Error when training a transcription model

Hi Teklia team,

I have tried to train a transcription model with the updated version of the PyLaia Training worker, but this results in a 400 Client Error. I have tried it with two different datasets and also with another base model. What could be the issue here?

Thank you for your help!
Marthe

Hi @marthelamote,

Thanks for reaching out to us!

It is indeed an error on our side, that has been fixed in the 0.9.4 version of the worker. Thanks for the report and please try again with this new version.

Have a greay day!


Yoann Schneider

Hi,

I have tried it again and now I have another error:


What could be the issue here?

Kind regards,
Marthe

Hi Marthe,

your dataset is not ready for the training, it must be in “complete” state. I think that it is still open

The goal of make the dataset “complete” is to have a clear reference on which data was used to train the model.

You need to follow this step :

https://doc.arkindex.org/tutorial/segmentation-training/#make-your-dataset-immutable

Best,

Christopher