Custom REST end point is not called at all from doccano auto labelling

Question

I have filed an issue in the official doccano repo. Here . However I am posting here also in the hope of getting some idea on what I am doing wrong.

1. I have two EC2 instances both running Ubuntu 20

2. In one of them I have set up doccano and uploaded some data

3. I annotated a bit of that data and then trained a custom model using Hgging Face.

4. In the second EC2 instance I have uploaded the trained model and created a FastAPI based API to serve the result.

5. I want to set up auto labeling (it is a Sequence Labeling project).

6. I follow the steps in the official document and also take help from here .

7. Everything goes right, including at the second step when I am testing the api connectivity doccano could successfully connect and fetch the data.

8. Once all is done I go to one of the documents and try to do the auto labeling. And Surprise

9. NOTHING HAPPENS.

10. There is no log in the model server showing that no request has ever reached there!

11. Both the doccano and the model server and running via Docker inside the EC2 instances.

12. What am I doing wrong?

Please help.

Warm regards

Answer 1

Ok I found the reason. Thanks to a reply in the Github issue. Here I am going to quote the reply as it is.

The website is not behaving intuitively from the UX perspective. When you turn on Auto Labeling, just try going to the next example (arrow on the top right of the page) and your Auto Labeling API should be called. Then go back and it will be called on the first example. Also, it's called only on examples that are not marked as "labeled".

So anyone having difficulty with the same problem hopefully you will get some help from here.