NGC on Google Cloud Platform Virtual Machines

This NGC on Google Cloud Platform Cloud Guide explains how to set up an NVIDIA GPU Cloud Image on the Google Cloud Platform and also provides release notes for each version of the NVIDIA virtual machine image.

1. Using NGC with Google Cloud Platform

NVIDIA makes available on Google Cloud Platform (GCP) three different virtual machine images (VMIs). These are GPU-optimized VMIs for GCP VM instances with NVIDIA A100, V100 or T4 GPUs.

  • NVIDIA GPU-Optimized Image for Deep Learning, Machine Learning & HPC

    The base GPU-Optimized image Includes Ubuntu Server, the NVIDIA driver, Docker CE, and the NVIDIA Container Runtime for Docker

  • NVIDIA GPU-Optimized Image for TensorFlow

    The base image with NVIDIA’s GPU-Accelerated TensorFlow container pre-installed

  • NVIDIA GPU-Optimized Image for PyTorch

    The base image with NVIDIA’s GPU-Accelerated PyTorch container pre-installed

  • NVIDIA HPC SDK Image

    The base image with NVIDIA’s HPC SDK pre-installed

For those familiar with GCP, the process of launching the instance is as simple as logging into GCP and creating a deployment solution using the Google Cloud Launcher. After deploying the NVIDIA GPU-Optimized Image of choice,you can SSH into your VM and start building a host of AI applications in deep learning, machine learning and data science by leveraging the wide range of GPU-accelerated containers, pre-trained models and resources available from the NGC Catalog.

This document provides step-by-step instructions for accomplishing this, including how to use the gcloud CLI.

1.1. Prerequisites

  • You have a Google Cloud account - https://console.cloud.google.com/ .

  • Browse the NGC website and identify an available NGC container and to run on the Virtual Machine Instance (VMI).
  • You have installed the gcloud SDK if you plan to use the CLI. See setup instructions below.
  • Windows Users: The CLI code snippets are for bash on Linux or Mac OS X. If you are using Windows and want to use the snippets as-is, you can use the Windows Subsystem for Linux and use the bash shell (you will be in Ubuntu Linux).

1.2. Before You Get Started

Be sure you are familiar with the information in this chapter before starting to use the NVIDIA GPU Cloud Image on the Google Cloud Platform (GCP).

1.2.1. Set Up Your SSH Key

The Google Compute Engine generates and manages an SSH key automatically for logging into your instance (see the Google Cloud documentation Connecting to Instances.). However, to facilitate logging into the NGC container registry upon the initial connection to the VM instance, you need to -

  1. Generate your own SSH keys (see Creating a new SSH key for instructions), and then
  2. Add them to the metadata for your project (see Adding or Removing Project-Wide Public SSH Keys for instructions).
If you do not prepare your SSH keys before launching and connecting to your VM instance, you will not be able to access the NGC initially. In that case you will need to
  1. Add yourself to the docker group after connecting to the instance.
    sudo usermod -aG docker $USER
  2. Restart the session.

1.2.2. Set Up Firewall Rules

NVIDIA recommends setting firewall rules to allow external access to ports 443 (HTTPS), 8888 (DIGITS), and any other ports that may be needed. This should be done before launching an instance to avoid having to stop the instance when setting any firewall rules later.

  1. Log in to https://console.cloud.google.com
  2. Verify you are in the correct Project.
  3. Click the Products and Services menu icon, then scroll down to the Networking section and click VPC Network->Firewall Rules.

  4. Click Create Firewall Rule.
  5. Enter the following information to specify the firewall rule you want to create.
    • Name: NVIDIA recommends the following naming format

      For HTTPS: “default-allow-https”

      For DIGITS: “default-allow-digits”

      You can also create rules for other DIGITS versions, such as DIGITS4

    • Direction of traffic: "Ingress"
    • Action on match: "Allow"
    • Targets: "All instances in the network"
    • Source filter: "IP ranges"
    • Source IP ranges: "0.0.0.0/0"
    • Protocols and ports: "Specified protocols and ports", then enter

      For HTTPS: “tcp:443”

      For DIGITS: “tcp:8888”

      You can enter ports for other DIGITS versions as well

  6. Click Create.

    Your new firewall rules should appear on the Firewall Rules page.

