Deploy a model to H2O MLOps through the H2O Hydrogen Torch UI
Overview
After building an H2O Hydrogen Torch model, you can upload it to H2O MLOps from the H2O Hydrogen Torch UI. Deployments are created and managed in the H2O MLOps UI, where all available scoring runtimes (including custom scoring runtimes) are accessible, along with security, monitoring, and Kubernetes resources.
You can also deploy a built model to H2O MLOps by using the model's H2O MLOps pipeline. To learn more, see Download an experiment's H2O MLOps pipeline.
Prerequisites
You upload a model from the MLOps tab of the experiment. The tab appears only when the following requirements are met:
- The experiment is a training experiment. Predictions do not have an MLOps tab.
- H2O MLOps is turned on for your H2O Hydrogen Torch instance.
- The experiment has a Finished or Uploaded status.
- The experiment is not a demo experiment.
You also need an H2O MLOps workspace to upload the model to. H2O Hydrogen Torch uploads to existing workspaces only, so if no workspace is available to you, create one in the H2O MLOps UI first.
If the experiment was built with the Native MLOps output training setting turned on, the connected H2O MLOps must be version 1.2.0 or later.
Instructions
Upload a model to H2O MLOps
To upload a built model (experiment) to H2O MLOps from the H2O Hydrogen Torch UI:
- Open the experiments list
In the H2O Hydrogen Torch navigation menu, click View experiments.
- Open the experiment
Click the name of the experiment (model) you want to upload to H2O MLOps.
- Go to the MLOps tab
Click the MLOps tab.
- Select an H2O MLOps workspace
In the Workspace list, select an H2O MLOps workspace. Each entry shows the workspace name followed by its UID.
note- To upload the model to a new workspace, create the workspace in the H2O MLOps UI first.
- H2O Hydrogen Torch preselects the workspace of your most recent upload. If you have not uploaded a model before, it preselects your personal H2O MLOps workspace, or the first workspace in the list. H2O MLOps lists the workspaces you can see, which are not necessarily the workspaces you can upload to. If H2O MLOps rejects an upload, the error message states the reason.
- Upload the model
Click Upload to MLOps.
After the upload completes, H2O Hydrogen Torch displays a success message and records the upload in the Uploaded models table. When H2O Hydrogen Torch can resolve the address of the H2O MLOps UI, the message also includes an Open in H2O MLOps button that opens the uploaded model in H2O MLOps.
H2O Hydrogen Torch registers the model in H2O MLOps under the experiment's name. If the workspace already contains a registered model with that name, H2O Hydrogen Torch adds the upload as a new version of that model.
If the experiment was built with Native MLOps output turned on and the connected H2O MLOps is earlier than version 1.2.0, H2O Hydrogen Torch stops the upload and reports the reason. If H2O Hydrogen Torch cannot detect the version, the upload proceeds, but the model can fail to score in H2O MLOps. In either case, upgrade H2O MLOps or run the experiment again with Native MLOps output turned off.
Deploy the uploaded model in H2O MLOps
Deployments are created and configured in the H2O MLOps UI. There, you can select any available scoring runtime, turn on security and monitoring, enter a deployment name and description, and set Kubernetes resource requests and limits.
To learn how to create a deployment in H2O MLOps, see Create a deployment in the H2O MLOps documentation. To learn about the states that a deployment can be in, see States in the H2O MLOps documentation.
- Submit and view feedback for this page
- Send feedback about H2O Hydrogen Torch to cloud-feedback@h2o.ai