Implement beam search

[abelriboulot]

A next step for better generation is to implement a beam search for the generation. An example of it can be seen on the huggingface repo here, and this would need adding such a function to the GenerativeT5 model in onnxt5/models.py

[patil-suraj]

Yes, beam search is especially useful for summarization, translation, and other fixed length generation tasks. Also all the generative results in the T5 paper are given using beam search.

And thank you for building awesome project! This is going to be super useful as people are using T5 for lot's of tasks both generative and discriminative.

[abelriboulot]

Thanks for the help on the generation patil, that was very useful!