The post aims to show how to create Jupyter environments and how to debug any issues. It also provides some commands that are good for general debugging.

Table of Contents

• Debugging Commands
  • Conda environments and Jupyter kernels
  • What Python interpreter am I using?
  • What kernels are available to Jupyter?
  • Where are my kernels located?
  • What is the active kernel?
  • Removing Kernels
• Windows
• Python and environment not matching

Debugging Commands

Here are some commands I’ve found useful when debugging issues with my Jupyter environment. You’ll see them used in the examples below.

Conda environments and Jupyter kernels

Two things you’ll want to know when debugging Jupyter environment problems are what conda environments you have and what kernels are available to Jupyter. You can find your environments with either

conda env list or conda info --envs

They provide the same information.

What Python interpreter am I using?

In Python:

import sys
sys.executable

C:\\Users\\Julius\\anaconda3\\envs\\tf\\python.exe

What kernels are available to Jupyter?

jupyter kernelspec list

Where are my kernels located?

You should also be aware that different kernels will have different paths. For example, import my_package may work in one kernel but not in another. Check sys.path to see which paths are being called.

Keep in mind that you can run all of the commands either from a terminal or from a Jupyter notebook. To run a Unix command from Jupyter, you simply need to add a ! before it. So you can run !conda info --envs or !jupyter kernelspec list from a notebook.

OK, now let’s talk about some examples where you might need these.

What is the active kernel?

Here’s the only way I’ve found how to do this (source):

%%javascript
var kernel = Jupyter.notebook.kernel
kernel.execute('kernel_name = ' + '"' + kernel.name + '"')

print(kernel_name)
# my-kernel

from jupyter_client import kernelspec
spec = kernelspec.get_kernel_spec(kernel_name)
print(spec.resource_dir)

Removing Kernels

You can remove a kernel like so:

jupyter kernelspec uninstall my_kernel

Windows

For Windows users, you can use the Anaconda Prompt

jupyter kernelspec list

Should see something like C:\Users\Julius\anaconda3\share\jupyter\kernels\python3

cd there

You should see a kernel.json file. You can look inside it with type kernel.json.

{
 "argv": [
  "C:/Users/Julius/anaconda3\\python.exe",
  "-m",
  "ipykernel_launcher",
  "-f",
  "{connection_file}"
 ],
 "display_name": "Python 3",
 "language": "python"
}

Now you can add the kernel as shown above.

Python and environment not matching

If you’re in your PT environment but that’s not the Python distribution you’re seeing when you sys.executable, you’ll need to edit your kernel.json file that you found with jupyter kernelspec list.