How return custom inference in AWS SageMaker without clone the repo?

Hi everyone!

I’m a begginer of Amazon SageMaker and I’m testing the this tool. I’ve fine-tuned a BERT model for sequence classification in which the text is classified in one of twenty classes (my reference was this notebook). I’ve follow this notebook so I defined an entrypoint and customized the file for the training.The problem is thaht the goal is to obtain the logits associated at each class so I have to custom the inference. The solution I’ve found is to clone the repository of the model, unzip it and then add the code/ file in which define the predict_fn function. My question is if there is any way to define the custom inference without clone the repository?