1.3. Creating an NGC Certified Virtual Machine Using the GCP Console

1.3.1. Log In and Create VM

  1. Log in to https://console.cloud.google.com.
  2. Verify you are in the correct project.
  3. Open the Google Cloud Platform Marketplace page.

    Click the Products and Services menu icon and select Marketplace.

  4. Search for "nvidia".
  5. Select your choice of the three flavors of NVIDIA GPU-optimized images published by NVIDIA.
  6. From the image information page, click Launch.

  7. Configure the NVIDIA GPU Cloud Image deployment.
    1. In "Name", enter your new deployment name.
    2. In "Zone", select the zone to create the instance (select one that features the appropriate GPU).
    3. In the "Machine Type" section, click Customize to open the customize view.
    4. Under the GPU section, select the GPU type and Number of GPUs.

      Assign the Cores (vCPUs) and Memory. The following ratio is recommended: 1x GPU : 10x vCPU: 60 GB mem

    5. In the "Boot disk" section, select Standard Persistent Disk.
    6. Make other changes as needed for Networking, Firewall and IP.

  8. Click Deploy from the bottom of the page. It may take a few minutes for the deployment process to complete.

1.3.2. Connect to Your VM Instance

  • If you are still on the Deployment page, you can click SSH to connect to your instance.
  • If you are no longer on the Deployment page, you can return to your instance and connect as follows.
  1. Click the Products and Services menu icon, then scroll down to the Compute Engine section and click VM Instances.
  2. Either click SSH by your listed deployed instance, or click your deployed instance and then click SSH from the VM instance details page.

Start/Stop Your VM Instance

Select your GPU instance, either from the Deployment Manager->your deployment page or from the Compute Engine->VM Instances page.

The top menu lets you edit, stop a running instance, or start a stopped instance.

Delete VM and Assocated Resources

Select your GPU instance, either from the Deployment Manager->your deployment page or from the Compute Engine->VM Instances page.and then click Delete.

1.4. Create an NGC Certified Virtual Machine Using the gcloud CLI

This section explains how to create a GPU Cloud instance using the gcloud CLI.

This flow and the code snippets in this section are for Linux or Mac OS X. If you are using Windows, you can use the Windows Subsystem for Linux and use the bash shell (where you will be in Ubuntu Linux). Many of these CLI commands can have significant delays.

For more information about creating a deployment using gcloud CLI, see Creating a Deployment using gcloud or the API.

1.4.1. Install and Set Up gcloud CLI

Follow the instructions at https://cloud.google.com/sdk/docs/quickstarts. These include instructions for Linux, Mac, and Windows.

The instructions walk you through the platform specific install and initial gcloud login.

For at least the Mac, you will be given a large list of additional gcloud components to install such as extensions for GO, Python and Java. You can use the defaults for now, and use the gcloud components command later to list, install, or remove them.

Once the setup is complete, start a new shell since your environment has been updated.

1.4.2. Set Up Instance Options

You will need to specify the following options when creating the custom GPU instance.

OPTION [1] VALUE NOTES
<instance-name> Name of your choosing. Ex. “my-ngc-instance” Must be all lowercase, with no spaces. Hyphens and numbers are allowed.
--project "<my-project-id>"

This is the project in which the VM will be created.

Use

gcloud projects list

to view PROJECT ID to use for this field.

