Model deployment using `run.deploy()` or `verta.deployment.DeployedModel` not working as expected

[Aid91]

Hi,

Is open-source version of the ModelDB supporting local model deployment (from logged models in ExperimentRuns), or this is only working on the SaaS and Enterprise solutions?

I am using open-source versions:

• ModelDB: 2.0.7
• ModelDB client: 0.14.6

In my use-case, logging everything in ModelDB works, only model deployment using run.deploy(wait=False) throws the following exception: _HTTPError: 404 Client Error: Not Found for url: http://HOST/api/v1/deployment/models/9ba21a96-9ea5-41c1-895c-9a26e623fbb2

Thanks!

[convoliution]

Hi @Aid91,

Thanks for the interest in ModelDB! Indeed, model deployment is currently only available in our enterprise offering, and not in our open-core platform.

We are currently evaluating opening part of our deployment solution, however. I’d love to schedule some time to hear more about your use case.

[Aid91]

@convoliution thanks for the quick answer!

I was trying out the possibilities of the ModelDB in our on-premise k8s cluster, and was interested if simple deployments are working, to quickly try out the REST API calls for instance. Something similar like you can find in MLFlow serving.

It was also documented here that the Model Deployment is "Limited" in the Open Source version, so was wondering what that means. Maybe it makes sense to write "Not supported" if that is the case.