Newbie question trying to use torchserve to serve a Satrn model

[jaydeboer]

Hello,

I am new to the whole MMLabs project, and to machine learning, so I am guessing I am just missing some fundamental things, so please bear with me. I wanted to host a model in torchserve to expose an inference api. With the update to 1.x, they deployment tools have been removed, and from what I understand, we should use MMDeploy to make create the correct type of model for torchserve.

I was wanting to try out the Satrn model, using just the default to start to figure out the process. Eventually I want to try some custom kie stuff, but I wanted to start simple.

I seem to get lost on how to create .mar file for torchserve. Any clues would be very much appreciated. I have spent a day hacking at all kinds of things and have managed to get nowhere. If there is a portion of documentation that I have missed, I would love to be pointed in the right direction.

Thanks so much,

Jay