Model Configuration and Deployment

This lesson details the process of foundation model creation and deployment utilizing the SAP AI Launchpad. It should be noted that the subsequent steps are provided for informational purposes only, as workshop users lack the requisite authorization for independent model deployment. However, replication of these steps within a user’s own Business Technology Platform (BTP) account will facilitate the creation and deployment of user-specific models.

---

Note: The following steps are designated as Read-Only and are intended solely for individual reference! Participants will not have all the menu options available.

1. From the menu pane, expand ML Operations and select Configurations.

2. All current foundation model configurations are displayed. Selecting the Create button will create a new configuration.

3. The following fields should be populated with relevant details, and then select Next to continue.

• Configuration Name
• Scenario
• Version
• Executable

4. For the Input Parameters section, the name and version of the required model can be filled in. In this case, the default values are used.

Note: The model version can be given as the latest. If the model version is not specified, it will be set to the latest by default.

5. On the next screen - Input Artifacts - select Review to verify the details of the configuration.

6. On the Review screen, verify all the details of the configuration and select Create to create the configuration.

7. The Configuration Details page now appears, displaying attributes such as name, configuration ID and model version.

Creating a Deployment

The LLM can be made available for use by creating a virtual LLM deployment. This can be done once for each model and model version.

1. After creating the configuration in the previous section, click on Create Deployment to start the process.

2. Set the duration or a timeframe for this deployment to be active and then select Review.

3. Check the status of the deployment. Once it is running, the model can be accessed in the Generative AI Hub.

For the subsequent lessons, the prompts in AI Launchpad will be executed against a pre-deployed ChatGPT 3.5 Turbo LLM.