConversionOptions object at 0x7f426f54f6d0>

    def wrapper(*args, **kwargs):
      """Wrapper that calls the converted version of f."""
      options = converter.ConversionOptions(
          recursive=recursive,
          user_requested=user_requested,
          optional_features=optional_features)
      try:
        with conversion_ctx:
          return converted_call(f, args, kwargs, options=options)
      except Exception as e:  # pylint:disable=broad-except
        if hasattr(e, 'ag_error_metadata'):
>         raise e.ag_error_metadata.to_exception(e)
E         NotImplementedError: in user code:
E         
E                 audio = tfio.audio.decode_flac(tensor, dtype=tf.int16)
E             
../lib/python3.7/site-packages/tensorflow_io/core/python/ops/audio_ops.py:105 decode_flac  **
E                 return core_ops.io_audio_decode_flac(input, shape=shape, dtype=dtype, name=name)
E             
../lib/python3.7/site-packages/tensorflow_io/core/python/ops/__init__.py:88 __getattr__
E                 return getattr(self._load(), attrb)
E             
../lib/python3.7/site-packages/tensorflow_io/core/python/ops/__init__.py:84 _load
E                 self._mod = _load_library(self._library)
E             
../lib/python3.7/site-packages/tensorflow_io/core/python/ops/__init__.py:71 _load_library
E                 + "{}, from paths: {}\ncaused by: {}".format(filename, filenames, errs)
E         
E             NotImplementedError: unable to open file: libtensorflow_io.so, from paths: ['
../lib/python3.7/site-packages/tensorflow_io/core/python/ops/libtensorflow_io.so']
E             caused by: ['
../lib/python3.7/site-packages/tensorflow_io/core/python/ops/libtensorflow_io.so: undefined symbol: _ZNK10tensorflow10FileSystem8BasenameEN4absl14lts_2020_02_2511string_viewE']


../lib/python3.7/site-packages/tensorflow/python/autograph/impl/api.py:695: NotImplementedError

[yongtang]

@lminer The issue is very likely the same issue as PR #1336. I tends to believe the latest nightly pip install tensorflow-io-nightly==0.18.0.dev20210323183735 fixed it. Can you give the new nightly a try?

[lminer]

That fixed it!