Port Data Parallel Training example to multi-backend

[shivance]

Link : https://github.com/keras-team/keras-io/blob/master/examples/nlp/data_parallel_training_with_keras_nlp.py

The tutorial was originally written in Keras, we need to port this to Keras Core with multi backend support.

[rajchordia14]

Can I work on this issue?

[shivance]

Sure @rajchordia14 ! You are welcome to contribute. Please open a PR at keras-team/keras-io ! We can use this thread to discuss if you need any help!

[shivance]

Hi @rajchordia14 any update? are you stuck somewhere and need help?

[rajchordia14]

Working on it, first ever contribution might take time. Apologies for delay

[pranavvp16]

@rajchordia14 are you still working on this or can I take over and start working on it

[shivance]

@pranavvp16 you may start working on this if you still are interested. Please try to come up with a colab. Then we could iterate on it.

[pranavvp16]

fine I'll start working today

[pranavvp16]

@shivance I changed the imports according to the new keras-core and ran the example, the example ran fine with pytroch as backend but when backend is set to tensorflow there's an error. Below is the link to collab notebook pytorch as backend tensorflow as backend Seems there is something missing in optimizers file in tensorflow backend

[shivance]

This is strange, can you open a seperate issue to file this bug?

[pranavvp16]

I should be opening the issue in keras-core repository right ?

[shivance]

keras-nlp for now

[pranavvp16]

@shivance can I make the PR for converted code sample as it works with other backends, and once the above issue is fixed it will work with tensroflow too

[shivance]

Yes, please go ahead! Please open seperate issues for each bug you face. Keras team tracks all the work through github issues.

[github-actions[bot]]

This issue is stale because it has been open for 180 days with no activity. It will be closed if no further activity occurs. Thank you.