Request: DOI

#9
by sandeepmbm - opened

Could you please help me to train the layoutlm-document-qa model .
Any documentation or guideline to follow the steps with training code module which I can modify for research purpose.
I am a PhD scholar and trying to give my contribution to improve the accuracy and performance.
Kindly write to [email protected]

Impira org

Hi @sandeepmbm this project is no longer actively developed but we'd welcome any contributions you have! The training code is roughly adapted from @nielsr 's excellent notebooks here: https://github.com/NielsRogge/Transformers-Tutorials/tree/master/LayoutLMv2/DocVQA

Hi @sandeepmbm , I am trying to train this model too.
Could you share the process for training ?
Would be really helpful

@ankrgyl I am trying to fine tune this model for visual question answering task, but unable to find a relevant script to do the same. I tried using Niels' LMv2 script but realized that layoutLM uses a different tokenizer and model architecture hence cannot be trained using that notebook. The pre-existing notebooks for LayoutLM focus on token and sequence classification rather than question answering. It would be really helpful if you could share some relevant code for fine tuning this model on question answering.

This comment has been hidden
sandeepmbm changed discussion status to closed
sandeepmbm changed discussion status to open

Hi @ankrgyl ,
I am trying to fine tune this model for visual question answering task, but unable to find a relevant script to do the same. I tried using Niels' LMv2 script but in output it gave me answer in chunks of data.
It would be really helpful if you could share some relevant code for fine tuning this model on question answering.

Sign up or log in to comment