Skip to content

TF_ADDONS_PY_OPS does not appear to work with beam_search_decoder.py unit test #2104

@jimthompson5802

Description

@jimthompson5802

System information

  • OS Platform and Distribution (e.g., Linux Ubuntu 16.04): MacOS 10.15.6 running Docker for Mac 2.3.0.4
  • TensorFlow version and how it was installed (source or binary): TF 2.3.0, from TF docker image tensorflow/tensorflow:2.3.0-jupyter
  • TensorFlow-Addons version and how it was installed (source or binary): In docker build, pip install tfa-nightly==0.12.0.dev20200809083253
  • Python version: 3.6.9
  • Is GPU used? (yes/no): No

Describe the bug
Several failures in unit test tensorflow_addons/seq2seq/test/beam_search_decoder_test.py. All appear to be this error

tensorflow.python.framework.errors_impl.NotFoundError: /opt/project/tensorflow_addons/custom_ops/seq2seq/_beam_search_ops.so: cannot open shared object file: No such file or directory

This error occurs even though environment variable TF_ADDONS_PY_OPS=1 is set.

Based on TFA documention I expected the unit test to complete successfully.

Code to reproduce the issue

  • git clone the TFA github
  • build docker image with TF 2.3.0 and TFA from nightly build. build command: docker build --progress auto -t addons:tf23 -f Dockerfile .
    Dockerfile
  • cd position into root directory of the TFA local repo
  • start container docker run -it -v ${PWD}:/opt/project addons:tf23 /bin/bash
  • within the container execute this command:
TF_ADDONS_PY_OPS=1  pytest -v /opt/project/tensorflow_addons/seq2seq/tests/beam_search_decoder_test.py

Other info / logs

Log file
containing all the output.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions