This repository contains the code used to conduct fine-tuning on OrthodonticQA (OQA).
To download the assets (dataset and pretrained model weights) required to reproduce the experiments use the following command.
python etc/build_cache.py assests.toml
We use TOML to configure the hyper-parameters for the run. The config/
directory contains the various configurations for the experiments from the
paper.
scripts/ contains examples for usage with a SLURM cluster.
To launch a training run:
python main.py configs/pubmedbert_sft.toml