minrk
commented
5 years ago This is typically the result of the kernel not being in the same env where packages are being installed.
The quick way to check is:
- which Python is pip using (
head -n 1 $(which pip)in a terminal where you are installing packages) - which Python is the notebook using (
import sys; print(sys.executable))
If you are installing packages that you can't import, these two will likely not be the same. I suspect that one or the other is not the one you mean to be using.
You might want to check:
jupyter kernelspec listit's possible that you still have a python3 kernel registered pointing to an older location. If that's the case, you can remove it:
jupyter kernelspec remove python3and make sure that python3 in your env points to the env itself:
ipython kernel install --sys-prefix
gcgruen
It seems I am not the first one with this issue, however, the other solutions I found around the web (stackoverflow, github issues list, medium blog posts) don't solve my issue.
I have created a virtual environment in which Python 3.7 is successfully running. I also installed some modules within the virualenv (jupyter, pandas, matplotlib, requests, BeautifulSoup etc).
If I run a .py file from the command line, it works perfectly fine, all modules are loaded and used.
However, if I start a jupyter notebook, it already fails upon importing basically any of the installed modules, like for example
import pandas as pd(not exclusively for pandas though, but basically for any module.)It fails with the error message
ModuleNotFoundError: No module named 'pandas'I went on searching for an answer. The most common advice: pandas is not installed. However, it is. At first it wasn't in the right place. I did
!which jupyter,which !jupyter-notebookand!which python(separately from the command line as well as from within the jupyter notebook). All three returned with the same path to my virtual environment:/Users/lc_admin.grueng/.virtualenvs/dwdata/bin/jupyterWhen I do
!which pandasit doesn't return a thing, hinting at pandas not being installed. However, when installing it (no matter via what method), it returned with(dwdata) bmac1025559:bin lc_admin.grueng$ pip install pandas Requirement already satisfied: pandas in /Users/lc_admin.grueng/.virtualenvs/dwdata/lib/python3.7/site-packages (0.23.4)Looking at the path, I realized it is different from the path for jupyter. Thus, I installed it in the same directory using
pip install pandas --force-reinstall --target /Users/lc_admin.grueng/.virtualenvs/dwdata/bin.If I do
!which pandasagain, it still doesn't return any path. However, if I cd into said path andlsit definitely is there, together with all the other modules.I guess this is the first part of my problem which I don't know how to solve.
Then I thought: If the module is there, but Jupyter doesn't find it, maybe I need to tell Jupyter where to look. So then in the command line I added the path to the JUPYTER_PATH for datafile directory locations to my .bash_profile using
echo "export JUPYTER_PATH=/Users/lc_admin.grueng/.virtualenvs/dwdata/bin:$JUPYTER_PATH" >> ~/.bash_profile, I reloaded the bash profile withsource ~/.bash_profileandcatit to see whether the path was successfully added (which it is).But upon restarting jupyter, still I get the
ModuleNotFoundError: No module named 'pandas'error and I am out of ideas how else I could fix it.Any hints much appreciated. Operating system is Mac OSX High Sierra version 10.13.6, Python version within my virtual environment is Python 3.7.0 (outside Python 3.6.5 if it matters), I installed both with pip and sudo pip3 and I'm also through the idea with upgrades and setuptools for the modules. And I'm not a user of (ana)conda, so if there is a way to solve this without it that would be perfect.