New deployment requirements

You need 3 things to create a new deployment:

1. An artifact file. It will be copied to /home/mlstudio/models for later usage and reproducablity.

2. A valid requirements.txt file. It will be validated on submission to make sure everything there works.

3. A python class for running the model. A python class with your inference code.

class UserModel(object):
​
    def __init__(self):
        # Load your model here
        pass
​
    def predict(self, data):
        # Your predication code goes here
        pass

Where data is json you would sent when calling the generated endpoint.

Create a new deployment

Once you have the above files, open ML Studio and go to Models under Serving.

kubectl get svc istio-ingressgateway -n istio-system

To add a new deployment, press New Deployment where you can add your model name, version and short description.

Then you will be prompted to select the files from the start of this page. And choose your prefered deployment infrastructure.

Finally, just wait a few minutes untill it is ready.

{
    "predication": response
}

{
    "predication": response
}

​