Skip to content

GPU support#

Hopsworks can harness the power of GPUs to speed up machine learning processes. You can take advantage of this feature in Hopsworks.ai by adding GPU equipped workers to your cluster. This can be done in two way: creating a cluster with GPU equipped workers or adding GPU equipped workers to an existing cluster.

Creating a cluster with GPU equipped workers#

When selecting the workers' instance type during the cluster creation, you can select an instance type equipped with GPUs. The cluster will then be created and Hopsworks will automatically detect the GPU resource.

Create cluster with GPUs
Create cluster with GPUs

Adding GPU equipped workers to an existing cluster.#

When adding workers to a cluster, you can select an instance type equipped with GPUs. The workers will then be added to the cluster and Hopsworks will automatically detect the new GPU resource.

Add GPUs to cluster
Add GPUs to cluster

Using the GPUs#

Once workers with GPUs have been added to your cluster you can use them by allocating GPUs to JupyterLab or Jobs.

Using GPUs in JupyterLab
Using GPUs in JupyterLab

Using GPUs in jobs
Using GPUs in jobs

For more information about using GPUs in Hopsworks you can consult Hopsworks Experiments documentation.