Inspiration : There exists a lot of content around how to create various types of ML models using PyTorch but hardly any tutorials on how to deploy the same. This project tries to fill that gap by showing model deployment from the cloud for enterprise applications.
What it does : This turoail is a step by step guide in a simple GitHub repo that includes everything from setting up a cloud account for free, logging into a remote machine, installing tools and sample code to running TorchServe to deploy the Image Classification model and finally testing it using HTTP Post Reqest using CURL command.
How I built it : I build it by brining my knowledge of Oracle Cloud and referencing official documentation about from TorchServe's GitHub repository.
Challenges I ran into : I ran into several challenges with installing dependencies specially Anaconda on a Linux (Debian) environment, inhaling TorchServe but finally with persistence managed to overcome them.
Accomplishments that I'm proud of : I'm proud of adding this tutorial to help engineers take pre-trained model and deploy from Oracle Cloud.
What I learned :
What's next for How to serve PyTorch Models from the Cloud using TorchServe : Next I plan to write a client app using ReactJS that will call the ML Model for inference. Currently this version runs from the command line only.