Model Serve > Introducing Informatica Model Serve > Deployment process
  

Deployment process

Deploying a model in Model Serve involves tasks that you perform in multiple applications. The process differs based on the type of model that you use.
The following image shows the process to deploy a model:
Using a machine learning framework, build and train your machine learning algorithm. Then, in Model Serve, register the model and deploy it as a URL endpoint. In your own application, send requests to the endpoint to generate predictions. Finally, monitor and stop the deployment in Model Serve.
The following table describes the tasks in the deployment process:
Task
Required for
Description
Build and train a machine learning algorithm.
User-defined models
Use a third-party machine learning framework, such as TensorFlow, to build and train an algorithm.
Register a machine learning model and a model deployment.
User-defined models
In Model Serve, configure a machine learning model where you upload the algorithm files. Then create a model deployment to define the runtime properties for deploying the model.
Deploy the model as an endpoint URL.
Quick start models and user-defined models
Start the quick start model or model deployment to make the endpoint URL available for requests.
Send requests to the endpoint to generate predictions.
Quick start models and user-defined models
In your application, send API calls to the endpoint URL to request predictions from the deployed model.
Monitor and stop the deployment.
Quick start models and user-defined models
Use Model Serve to monitor the status of your deployments, download logs for troubleshooting, and stop deployments to release the cloud resources.