Deploys a fastai model to a sagemaker endpoint using torchserve. This notebook can be run on a CPU based Sagemaker notebook instance.

This is based on other guides on the internet that use pytorch 1.0 and pytorch 1.4. This guide uses the newer deployment mechanism of torchserve which is only available in pytorch >= 1.6.

Feel free to use this as a template for deploying your own models. I suffered through a lot of issues getting this working so hopefully I can save you some of the pain.