using unsloth mistral instruct-tuned, finetuning the model on new data that contain question, student answer, reference answer, and mark scheme and the model generate the rational and score

Finetuning unsloth model(instruct-tuned) from huggingface to generate score and Rational