Would be nice if you can add some examples for fine-tuning for example with any pretrained bert as decoder !? :) Do we have also a chance to export these after training to ONNX ? However, I think even if then only with a own greedy search implementation

I have started to play a bit with it but struggle currently with the greedy search i shared my questions and notebook also in huggingface forum: Question

Thanks a lot :)

Hi @felixdittrich92

Do you know if there is any news about it? I am trying to create a training pipeline for different languages and some examples would be very helpful.
How many examples do you think would be useful per fine-tuning in a new language?

Thanks a lot!