[ARROW-2805] [Python] TensorFlow import workaround not working with tensorflow-gpu if CUDA is not installed

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.10.0
• Component/s: Python
• Labels: pull-request-available tensorflow
• External issue URL: https://github.com/apache/arrow/issues/19184
• Assignee: Philipp Moritz
• Reporter: Philipp Moritz
• Created: 06/Jul/18 22:10
• Updated: 11/Jan/23 07:23
• Resolved: 07/Jul/18 03:58

Description

TensorFlow version: 1.7 (GPU enabled but CUDA is not installed)

tensorflow-gpu was installed via pip install

```

import ray
File "/home/eric/Desktop/ray-private/python/ray/_init_.py", line 28, in <module>
import pyarrow # noqa: F401
File "/home/eric/Desktop/ray-private/python/ray/pyarrow_files/pyarrow/_init_.py", line 55, in <module>
compat.import_tensorflow_extension()
File "/home/eric/Desktop/ray-private/python/ray/pyarrow_files/pyarrow/compat.py", line 193, in import_tensorflow_extension
ctypes.CDLL(ext)
File "/usr/lib/python3.5/ctypes/_init.py", line 347, in __init_
self._handle = _dlopen(self._name, mode)
OSError: libcublas.so.9.0: cannot open shared object file: No such file or directory

```

Issue Links

• links to: GitHub Pull Request #2224