AI Training - Start a job with a notebook Docker image – Support Guides

Learn how to start a simple interactive notebook leveraging GPUs over AI Training service.

Requirements

Access to the OVHcloud Control Panel
An AI Training project created inside a public cloud project
A user for AI Training

Instructions

Step 1 - Begin as a classic job submission

Follow the same steps as a classic job submission described here until you reach the Docker image step.

Step 2 - Select the notebook corresponding to your needs

A job is basically a Docker container that is run within the OVHcloud infrastructure.

Notebooks are daemon jobs, meaning they will run indefinitely until the user requests an interruption.

AI Training offers several notebook images with different configurations. You can choose the configuration that best suits your needs.

Currently, the following configurations are available:

PyTorch: An OVHcloud preset image including JupyterLab notebook, Visual Studio Code IDE, and pytorch libraries
Tensorflow 2: An OVHcloud preset image containing JupyterLab notebook, Visual Studio Code IDE, and tensorflow 2 libraries
Hugging Face Transformers: An OVHcloud preset image containing JupyterLab notebook, Visual Studio Code IDEand hugging face libraries
MXNet: An OVHcloud preset image containing JupyterLab notebook, Visual Studio Code IDE, and mxnet libraries
Fast.ai: An OVHcloud preset image containing JupyterLab notebook, Visual Studio Code IDE, and fast.ai libraries
auto gluon: An OVHcloud preset image containing JupyterLab notebook, Visual Studio Code IDE, and AutoGluon + mxnet libraries

Step 3 - Continue as a classic job submission

Continue to follow the same steps as a classic job submission described here until you reach Step 3 - Consulting your job.

NOTE: If you want to be able to save your notebook files on your object storage, we strongly advise that you plug a read and write volume on your job before submitting. That volume will be synchronized with your object storage at the end of the job.

Step 4 - Access notebook URL

Once your job is Running, in the job Access panel, you should see the Container access link. Click on it, and you will be redirected to your job URL.

If you are not authenticated as an AI Training user, you should see a screen asking for your username and password.

If you have never created a user for AI Training yet, you can follow the instructions here.

Fill in the fields and click Connect.

Step 6 - Use your notebook

In most provided preset images, you can choose which editor you prefer between JupyterLab and VisualStudio code.

Select the one that you want to use, and you will be redirected to the corresponding editor.

By default, the home directory of your job is located under /workspace. It means that you will have read and write access to that directory as well as your read and write mounted volumes.

NOTE: If you are missing a library or a configuration, you can add it directly in the command line of the notebook's console as long as you don't need privileged access (root access). Example: pip install <...>

For installing specific libraries that require privileged access, you will have to build your own notebook image and use it as a custom image at Step 2 instead of a preset image. More information about creating your own Docker image can be found here.

If you open a console tab in your notebook and type nvidia-smi, you will see the available GPUs that you can use on your notebook.

From the Jupyter Notebook interface, the !nvidia-smi command will retrieve the same information.

Step 7 - Stop your notebook

Once you are done working with your notebook, don't forget to stop it.

You can do it by selecting more options ... and then Stop.

Then, in the pop-up window, click Stop.

After some time, your job should go into an Interrupted state, meaning that the job has been stopped.

Before going into the Interrupted state, your job may run through the Finalizing state. During this phase, all data inside read & write volumes are saved inside their linked containers in your object storage.

Go further

For more information and tutorials, please see our other AI & Machine Learning support guides or explore the guides for other OVHcloud products and services.

If you need training or technical assistance to implement our solutions, contact your sales representative or click on this link to get a quote and ask our Professional Services experts for a custom analysis of your project.