How to setup up a notebook editor

There are many notebook editor and executor software. This guide shows how to set up two of them: Jupyter Lab and Visual Studio Code.

Jupyter Lab

It is strongly recommended to create a separated environment for installing and running Jupyter Lab, to avoid version collisions with the dependencies of AI SDK or the end-to-end tutorial.

First, deactivate the image_classification environment. Then create jupyter_env, activate it, and install the jupyterlab package.

deactivate
python -m venv ~/venv/jupyter_env
source ~/venv/jupyter_env/bin/activate
python -m pip install --upgrade pip
pip install jupyterlab

You can simply start Jupyter Lab from this environment. When the server initializes, it prints out the URLs with which it can be accessed from your browser. Just copy the address pointing to localhost, and open it in a browser.To start the server, run:

cd ~/image_classification
jupyter lab --no-browser

When the Jupyter server is not running in a virtual machine, it can open the browser automatically. To make this work from WSL2, it requires the following command alias to be created:

echo -ne "\nalias jupyter-lab='(source ~/venv/jupyter_env/bin/activate && jupyter lab --ServerApp.use_redirect_file=False --browser=\"cmd.exe /c start %s\")'\n\n" >> ~/.bashrc

With this alias, you don't need to activate the jupyter_env before starting the server. Just step into the end-to-end tutorial directory, and invoke the jupyter-lab command.

cd ~/image_classification
jupyter-lab

Visual Studio Code

Visual Studio Code is a versatile code editor, and with the right extensions, it can connect to the virtual Python environments on WSL2, and use the registered kernels to run Jupyter notebooks.Download the VSCode installer, and install it on your system.

Once the VSCode is installed, start VSCode, select 'Open folder' and select one of our tutorial folder from the filesystem.Alternatively, you can start VSCode from a shell, also specifying the folder to open.

code ~/image_classification

At first start, it will show a popup message:

Do you trust the authors of the files in this folder?

Select Yes, I trust the authors.

When you first open a Jupyter notebook, you will be prompted to install the necessary extensions.

Do you want to install the recommended extension for Python?

Answer Yes, then wait for the extensions to install, then restart VSCode.Reopen the Jupyter notebook, and try to run it or click on select kernels in the top right corner. Select the Python environments option, then choose the Python (image_classification) kernel to run the notebook.

Known issue with updates

The proxy settings configured in your ~/.bashrc will not take effect when VSCode is started from Windows, and it tries to download its updates. The simple workaround is to first start a Linux shell, so the environment variables are initialized, then start VSCode from this shell.

NOTICE

If VSCode is stuck downloading updates on startup, try to start it from a Linux shell with the command `code`.

VSCode Extensions

To manually install the notebook extensions for Python and Jupyter, run the following script:

code --disable-telemetry --install-extension ms-python.python
code --disable-telemetry --install-extension ms-python.vscode-pylance
code --disable-telemetry --install-extension ms-toolsai.jupyter
code --disable-telemetry --install-extension ms-toolsai.jupyter-keymap
code --disable-telemetry --install-extension ms-toolsai.jupyter-renderers
code --disable-telemetry --install-extension ms-toolsai.vscode-jupyter-cell-tags
code --disable-telemetry --install-extension ms-toolsai.vscode-jupyter-slideshow