Introduction to user-managed notebooks
Vertex AI Workbench user-managed notebooks instances let you create and manage deep learning virtual machine (VM) instances that are prepackaged with JupyterLab.
User-managed notebooks instances have a preinstalled suite of deep learning packages, including support for the TensorFlow and PyTorch frameworks. You can configure either CPU-only or GPU-enabled instances.
Your user-managed notebooks instances are protected by Google Cloud authentication and authorization and are available by using a user-managed notebooks instance URL. User-managed notebooks instances also integrate with GitHub and can sync with a GitHub repository.
User-managed notebooks instances save you the difficulty of creating and configuring a Deep Learning virtual machine by providing verified, optimized, and tested images for your chosen framework.
You can configure a user-managed notebooks instance to include the following:
JupyterLab (see version details)
Python 3, with key packages:
- fairness-indicators for TensorFlow 2.3 and 2.4 user-managed notebooks instances
- many others
R version 4.x, with key packages:
- rpy2 (an R package for accessing R in Python notebooks)
- many others
Nvidia packages with the latest Nvidia driver for GPU-enabled instances:
- CUDA 9.x, 10.x, and 11.x
- CuDNN 7.x
- NCCL 2.x
JupyterLab version details
JupyterLab 3.x is preinstalled on new user-managed notebooks instances by default. For instances created before the M80 Deep Learning VM release, JupyterLab 1.x was preinstalled.
To create an earlier version of a user-managed notebooks instance, see Create a specific version of a user-managed notebooks instance.
VPC Service Controls
VPC Service Controls provides additional security for your user-managed notebooks instances. For more information, see the Overview of VPC Service Controls. To use user-managed notebooks within a service perimeter, see Use a user-managed notebooks instance within a service perimeter.
You can upgrade your environment to use new capabilities and to benefit from framework updates, package updates, and bug fixes. You can upgrade environments manually or through an automatic update setting. To learn more, see Upgrade the environment of a user-managed notebooks instance.
User-managed notebooks and Dataproc Hub
Dataproc Hub is a customized JupyterHub server. Administrators can create Dataproc Hub instances that can spawn single-user Dataproc clusters to host user-managed notebooks environments. For more information, see Configure Dataproc Hub.
User-managed notebooks and Dataflow
You can use user-managed notebooks within a pipeline, and then run the pipeline on Dataflow. For information about how to create an Apache Beam user-managed notebooks instance that you can use with Dataflow, see Developing interactively with Apache Beam notebooks.
Consider the following limitations of user-managed notebooks when planning your project:
User-managed notebooks instances are highly customizable and can be ideal for users who need a lot of control over their environment. Therefore, user-managed notebooks instances can require more time to set up and manage than managed notebooks instances. Managed notebooks instances can be more ideal for users who don't need a lot of control over their environment. For more information, see Introduction to managed notebooks.
Third party JupyterLab extensions are not supported.
For Dataproc Hub user-managed notebooks instances, disabling file downloading from the JupyterLab user interface is not supported. User-managed notebooks instances that use the Dataproc Hub framework permit file downloading even if you don't select Enable file downloading from JupyterLab UI when you create the instance.
When you use Access Context Manager and BeyondCorp Enterprise to protect managed notebooks instances with context-aware access controls, access is evaluated each time the user authenticates to the instance. For example, access is evaluated the first time the user accesses JupyterLab and whenever they access it thereafter if their web browser's cookie has expired.
To get started with user-managed notebooks, create a user-managed notebooks instance, open JupyterLab, and try one of the samples in the tutorials folder.
Then install dependencies that you'll need to do your work.