--zone

One of the following zones that contain GPUs:

"us-west1-b"

"us-east1-c"

"us-east1-d"

"europe-west1-b"

"europe-west1-d"

"asia-east1-a"

"asia-east1-b"

Pick one nearest you and with the GPUs you want to use.
--machine-type

One of the following:

"custom-10-61440" (for 1x P100 or V100)

"custom-20-122880" (for 2x P100)

"custom-40-212992" (for 4x P100)

"custom-80-491520" (for 8x V100)

vCPU/Memory configuration of the VM in "custom-<#vCPUs>-<memory MB>" format.

Recommended ratio is 1 GPU : 10 vCPUs : 60 GB memory

--subnet “default”, or the name of the VPC network to use
--metadata "ssh-keys=<user-id>:ssh-rsa <ssh-key> <user-email>"
--maintenance-policy

"TERMINATE"

What to do with your instance when Google performs maintenance on the host
--service-account

Compute Engine identity attached to the instance.

Use

gcloud iam service-accounts list 
to view the email for your account.
--scope

"https://www.googleapis.com/auth/devstorage.read_only",

"https://www.googleapis.com/auth/logging.write",

"https://www.googleapis.com/auth/monitoring.write",

"https://www.googleapis.com/auth/servicecontrol",

"https://www.googleapis.com/auth/service.management.readonly",

"https://www.googleapis.com/auth/trace.append"

Default values (recommended). Specifies the permissions for your instance.
--accelerator

nvidia-tesla-p100,count=[1,2,4]

Which GPU to attach, and how many
--min-cpu-platform

"Intel Broadwell"

(for P100 instances)

--image Name of the latest NVIDIA GPU Cloud Image (See the NGC GCP VMI Release Notes for the current name.)
--image-project "nvidia-ngc-public" Project name in which the NVIDIA GPU Cloud Image is located
--boot-disk-size 32
--boot-disk-type "pd-standard"
--boot-disk-device-name Name of your choosing Recommend using the same name as your VM instance for easy correlation

1.4.3. Launch Your VM Instance

Use the Python scripts provided at https://github.com/nvidia/ngc-examples/tree/master/ncsp to create your custom GPU instance. You can also enter the following, using the information gathered in the previous section:

gcloud compute \
--project "<project-id>" \
instances create "<instance-name>" \
--zone "<zone>" \
--machine-type "<vCPU-mem-config>" \
--subnet "<subnet-name>" \
--metadata "<your-public-ssh-key>" \
--maintenance-policy "<maintenance-policy>" \
--service-account "<service-account-email>" \
--scopes "https://www.googleapis.com/auth/devstorage.read_only","https://www.googleapis.com/auth/logging.write","https://www.googleapis.com/auth/monitoring.write","https://www.googleapis.com/auth/servicecontrol","https://www.googleapis.com/auth/service.management.readonly","https://www.googleapis.com/auth/trace.append" \
--accelerator type=<accelerator-type> \
--min-cpu-platform "<CPU-platform>" \
--image "<nvidia-gpu-cloud-image>" \
--image-project "<project-name>" \
--boot-disk-size "32" \
--boot-disk-type "pd-standard" \
--boot-disk-device-name "<boot-disk-name>"

1.4.4. Connect to Your VM Instance

(Use a CLI on Mac or Linux. Windows users: use OpenSSH on Windows PowerShell or use the Windows Subsystem for Linux)

If you ran the scripts from https://github.com/nvidia/ngc-examples/tree/master/ncsp you should be connected to your instance. Otherwise, run ssh to connect to your GPU instance, or enter the following gcloud command.

Command syntax:

gcloud compute --project "<project-id>" ssh --zone "<zone>" "<instance-name>"

See https://cloud.google.com/compute/docs/instances/connecting-to-instance for more information about connecting to your GPU instance.

Stop/Stop Your VM Instance

Once an instance is running, you can stop and (re)start your instance.

Stop:

gcloud compute instances stop <instance-name>

Start or Restart:

gcloud compute instances start <instance-name> <zone>

Access Jupyter Notebooks in Your GPU Virtual Machine

Accessing Jupyter notebooks you create or download from the NGC Catalog in your Google Cloud virtual machine is simple and straightforward.

