A notebook runtime defines what Python libraries you have available within your notebook. Within ArcGIS Notebooks in Online there are 3 different runtimes available:
- ArcGIS Notebook Python 3 Standard
- ArcGIS Notebook Python 3 Advanced
- ArcGIS Notebook Python 3 Advanced with GPU support
All notebook runtimes use Python 3 libraries, older versions of Python are not supported.
ArcGIS Notebook Python 3 Standard
This includes open source libraries and the ArcGIS API for Python (it does not include ArcPy). Use this runtime for data exploration and visualization using open source libraries such as pandas, seaborn, or matplotlib, perform analysis using the Standard Analysis tools, and perform administrative tasks such as managing users or content in your ArcGIS Online organization.
ArcGIS Notebook Python 3 Advanced
This includes all open source libraries, the ArcGIS API for Python, and ArcPy. Use this runtime when you want to access tools within ArcPy that are not yet available within the Standard Analysis tools. Note that you will need to set up "local" file geodatabases and scratch environments in the Files section of your Notebook environment as part of your workflow. There are many sample notebooks available in the Samples gallery than can show how to do this.
ArcGIS Notebook Python 3 Advanced with GPU support
This includes open source libraries, the ArcGIS API for Python, ArcPy, and NVIDIA libraries that will enable GPU versions of the open source libraries to be used. The GPU runtime is designed to support resource intensive data science workflows such as model training or inferencing.
You will notice that for the time being in the beta the runtimes are tagged with their respective counterpart version in the ArcGIS Notebook Server ecosystem (10.7.1) - you can use the ArcGIS Notebook Server Python Manifest for 10.7.1 to see a full list of libraries included with each runtime version (note that in ArcGIS Notebook Server, GPU support is not explicitly called out as a version because ArcGIS Enterprise users must provide, configure, and enable their own GPU infrastructure).
