AttributeError: module 'tensorflow' has no attribute 'estimator'

I have py36-tensorflow-1.14.0 and py36-tensorflow-estimator-1.14.0_1 installed (both are latest versions), but tensorboard fails:

$ tensorboard < events.out.tfevents.1568855561.yv.noip.me
Traceback (most recent call last):
  File "/usr/local/bin/tensorboard", line 6, in <module>
    from tensorboard.main import run_main
  File "/usr/local/lib/python3.6/site-packages/tensorboard/main.py", line 40, in <module>
    from tensorboard import default
  File "/usr/local/lib/python3.6/site-packages/tensorboard/default.py", line 39, in <module>
    from tensorboard.plugins.beholder import beholder_plugin_loader
  File "/usr/local/lib/python3.6/site-packages/tensorboard/plugins/beholder/__init__.py", line 22, in <module>
    from tensorboard.plugins.beholder.beholder import Beholder
  File "/usr/local/lib/python3.6/site-packages/tensorboard/plugins/beholder/beholder.py", line 199, in <module>
    class BeholderHook(tf.estimator.SessionRunHook):
  File "/usr/local/lib/python3.6/site-packages/tensorflow/python/util/deprecation_wrapper.py", line 106, in __getattr__
    attr = getattr(self._dw_wrapped_module, name)
AttributeError: module 'tensorflow' has no attribute 'estimator'

tensorboard is of version 2.0.0, also the latest from PyPI.

So tensorboard is broken, fails to run.

[wchargin]

If TensorFlow is at one stable release number, you should keep TensorBoard’s release compatible with the same TF release.

This is the case: TensorBoard 2.0 and TensorFlow 2.0 are compatible.

Otherwise, how would users figure out what is compatible with what?

For any given minor series x.y, TensorBoard x.y is compatible with TensorFlow x.y, and no other guarantees are made.

It is extremely difficult to build a new version of TF from scratch

You don’t need to build TensorFlow from scratch. (I don’t usually—it takes a long time.) Release candidate binaries are provided on PyPI.

We don’t control what packages are available in the FreeBSD repositories, so I can’t advise you there.

(Bazel also fails to build TB from source.) In fact. Bazel doesn’t work well outside of Google on any project.

This is not true. I can build TensorBoard on my personal laptop just fine. Also, all daily development of TensorBoard is done using public Bazel binaries, building the public TensorBoard source tree, in exactly the same way that is specified in the documentation on this repository. We receive contributions from third-party contributors who successfully use Bazel to build and test their code.

In the light of the above, could you please tell me how can I follow your suggestion?

pip uninstall \
    tensorflow tf-nightly tf-nightly-2.0-preview tensorflow-estimator \
    tensorflow-estimator-nightly tensorflow-estimator-2.0-preview \
    tensorboard tb-nightly \
    &&
pip install --upgrade pip &&
pip install tensorflow==2.0.0rc1 &&
pip uninstall -y tb-nightly tensorboard &&
pip install tensorboard==2.0.0

[wchargin]

Please feel free to send patches that improve support for FreeBSD. We’d be happy to accept them, provided that they don’t complicate the overall system too much. Bazel also has “community support” for FreeBSD, so I suspect that they have a similar policy.