Follow the below steps to configure your virtual machine instance with the right network settings to be able to run a JupyterLab server on your VM and access Jupyter notebooks via a browser on your local machine.

  1. Create a NGC certified virtual machine instance on Google Cloud by following the steps listed above. (LINK TO ‘CREATING AN NGC CERTIFIED VIRTUAL MACHINE USING THE GOOGLE CLOUD CONSOLE’ SECTION ABOVE) Refer to Creating an NGC Certified Virtual Machine Using the GCP Console
  2. Create a static external IP address.

    This static external IP address will be used as hostname when you access the Jupyter notebook from your local browser. (example: http://<external IP address>:8080)

    1. Navigate to Networking (from menu) → VPC network → External IP addresses.

    2. Identify the virtual machine instance you created and change the the type from “Ephemeral” to “Static

    3. Copy the external IP in the corresponding External Address column for use later.
  3. Navigate to Networking (from menu) → VPC network → Firewall rules to create a new firewall rule with the following parameters (or add a new rule to an existing VPC if you already created one).
    • Name: <Enter a firewall name>
    • Targets: All instances in the network
    • Source IP ranges: 0.0.0.0/0
    • Protocols and ports: Select “Specified protocols and ports” option. tcp: 8080 <You can change any other port number>

That’s it! Now you’re all set to create and edit Jupyter notebooks that are in your virtual machine instance.

Now you can pull any container from NGC and access Jupyter notebooks from within the container as well.

While running the container, make sure to include the port you configured for the JupyterLab while creating the VPC (in this example, the port used was 8080)

For example:

$ docker run --gpus '"device=1"' --rm -it \
    -p 8080:8080 \
    -p 6006:6006 \
    --shm-size=1g \
    --ulimit memlock=-1 \
    --ulimit stack=67108864 \
    --name bert_gcp \
$ docker pull nvcr.io/nvidia/tensorflow:20.08-tf1-py3

You can now access the Jupyter notebooks in your Google Cloud virtual machine by simply navigating to https://<externalip>:8080 on any browser on your local machine. (External IP to be included in the URL is the same as the external IP you made a note of in step 2c)

1.6. Persistent Data Storage for GCP Virtual Machines

GCP recommends using Persistent SSD Disks for Compute Engine storage. A minimum of 1 TB of storage is recommended for storing deep learning datasets. However, a much larger disk or a software RAID, using mdadm, can be used to create a volume with multiple SSD Persistent Disks for achieving the the maximum performance supported by GCP on a Compute Engine instance. See instructions on how to set up software RAID on local disks. Persistent SSD disks can also be set up for software RAID using the same instructions.

1.6.1. Create a Data Disk Using the GCP Console

You can create a persistent SSD dataset disk from the GCP console as follows.
  1. Log on to the Google Cloud Platform.
  2. Create the SSD disk.
    1. Click Compute Engine-> Disks in the left-side navigation pane.
    2. Click Create Disk from the top of the page.

    3. Specify the following and then click Create when done:
      • Zone: Select the same zone as the VM instance you created.
      • Disk Type: SSD persistent disk
      • Source type: None (blank disk)
      • Size: At least 1024 GB
      If you choose to provide your own encryption key, You must provide a key that is a 256-bit string encoded in RFC 4648 standard base64 to Compute Engine. See Customer-Supplied-Encryption-Keys for details on how to provide a custom Encryption Key globally for all your operations.
  3. Attach the disk to the VM instance.
    1. Go to the Compute Engine->VM Instance page.
    2. Click your VM instance from the list.
    3. Click Stop.

      You must stop a running VM Instance as changes cannot be performed when the instance is running.

    4. Click Edit.
    5. Scroll down to the Additional Disks and click + Add Item.
    6. Under Name, select the disk that you created and want to attach to the VM instance.
    7. Click Save.
    8. Start the VM instance.

