Investigate Integration with tensorflow

[parmidaatg]

Is your feature request related to a problem? Please describe. As per conversation with @tboquet, this feature could help him working with BaaL in his research at Unity.

Describe the solution you'd like A clear and concise description of what you want to happen.

Describe alternatives you've considered A clear and concise description of any alternative solutions or features you've considered.

Additional context Add any other context or screenshots about the feature request here.

[Dref360]

Do we know what is missing? BALD is in numpy, I do not know how to have a patch_module for TF, but we could do without for now.

[tboquet]

We might want to have a look at how the transformers repo at hugging works w.r.t having 2 or more backends in the future:

• https://github.com/huggingface/transformers/blob/1a3deae8202bb9d64ef9419ca18f7392255b9005/src/transformers/file_utils.py#L80-L111
• https://github.com/huggingface/transformers/blob/1a3deae8202bb9d64ef9419ca18f7392255b9005/src/transformers/file_utils.py#L80-L111
• https://github.com/huggingface/transformers/blob/75e63dbf70d566f6be122b16bb2fb22e1bb6e558/src/transformers/models/mt5/init.py#L48-L72

But we can do without right? There would be a model wrapper to implement and the dirichlet calibration?

[parmidaatg]

yes mainly model_wrapper, and patch_module (this is the main one which changes droput layer) and dirichlet. the heuristics are numpy so we are good