Access to models
#3
by
MathieuBartels
- opened
Dear authors,
What a great paper and you've localized a problem in document processing which is very costly in many sectors. I'm interested in your approach and have two questions.
Firstly I want to ask if the trained model is publicly available?
Secondly I was wondering how you deal with prompts larger than the token size?
Hi,
We don't have a trained model which is publicly available.
But we provided code to train one in: https://github.com/IBM/KVP10k
In this code we use Mistral 7B where the max token size is around 7000.
We don't support now prompts which are larger than the max token size.
ophir
changed discussion status to
closed
Thank you for your answer!