Is it possible to take a ML Azure Container Instance from the registry and deploy it elsewhere?

Question

Hi,

I was wondering if it was possible to take a ML Model deployed to a container instance and then take that and deploy that in docker elsewhere. I have managed to sync a repo with the Azure registry and download the container which was created in Azure ML. When i started the container though it didn't do anything; i logged into the container and nothing was running; i could start NGINX but wasn't sure if this was a supported?

I was that you can do this with the Azure Sentiment Analysis Containers but wondering if you could do with ML Azure Container Instances?

Any pointers would be good.

Answer 1

Thanks for reaching out. Yes, however, you need to run az ml model package to get an image with your model baked into it, otherwise, the model file is mounted by the deployment engine. Here's more information on how to package a registered model with Docker. Hope this helps.