1.6.2. Create a Data DIsk Using the gcloud CLI

  1. Create the disk using the following command.
    $ gcloud compute disks create ngc-ssd --zone <zone> --description “<your-description>” --type=pd-ssd --size=1000GM
  2. Attach the disk to a VM instance using the following command..
    $ gcloud compute instances attach-disk <instance-name> --disk ngc-ssd --zone <zone>

1.6.3. Delete a Data Disk

Be aware that once you delete a Persistent SSD Disk, you cannot undelete it.

gcloud CLI

$ gcloud compute instances detach-disk <instance-name> --disk ngc-ssd --zone <zone>

GCP Console

  1. Click the disk to delete from the Compute Engine->Disks page.
  2. On the top of the page, click Delete.
  3. Click Delete at the Delete a disk confirmation dialog. .

2. Release Notes for NVIDIA Virtual Machine Images on the Google Cloud Platform

NVIDIA makes available on the Google Cloud Platform a customized NVIDIA virtual machine image optimized for the NVIDIA® Volta™ GPU. Running NVIDIA GPU Cloud containers on this instance provides optimum performance for deep learning jobs.

See the NGC with Google Cloud Platform Setup Guide for instructions on setting up and using the VMI.

2.1. Version 20.03.1

Image Name

  • NGC Image: nvidia-gpu-cloud-image-20200324
  • NGC with TensorFlow Image: nvidia-gpu-cloud-image-tensorflow-20200324
  • NGC with PyTorch Image: nvidia-gpu-cloud-image-pytorch-20200324

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 18.04 LTS
  • NVIDIA Driver:  440.64.01
  • Docker Engine:   19.03.6
  • NVIDIA Container Toolkit v1.0.5-1

    Includes new command to run containers: docker run --gpus all <container>

  • (NGC TensorFlow Image only) nvcr.io/nvidia/tensorflow:20.02-py3
  • (NGC PyTorch Image only) nvcr.io/nvidia/pytorch:20.02-tf1-py3, nvcr.io/nvidia/pytorch:20.02-tf2-py3

Key Changes

  • Updated Docker Engine to 19.03.6
  • Updated NVIDIA Driver to 440.64.01

Known Issues

Installing GPU drivers on the VM via a CUDA Install Succeeds Erroneously

Issue

Attempting to install CUDA on the VM will succeed, resulting in a potential conflict with the NVIDIA GPU driver included in the VM image.

Explanation

The configuration file to prevent driver installs is not working. This will be resolved in a later release of the VM image.

2.3. Version 19.11.3

Image Name

  • NGC Image: nvidia-gpu-cloud-image-20191114
  • NGC with TensorFlow Image: nvidia-gpu-cloud-image-tensorflow-20191114
  • NGC with PyTorch Image: nvidia-gpu-cloud-image-pytorch-20191114

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 18.04 LTS
  • NVIDIA Driver:  440.33.01
  • Docker CE:   19.03.4-ce
  • NVIDIA Container Toolkit v1.0.5-1

    Includes new command to run containers: docker run --gpus all <container>

  • (NGC TensorFlow Image only) nvcr.io/nvidia/tensorflow:19.10-py3
  • (NGC PyTorch Image only) nvcr.io/nvidia/pytorch:19.10-py3

Key Changes

  • Updated Docker-CE to 19.03.4
  • Updated NVIDIA Driver to 440.33.01

Known Issues

Installing GPU drivers on the VM via a CUDA Install Succeeds Erroneously

Issue

Attempting to install CUDA on the VM will succeed, resulting in a potential conflict with the NVIDIA GPU driver included in the VM image.

Explanation

The configuration file to prevent driver installs is not working. This will be resolved in a later release of the VM image.

