The docs for this PR live here. All of your documentation changes will be reflected on that endpoint.

Hi @younesbelkada !
I have made two additional changes to the branch to solve the test problems.
For the second change especially:

1. I created a tiny ast test model, tiny-random-ast
   Compared with the original ast model, I did the following changes :

• changed hidden_size from 768 to 32
• changed intermediate_size from 3072 to 37
• changed patch_size from 16 to 1
• changed num_attention_heads from 12 to 4
• changed num_hidden_layers from 12 to 5

2. I separated the ast model from BetterTransformersAudioTest class and created another class BetterTransformersASTTest in test_bettertransformer_audio.py
   file to test the ast model.
   This exactly solved the test problems since the ast model does not have the processor as other audio models do. Instead, it uses featureExtractor, which has similar functionality.

Please let me know what you think about these changes and let me what else I need to do on this topic.

Thank you very much!

Hi @ravenouse !
This is great! Thanks for adding this, could you please rebase with main branch so that we can merge this PR? 🙏
Thanks!

Hi @younesbelkada ! Happy New Year!

I have just rebased this branch with the current optimum main branch, and I re-run the test, pytest tests/bettertransformer/test_bettertransformer_audio.py::BetterTransformersASTTest.
The tests went well this time, except for one error with collections packages.
I believe this is caused by the version of python. I run the test with python 3.10.4.

Please let me know what else I can do!

Thank you so much!

Hi @younesbelkada @fxmarty!

I did rebase some time ago. Could you spare some time to review this?
I deeply appreciate all the help!
Thank you so much!

This PR has been marked as stale because it has been open for 90 days with no activity. This thread will be automatically closed in 30 days if no further activity occurs.