Override KServe configuration

You can override KServe settings to customize your model-serving environment.

Follow the general steps for your operating system.

Option 1

  1. Make a copy of the existing inferenceservice-config config map file in the redhat-ods-applications namespace.

  2. Edit the settings you want to change.

  3. Overwrite the existing ConfigMap object.

  4. Restart KServe by either deleting the pod or scaling the Deployment pod parameter down to 0 and then back up to 1.

Option 2

  1. Copy the /usr/lib/microshift/manifests.d/010-microshift-ai-model-serving-kserve/inferenceservice-config-microshift-patch.yaml config map file.

  2. Edit the settings you want to change.

  3. Apply the ConfigMap object.

  4. Restart KServe by either deleting the pod or scaling the Deployment pod parameter down to 0 and then back up to 1.

For RHEL for Edge and image mode for RHEL systems

  1. Create a new manifest with the ConfigMap file, based on either the /usr/lib/microshift/manifests.d/010-microshift-ai-model-serving-kserve/inferenceservice-config-microshift-patch.yaml or inferenceservice-config file, in the redhat-ods-applications namespace.

  2. Place the new manifest in the /usr/lib/microshift/manifests.d/ directory. Staring with prefix 011 is recommended so that your manifest is applied after the /usr/lib/microshift/manifests.d/010-microshift-ai-model-serving-kserve/ directory contents.

← Previous: Export model-server metrics by using Open Telemetry | ↑ Up: AI | ⌂ Home: Red Hat build of MicroShift | Next: Develop →