This repo contains a Python script to finetune the Phi 3 model on the CoEdit Dataset. The goal is to make the Phi 3 model work well on text editing in the English language by fine-tuning it on the Coedit Dataset.
The script is based on the notebook of Unsloth AI and was run on a Google Cloud Machine with L4 GPU. The checkpoint was saved here which was trained on full the CoEdit dataset on 2 epochs.