Multiple Jupyter Notebook Kernels

Goal

Use Coda or venv to manage packages environment with specific Python versions

What is virtual environment

A cooperatively isolated runtime environment that allows Python users and applications to install and upgrade Python distribution packages without interfering with the behaviour of other Python applications running on the same system.

Reference: Virtual Environments and Packages ⇗

We can choose the Conda way or the standard Python way to achieve it.

Use Conda

Ensure Conda is installed

conda –V

Update Conda

conda update conda

Create a new environment with desired Python version

conda create --name myenv python=3.6.8

Activate the new environment

source activate myenv

In the new environment, install ipykernel

conda install ipykernel

Register a new ipykernel

python -m ipykernel install --user --name myenv

Launch a new Notebook using the myenv kernel

In the notebook script !python -V may return the incorrect version. The system level Python may be returned, and not the version in use in the current kernel environment.

---

Use virtualenv(venv)

Open Terminal and run, it will create a venv corresponding to the specific_python_version

/path/to/specific_python_version -m venv myenv_py_version

and activate/switch to the venv:

source myenv_py_version/bin/activate

then install iPython kernel package and register a new ipykernel in the current venv:

pip install ipykernel
python -m ipykernel install --user --name=myenv_py_version --display-name "Python <specific_version>"

list available kernel spec for the verification

jupyter kernelspec list

# e.g. output
$ Available kernels:
  myenv_py_version      /home/jovyan/.local/share/jupyter/kernels/myenv_py_version
  python3    /home/jovyan/.local/share/jupyter/kernels/python3

Activate virtualenv

source myenv_py_version/bin/activate

Launch Notebook with specific kernel/venv

Switch Notebook Kernel anytime

Verify Python version in Notebook

# run the code in a cell
import sys
sys.version_info

# e.g. output
# sys.version_info(major=3, minor=6, micro=8, releaselevel='final', serial=0)

Install the specific library version in the virtual environment

Activate the specific virtual environment.

The Conda way

source activate myenv

or the standard Python way

source myenv_py_version/bin/activate

Then install the specific library version in this virtual environment. The dependency in this environment is independent from others.

pip install tensorflow==2.1

The library/package version varies with different kernels(virtual environment)