TensorFlow 2.6 upgrade

[anargyri]

Description

Changes required for upgrade to TensorFlow 2.6. This is a migration of the existing code to TF2 keeping the v1 style and using the tf.compat.v1 API. See https://www.tensorflow.org/guide/migrate/upgrade and https://www.tensorflow.org/guide/migrate.

In detail

• Ran update script provided by TensorFlow that replaces syntax and fixed any errors caused.
• Replaced references from tf.contrib (now deprecated) with other modules or packages
• Edited setup.py, conda script and documentation.
• Made changes to some classes (e.g. DKN, LSTUR) to work.
• Replaced tf.summary, tf.logging, hparams.
• Replaced batch_dot() to fix self attention layer.
• Replaced optimizers that were not working e.g. LazyAdam.
• Replaced deprecated imports of RNN cells.
• Replaced fully connected layer with tf-slim implementation.
• Pinned commit from pymanopt in order to fix GeoIMC.
• Removed batch inference step from SLIRec notebook.

Related Issues

Checklist:

• I have followed the contribution guidelines and code style for this project.
• I have added tests covering my contributions.
• I have updated the documentation accordingly.
• This PR is being made to staging branch and not to main branch.

[Leavingseason]

Great job! I have tested some deeprec-related jupyter-notebooks, and everything is running normally.

[laserprec]

This is awesome! ❤️

[miguelgfierro]

@anargyri, it's weird, do you know why the ADO pipelines are not being executed?

[anargyri]

@anargyri, it's weird, do you know why the ADO pipelines are not being executed?

The GitHub hosts don't have CUDA 11 installed. About spark, I am not sure, I am trying to figure what happens.

[miguelgfierro]

The GitHub hosts don't have CUDA 11 installed. About spark, I am not sure, I am trying to figure what happens.

But the ADO pipelines should execute. This is weird

[anargyri]

The GitHub hosts don't have CUDA 11 installed. About spark, I am not sure, I am trying to figure what happens.

But the ADO pipelines should execute. This is weird

Ah sorry, I missed the "ADO" in your question. Yes, it is strange, let me double check.