Serverless LLMs

I've kinda multiple questions to ask

1. Can I deploy a model serverlessly on Sagemaker? My current requirements are some pretrained models with around 8-12GB of weights and stuff.
2. What is the best approach for LLM deployment and functioning on sagemaker? if I wanna plug some models from github and choose to manage them here on sagemaker. I might need fine-tuning thing along the way.