NVIDIA GPU Cloud (NGC) offers pre-built containers for machine learning and deep learning workloads on NVIDIA GPUs. They have both general purpose and domain-specific offerings.
Domino enhances NGC containers for use in Domino and for general data science work.
- Add Domino compatibility
-
Domino automatically manages code and data versioning as part of the container lifecycle and as a result we require specific additional software in the container to work
- Add data source drivers
-
Domino adds drivers for data sources such as Snowflake, Oracle and Microsoft SQL to make it easy to connect to a data source without needing to manually add drivers.
- Add workspaces
-
When an NGC container doesn’t include an interactive notebook, Domino adds Jupyter to the image so that you can interactively develop in the image. When there’s a notebook already included, Domino configures the existing notebook to work.
Domino builds the latest versions of NGC containers with a preference for Ubuntu base images. If there’s an environment that you’re looking for which is not available, request it from ngc-request@dominodatalab.com.
-
Choose the NGC container from the list of available containers.
-
Create a new compute environment in Domino.
-
Use the link from the container you selected as the base compute environment (for example,
quay.io/domino/ngc-pytorch:20.12-py3) -
Enter the following in the Workspace Definition area:
jupyter:
title: "Jupyter (Python, R, Julia)"
iconUrl: "/assets/images/workspace-logos/Jupyter.svg"
start: [ "/opt/domino/workspaces/jupyter/start" ]
httpProxy:
port: 8888
rewrite: false
internalPath: "/{{ownerUsername}}/{{projectName}}/{{sessionPathComponent}}/{{runId}}/{{#if pathToOpen}}tree/{{pathToOpen}}{{/if}}"
requireSubdomain: false
jupyterlab:
title: "JupyterLab"
iconUrl: "/assets/images/workspace-logos/jupyterlab.svg"
start: [ "/opt/domino/workspaces/jupyterlab/start" ]
httpProxy:
internalPath: "/{{ownerUsername}}/{{projectName}}/{{sessionPathComponent}}/{{runId}}/{{#if pathToOpen}}tree/{{pathToOpen}}{{/if}}"
port: 8888
rewrite: false
requireSubdomain: falseRAPIDS
Domino registry path:
quay.io/domino/ngc-rapids:0.18-cuda11.0-runtime-ubuntu20.04-py3.8
Versions:
RAPIDS 0.18, Ubuntu 18.04, CUDA 11
Base Image:
docker pull nvcr.io/nvidia/rapidsai/rapidsai:0.18-cuda11.0-base-ubuntu18.04
Notes:
N/A