ocr-T5 / README.md
GaborMadarasz's picture
Update README.md
71a5729 verified
|
raw
history blame contribute delete
No virus
316 Bytes
metadata
language:
  - hu
widget:
  - text: 'ocr: A mútt hé ten még gyengütt a magyar fizetóeszköz az euróval szemben.'
license: apache-2.0
metrics:
  - bleu

An mT5-large model finetuned for Post-OCR correcting Hungarian texts.

max_token = 512 (preferably just one sentence)

prefix: "ocr: "

More details later. :)