BERT Multi input - Multi output

Hello everybody,

I have been trying to implement a Fastai BERT multi-task text classification model which takes two inputs into “bert-base-uncased”, do some operations on the final pooling layer outputs, get a shared representation and then classifies the shared representation into two outputs, which are actually for a multi-labeled. Each classifier is a multi-label classifier.

I am struggling in creating the necessary databunch and learner object required to run my model using Fastai. I am not skilled at creating custom databunch functions or learner functions.

Please advice on what I should do.