Can't import TQDMProgressBar

[zaccharieramzi]

System information

• OS Platform and Distribution: Linux Ubuntu 16.04
• TensorFlow version and how it was installed: pip (2.0.0)
• TensorFlow-Addons version and how it was installed: pip install -q --no-deps tensorflow-addons~=0.6
• Python version: 3.6.8
• Is GPU used?: yes

Describe the bug

I can't import TQDMProgressBar.

Code to reproduce the issue

import tensorflow_addons as tfa
print(tfa.__version__)
tfa.callbacks.TQDMProgressBar

Other info / logs
This issue appeared when discussing with @shun-lin , I originally described it in tensorflow/tensorflow#34417, but felt it made more sense to discuss it here.
Apparently I could install from source to try and solve it, but it's not in my priorities right now.

[shun-lin]

@seanpmorgan is it because it will be included in the next release but not the current one? Thanks!

[Squadrick]

@zaccharieramzi Like @shun-lin mentioned, TQDMProgressBar isn't included in r0.6. You can either build TFA from source or install tfa-nightly.

[zaccharieramzi]

Sure makes sense. Probably the notebook introducing the TQMProgressBar needs to be updated to reflect that, because it says that you can use it with 0.6.

[zaccharieramzi]

I also have this other error with tfa-nightly (without dependencies):

AttributeError: module 'tensorflow_core.keras.utils' has no attribute 'register_keras_serializable'

If I install with dependencies:

NotFoundError: /mypath/tests/venv/lib/python3.6/site-packages/tensorflow_addons/custom_ops/text/_parse_time_op.so: undefined symbol: _ZN4absl9ParseTimeERKSsS1_PNS_4TimeEPSs

[Squadrick]

We currently have a breakage due to upstream tf-nightly. We're working on resolving it: #709 #710

In the meantime, you can install tf-nightly==2.1.0.dev20191119.