Version 19.10.2

Image Name

  • NGC Image: nvidia-gpu-cloud-image-20190809
  • NGC with TensorFlow Image: nvidia-gpu-cloud-image-tensorflow-20190809
  • NGC with PyTorch Image: nvidia-gpu-cloud-image-pytorch-20190809

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 18.04 LTS
  • NVIDIA Driver:  418.87
  • Docker CE:   19.03.2-ce
  • NVIDIA Container Toolkit v1.0.5-1
  • (NGC TensorFlow Image only) nvcr.io/nvidia/tensorflow:19.09-py3
  • (NGC PyTorch Image only) nvcr.io/nvidia/pytorch:19.09-py3

Key Changes

  • Updated Docker-CE to 19.03.2
  • Replaced the NVIDIA Container Runtime for Docker with the NVIDIA Container Toolkit.

    Includes new command to run containers: docker run --gpus all <container>

Known Issues

Installing GPU drivers on the VM via a CUDA Install Succeeds Erroneously

Issue

Attempting to install CUDA on the VM will succeed, resulting in a potential conflict with the NVIDIA GPU driver included in the VM image.

Explanation

The configuration file to prevent driver installs is not working. This will be resolved in a later release of the VM image.

2.4. Version 19.08.0

Image Name

  • NGC Image: nvidia-gpu-cloud-image-20190809
  • NGC with TensorFlow Image: nvidia-gpu-cloud-image-tensorflow-20190809
  • NGC with PyTorch Image: nvidia-gpu-cloud-image-pytorch-20190809

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 18.04 LTS
  • NVIDIA Driver:  418.87
  • Docker CE:   18.09.8-ce
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.1.0-1
  • (NGC TensorFlow Image only) nvcr.io/nvidia/tensorflow:19.06-py3
  • (NGC PyTorch Image only) nvcr.io/nvidia/pytorch:19.06-py3

Key Changes

  • Updated NVIDIA Driver to version 418.87.
  • Updated Docker-CE to 18.09.8
  • Updated NVIDIA Container Runtime for Docker to v2.1.0-1

Known Issues

Installing GPU drivers on the VM via a CUDA Install Succeeds Erroneously

Issue

Attempting to install CUDA on the VM will succeed, resulting in a potential conflict with the NVIDIA GPU driver included in the VM image.

Explanation

The configuration file to prevent driver installs is not working. This will be resolved in a later release of the VM image.

2.5. Version 19.07.0

Image Name

  • NGC Image: nvidia-gpu-cloud-image-20190628
  • NGC with TensorFlow Image: nvidia-gpu-cloud-image-tensorflow-20190628
  • NGC with PyTorch Image: nvidia-gpu-cloud-image-pytorch-20190628

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 18.04 LTS
  • NVIDIA Driver:  418.67
  • Docker CE:   18.09.7-ce
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.0.3
  • (NGC TensorFlow Image only) nvcr.io/nvidia/tensorflow:19.06-py3
  • (NGC PyTorch Image only) nvcr.io/nvidia/pytorch:19.06-py3

Key Changes

Known Issues

Installing GPU drivers on the VM via a CUDA Install Succeeds Erroneously

Issue

Attempting to install CUDA on the VM will succeed, resulting in a potential conflict with the NVIDIA GPU driver included in the VM image.

Explanation

The configuration file to prevent driver installs is not working. This will be resolved in a later release of the VM image.

2.6. Version 19.05.1

Image Name

  • NGC Image: nvidia-gpu-cloud-image-20190522
  • NGC with TensorFlow Image: nvidia-gpu-cloud-image-tensorflow-20190522
  • NGC with PyTorch Image: nvidia-gpu-cloud-image-pytorch-20190522

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 18.04 LTS
  • NVIDIA Driver:  418.67
  • Docker CE:   18.09.4-ce
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.0.3
  • (NGC TensorFlow Image only) nvcr.io/nvidia/tensorflow:19.04-py3
  • (NGC PyTorch Image only) nvcr.io/nvidia/pytorch:19.04-py3

Key Changes

19.05.1

19.05.0

  • Updated the Ubuntu Server to 18.04 LTS
  • Updated the NVIDIA Driver to 418.67
  • Updated Docker CE to 18.09.4-ce.

