I have been playing with this model and ran into the following issue
#1
by
chrisxx
- opened
I created this collab: https://colab.research.google.com/drive/1TO10E5fa9KeVyHQBhQQP3VESeigRTcsG?usp=sharing
and for some reason cannot get TrOCRProcessor.from_pretrained("microsoft/trocr-small-printed") to work (as compared to run.py where this is used).
Any ideas?
Have you solved the issue now?
I am not 100% sure. I still cannot load trocr-small-printed but I can load base and large. The model runs but the results are not that great.
Do you maybe have also trained a larger model?
No, I haven't trained a larger one. By the way, could you tell me why you use "microsoft/trocr-small-printed" instead of "yhshin/latex-ocr" here? I am confused about this.
Because your repository does not contain a Processor.