Run SDK on headless server without GPU? #82
Description
Feature request / 功能建议
I want to run the SDK locally. I'm not entirely sure what it all does but i do know that installing it drags in pytorch, torchvision, etc... I already have the model itself running elsewhere so i assume running the SDK is just a fancy way of communicating with that model.
Motivation / 动机
It's fairly simple really. A beefy pc runs the model.
Now i want to call that model from a lightweight headless server (no gpu at all).
Surely the SDK can run on any machine, right?