.. _enable_intel_gpu:

Enable Intel GPUs
=================

This guide describes how to configure Data Science Stack (DSS) to utilise the Intel GPUs on your machine. 

You can do so by enabling the Intel device plugin on your `Canonical K8s`_ cluster.

Prerequisites
-------------

* Ubuntu 24.04.
* DSS is :ref:`installed <install_DSS_CLI>` and :ref:`initialised <initialise_DSS>`.
* The `kubectl snap <https://snapcraft.io/kubectl>`_ package is installed.
* Your machine includes an Intel GPU.

Verify the Intel GPU drivers
----------------------------

To confirm that your machine has the Intel GPU drivers set up, first install the `intel-gpu-tools` package:

.. code-block:: bash

   sudo apt install intel-gpu-tools

Now list the Intel GPU devices on your machine as follows:

.. code-block:: bash

   intel_gpu_top -L

If the drivers are correctly installed, you should see information about your GPU device such as the following:

.. code-block::

   card0                    8086:56a0
   pci:vendor=8086,device=56A0,card=0
   └─renderD128 

.. note::
   For Intel discrete GPUs on Ubuntu versions older than 24.04, you may need to perform additional steps such as installing a `HWE kernel <https://ubuntu.com/kernel/lifecycle>`_. 

Enable the Intel GPU plugin
---------------------------

To ensure DSS can utilise Intel GPUs, you have to enable the Intel GPU plugin in your Canonical K8s cluster.

1. Use `kubectl kustomize` to build the plugin YAML configuration files:

.. code-block:: bash

  VERSION=v0.30.0
  kubectl kustomize https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/nfd?ref=${VERSION} > node_feature_discovery.yaml
  kubectl kustomize https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/nfd/overlays/node-feature-rules?ref=${VERSION} > node_feature_rules.yaml
  kubectl kustomize https://github.com/intel/intel-device-plugins-for-kubernetes/deployments/gpu_plugin/overlays/nfd_labeled_nodes?ref=${VERSION} > gpu_plugin.yaml

To allow multiple containers to utilise the same GPU, run:

.. code-block:: bash
                
  sed -i 's/enable-monitoring/enable-monitoring\n        - -shared-dev-num=10/' gpu_plugin.yaml

2. Apply the built YAML files to your Canonical K8s cluster:

.. code-block:: bash
                
  kubectl apply -f node_feature_discovery.yaml
  kubectl apply -f node_feature_rules.yaml
  kubectl apply -f gpu_plugin.yaml

The Canonical K8s cluster is now configured to recognise and utilise your Intel GPU.

.. note::
 After the YAML configuration files have been applied, they can be safely deleted.

Verify the Intel GPU plugin
---------------------------

To verify the Intel GPU plugin is installed and the Canonical K8s cluster recognises your GPU, run:

.. code-block:: bash

   kubectl get nodes --show-labels | grep intel

You should see an output with the cluster name such as the following:

.. code-block:: bash

   kubectl get nodes --show-labels | grep intel
   fluent-greenshank   Ready    <none>   18s   v1.30.3   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,intel.feature.node.kubernetes.io/gpu=true

Verify DSS detects the GPU
--------------------------

Verify DSS has detected the GPU by checking the DSS status. To do so, run the following command using the DSS CLI: 

.. code-block:: bash

  dss status

You should expect an output like this:

.. code-block:: bash
                
  Output:
  MLflow deployment: Ready
  MLflow URL: http://10.152.183.68:5000
  NVIDIA GPU acceleration: Disabled
  Intel GPU acceleration: Enabled

See also
--------

* To learn how to manage your DSS environment, check :ref:`manage_DSS`.
* If you are interested in managing Jupyter Notebooks within your DSS environment, see :ref:`manage_notebooks`.