Known Issues

Installing GPU drivers on the VM via a CUDA Install Succeeds Erroneously

Issue

Attempting to install CUDA on the VM will succeed, resulting in a potential conflict with the NVIDIA GPU driver included in the VM image.

Explanation

The configuration file to prevent driver installs is not working. This will be resolved in a later release of the VM image.

2.7. Version 19.03.0

Image Name

  • NGC Image: nvidia-gpu-cloud-image-20190318
  • NGC with TensorFlow Image: nvidia-gpu-cloud-image-tensorflow-20190318
  • NGC with PyTorch Image: nvidia-gpu-cloud-image-pytorch-20190318

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 18.04 LTS
  • NVIDIA Driver:  418.40.04
  • Docker CE:   18.09.2
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.0.3
  • (NGC TensorFlow Image only) nvcr.io/nvidia/tensorflow:19.03-py3
  • (NGC PyTorch Image only) nvcr.io/nvidia/pytorch:19.03-py3

Key Changes

  • Updated the NVIDIA Driver to 418.40.04

Known Issues

There are no known issues in this release.

Version 19.02.0

Image Name

  • NGC Image: nvidia-gpu-cloud-image-20190219
  • NGC with TensorFlow Image: nvidia-gpu-cloud-image-tensorflow-20190219
  • NGC with PyTorch Image: nvidia-gpu-cloud-image-pytorch-20190219

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 18.04 LTS
  • NVIDIA Driver:  410.104
  • Docker CE:   18.09.2
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.0.3
  • (NGC TensorFlow Image only) nvcr.io/nvidia/tensorflow:19.04-py3
  • (NGC PyTorch Image only) nvcr.io/nvidia/pytorch:19.04-py3

Key Changes

  • Initial release of images incorporating PyTorch and TensorFlow
  • Updated the Ubuntu Server to 18.04 LTS
  • Updated the NVIDIA Driver to 410.104
  • Updated Docker CE to 18.09.2.

Known Issues

There are no known issues in this release.

Version 18.11.1

Image Name

  • NGC Image: nvidia-gpu-cloud-image-20181121

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 16.04 LTS
  • NVIDIA Driver:  410.79
  • Docker CE:   18.06.1
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.0.3

Key Changes

  • Updated the NVIDIA driver to Release 410, version 410.79.

Known Issues

There are no known issues in this release.

2.10. Version 18.09.1

Image Name

  • Basic NGC Image: nvidia-gpu-cloud-image-20180911

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 16.04 LTS
  • NVIDIA Driver:  410.48
  • Docker CE:   18.06.1

Key Changes

  • Updated the NVIDIA driver to Release 410, version 410.48.
  • Updated Docker CE to 18.06.1

Known Issues

There are no known issues in this release.

2.11. Version 18.08.1

Image Name

nvidia-gpu-cloud-image-20180816

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 16.04 LTS
  • NVIDIA Driver:  396.44
  • Docker CE:   18.06-ce
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.0.3

Key Changes

  • Updated the NVIDIA driver to 396.44.
  • Updated Docker CE to 18.06

Known Issues

There are no known issues in this release.

2.12. Version 18.07.0

Image Name

nvidia-gpu-cloud-image-20180717

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 16.04 LTS
  • NVIDIA Driver:  396.37
  • Docker CE:   18.03.1-ce
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.0.3

Key Changes

  • Updated the NVIDIA driver to 396.37.

Known Issues

There are no known issues in this release.

2.13. Version 18.06.0

Image Name

nvidia-gpu-cloud-image-20180613

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 16.04 LTS
  • NVIDIA Driver:  396.26
  • Docker CE:   18.03.1-ce
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.0.3

Key Changes

  • Updated the NVIDIA driver to 396.26.

Known Issues

There are no known issues in this release.

