Package Details: python-tensorflow-probability 0.24.0-1

Git Clone URL: https://aur.archlinux.org/python-tensorflow-probability.git (read-only, click to copy)
Package Base: python-tensorflow-probability
Description: Probabilistic reasoning and statistical analysis in TensorFlow
Upstream URL: https://github.com/tensorflow/probability
Keywords: tensorflow
Licenses: Apache
Submitter: oysstu
Maintainer: oysstu
Last Packager: oysstu
Votes: 5
Popularity: 0.068122
First Submitted: 2019-02-20 14:59 (UTC)
Last Updated: 2024-03-13 12:53 (UTC)

Latest Comments

1 2 Next › Last »

oysstu commented on 2024-05-16 01:44 (UTC)

@pirofti you should upgrade to python 3.12 and rebuild the package. I don't understand why you are breaking on the exception though, are you running through a debugger? That exception should be caught and ignored afterwards.

pirofti commented on 2024-05-15 10:31 (UTC)

I get the following error at import:

File /usr/lib/python3.11/site-packages/tensorflow_probability/python/__init__.py:79, in _validate_tf_environment(package)                
     77 if required_tensorflow_version[0] == '2':                                                                                        
     78   try:                                                                                                                           
---> 79     import tf_keras  # pylint: disable=unused-import
     80   except (ImportError, ModuleNotFoundError):
     81     # Print more informative error message, then reraise.
     82     print('\n\nFailed to import TF-Keras. Please note that TF-Keras is not '
     83           'installed by default when you install TensorFlow Probability. '
     84           'This is so that JAX-only users do not have to install TensorFlow '
   (...)                          
     87           'This can be be done through installing the '
     88           'tensorflow-probability[tf] extra.\n\n')

ModuleNotFoundError: No module named 'tf_keras'

oysstu commented on 2024-03-13 12:55 (UTC) (edited on 2024-03-13 12:56 (UTC) by oysstu)

Looks like tf has become an optional dependency upstream, with the possibility of using jax as a backend instead. I haven't used tfp for a long time, so I'm not sure exactly how that works. I will keep tf as a dependency for now, as I imagine that's how most of you are using tfp, but it's possible to have tf/keras as optional dependencies instead if that makes more sense. Let me know.

carsme commented on 2024-02-14 14:58 (UTC)

Much appreciated, thanks!

oysstu commented on 2024-02-14 12:29 (UTC)

Fixed. Looks like the latest bazel version moved setup.py to a new location.

carsme commented on 2024-02-14 11:05 (UTC)

I'm running in to the following issue when building the package:

==> Starting build()...
WARNING: The following rc files are no longer being read, please transfer their contents or import their path into one of the standard rc files:
/tmp/makepkg/python-tensorflow-probability/src/probability-0.23.0/tools/bazel.rc
INFO: Analyzed target //:pip_pkg (64 packages loaded, 4766 targets configured).
INFO: Found 1 target...
Target //:pip_pkg up-to-date:
  bazel-bin/pip_pkg
INFO: Elapsed time: 26.941s, Critical Path: 0.42s
INFO: 941 processes: 4 internal, 937 linux-sandbox.
INFO: Build completed successfully, 941 total actions
python: can't open file '/home/carsme/.cache/bazel/_bazel_carsme/bfec90a73c6d9abcd351097554995d45/execroot/_main/bazel-out/k8-fastbuild/bin/pip_pkg.runfiles/tensorflow_probability/setup.py': [Errno 2] No such file or directory

Any ideas?

oysstu commented on 2020-07-25 15:15 (UTC)

Well, the arch wiki is pretty clear that the packages should track stable releases unless there are security issues or breakage that cannot easily be backported. Not sure why the arch tensorflow package tracks the release candidates. As it stands, I would rather track the stable releases and let those interested in the most recent features upgrade on their own, but I'm not opposed to upgrading to the tfp-rc if any breakage is found since it is tested against the tf-rc.

juacrumar commented on 2020-07-25 11:13 (UTC)

Wasn't sure wether to flag the package out of date. Technically the package is not out-of-date, but since Arch has updated now python-tensorflow to 2.3rc2, you'll need version 0.11rc1 of this package in order to run. I've installed: https://github.com/tensorflow/probability/archive/v0.11.0-rc1.tar.gz sha256sum: 24ceceffc3d1380d8aef16e7632f2ef856f94a480ac33a6fd670a91e6fec5d34 Also, now there is also a dependence on dm-tree.

oysstu commented on 2020-02-07 09:01 (UTC) (edited on 2020-02-07 09:02 (UTC) by oysstu)

Bazel requires OpenJDK 11, so the installation would fail on systems that had a different java environment set as default. I've updated the PKGBUILD to point to the specific version of java.

Bazel is unfortunately required, as it generates a bunch of python files before packaging with setuptools

wjhandley commented on 2020-02-07 04:01 (UTC) (edited on 2020-02-07 04:02 (UTC) by wjhandley)

I get the error:

==> Starting build()...
WARNING: The following rc files are no longer being read, please transfer their contents or import their path into one of the standard rc files:
/home/will/.cache/yay/python-tensorflow-probability/src/probability-0.9.0/tools/bazel.rc
Extracting Bazel installation...
FATAL: Could not find system javabase. Ensure JAVA_HOME is set, or javac is on your PATH.