Infercode fine-tuning

Thank you for our awesome work, i can run it successfully. Now i want to fine-tune Infercode model for a downstream task (**classification of programs by their functionalities**). May I know whether it is the right way to go? If so, would you mind sharing some ideas/insights about how the architecture of the new model looks like?