Gaudi accelerators from Habana Labs (an Intel company) deliver low-cost-to-train deep learning models for natural language processing, object detection, and image recognition use cases.
AWS currently offers a dl1.24xlarge EC2 image with eight Gaudi accelerators available. See Amazon EC2 DL1 Instances.
-
Create or add a Gaudi-enabled node to an existing node-pool in your Domino cluster.
-
Create a hardware tier so users can use this resource in Domino.
-
Select the Use custom GPU resource name checkbox.
-
In GPU Resource Name enter
habana.ai/gaudi
.
-
Prerequisite: Admin kubectl
permissions to your cluster
-
Use the kubectl command to add the Habana device plugin for Kubernetes.
kubectl create -f https://vault.habana.ai/artifactory/docker-k8s-device-plugin/habana-k8s-device-plugin.yaml
-
Run the following command to verify the plugin is running:
kubectl get pods -n habana-system
Many Intel Habana environment containers work natively in Domino. To use a custom image to create a new environment, paste the Docker registry path into the FROM field.
vault.habana.ai/gaudi-docker/1.5.0/ubuntu20.04/habanalabs/tensorflow-installer-tf-cpu-2.9.1
See the Habana Developer Catalog for more pre-built container images that can be used in Domino.