OCR models demo fails when improving model

neulab / cmulab

CMU Linguistic Annotation Backend

14 stars 1 forks source link

OCR models demo fails when improving model #35

Open oserikov opened 4 months ago

oserikov commented 4 months ago

In the online demo, OCR models aren't trainable. Hello! I was so happy to see your project allowing for SotA AI tools to be used in field lingiustics. However, when I tried to run the OCR model post-correction, I got errors in the logfile instead of the model. Please find it available on : https://cmulab.dev/annotator/media/oleg_model_log.txt.

zaidsheikh commented 3 months ago

Thanks for trying it out! For training, the OCR model requires a minimum of 10 source-target file pairs. We will update the code to show more user-friendly error messages!

oserikov commented 3 months ago

I see. Thank you! We are going to have a field trip soon and might be able to test it in the field. I will try to configure it properly. Ty