Instruction Fine-Tuning Approach
#2
by
bastistrauss
- opened
Unfortunately, I cant seem to find the instruction fine-tuning approach used in this model, does it use LoRa like in the first version of Med42? Would be very grateful for an answer!
Thank you for your interest in Med42.
For v2 we fully finetuned 8B and 70B. You'll find every detail for training as well as the list of public datasets in this paper: https://arxiv.org/pdf/2408.06142
bastistrauss
changed discussion status to
closed