Example's on a custom dataset
The repo shows how to fine tune on standard datasets but not how to run on your own data this limits the utility of the project.
This depends on the specific tasks (which needs you to adapt the model a bit). That's why we are preparing for the public research dataset and you can adapt based on your particular data.
That is a fair point, but I think it would be useful to provide a link to an annotation tool for generating text in the required format. The other thing that is missing for many of examples is how to deploy the model for evaluation. I have provided a baseline example of this for NER in my pull request would be happy to contribute to the other notebooks if it's deemed worthwhile.
@xiaoyongzhu , @aribornstein are there any plans to address this open issue?