TorchServe is an open source library from PyTorch and AWS intended to streamline the deployment of PyTorch models to production environments such as Amazon SageMaker, EC2 and Lambda. You can also quickly prototype serving PyTorch models on your local machine.

Check it out here.