2.14. Version 18.05.0

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 16.04 LTS
  • NVIDIA Driver:  384.125
  • Docker CE:   18.03.1-ce
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.0.3

Key Changes

  • Includes Ubuntu 16.04 security updates
  • Updated Docker CE to version 18.03.1-ce

Known Issues

There are no known issues in this release.

2.15. Version 18.04.0

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 16.04 LTS
  • NVIDIA Driver:  384.125
  • Docker CE:   18.03.0-ce
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.0.3

Key Changes

  • Updated the NVIDIA Driver to version 384.125
  • Updated Docker CE to version 18.03.0-ce

Known Issues

There are no known issues in this release.

2.16. Version 18.03.0

Contents of the NVIDIA GPU Cloud Image

  • Ubuntu Server: 16.04 LTS
  • NVIDIA Driver:  384.111
  • Docker CE:   17.12.1-ce
  • NVIDIA Container Runtime for Docker: (nvidia-docker2) v2.0.3

Known Issues

There are no known issues in this release.

Notices

Notice

THE INFORMATION IN THIS GUIDE AND ALL OTHER INFORMATION CONTAINED IN NVIDIA DOCUMENTATION REFERENCED IN THIS GUIDE IS PROVIDED “AS IS.” NVIDIA MAKES NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE WITH RESPECT TO THE INFORMATION FOR THE PRODUCT, AND EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE. Notwithstanding any damages that customer might incur for any reason whatsoever, NVIDIA’s aggregate and cumulative liability towards customer for the product described in this guide shall be limited in accordance with the NVIDIA terms and conditions of sale for the product.

THE NVIDIA PRODUCT DESCRIBED IN THIS GUIDE IS NOT FAULT TOLERANT AND IS NOT DESIGNED, MANUFACTURED OR INTENDED FOR USE IN CONNECTION WITH THE DESIGN, CONSTRUCTION, MAINTENANCE, AND/OR OPERATION OF ANY SYSTEM WHERE THE USE OR A FAILURE OF SUCH SYSTEM COULD RESULT IN A SITUATION THAT THREATENS THE SAFETY OF HUMAN LIFE OR SEVERE PHYSICAL HARM OR PROPERTY DAMAGE (INCLUDING, FOR EXAMPLE, USE IN CONNECTION WITH ANY NUCLEAR, AVIONICS, LIFE SUPPORT OR OTHER LIFE CRITICAL APPLICATION). NVIDIA EXPRESSLY DISCLAIMS ANY EXPRESS OR IMPLIED WARRANTY OF FITNESS FOR SUCH HIGH RISK USES. NVIDIA SHALL NOT BE LIABLE TO CUSTOMER OR ANY THIRD PARTY, IN WHOLE OR IN PART, FOR ANY CLAIMS OR DAMAGES ARISING FROM SUCH HIGH RISK USES.

NVIDIA makes no representation or warranty that the product described in this guide will be suitable for any specified use without further testing or modification. Testing of all parameters of each product is not necessarily performed by NVIDIA. It is customer’s sole responsibility to ensure the product is suitable and fit for the application planned by customer and to do the necessary testing for the application in order to avoid a default of the application or the product. Weaknesses in customer’s product designs may affect the quality and reliability of the NVIDIA product and may result in additional or different conditions and/or requirements beyond those contained in this guide. NVIDIA does not accept any liability related to any default, damage, costs or problem which may be based on or attributable to: (i) the use of the NVIDIA product in any manner that is contrary to this guide, or (ii) customer product designs.

Other than the right for customer to use the information in this guide with the product, no other license, either expressed or implied, is hereby granted by NVIDIA under this guide. Reproduction of information in this guide is permissible only if reproduction is approved by NVIDIA in writing, is reproduced without alteration, and is accompanied by all associated conditions, limitations, and notices.

Trademarks

NVIDIA and the NVIDIA logo are trademarks and/or registered trademarks of NVIDIA Corporation in the Unites States and other countries. Other company and product names may be trademarks of the respective companies with which they are associated.