[Ray Serve] using GRPC and DAG to host multiple models(or actors) in the same deployment

Hello Ray team, I was looking at the GRPC ingress (here) and I understand that it is still in alpha and I was just wondering if it is possible to use that with deployment graph (like this)?
If so, I am assuming we would be able to autoscale the different models independently?
If not, is this something that is in the roadmap for the future?

Hi @nihal , there shouldn’t have blockers to be used as deployment graph with current API, you can still follow the same idea as the doc suggested. Do you give a try? Please let me know if you have issues about it. :slight_smile:

Hi @Sihan_Wang
Yes, I had a misunderstanding regarding the deployment class in case of GRPC ingress. So now I was able to get it working, but I am not able to figure out what to give in import_path in the serve config file to deploy in a ray cluster using KubeRay. Could you please help?

Hi @nihal , glad to hear it worked from your side!

For import path, that is the package path for your deployment. Basically you should be able to get the import path by using serve build cli. Serve Config Files (serve build) — Ray 2.2.0