mlpractical/kernel_issue_fix.md

How to fix notebook's "kernel issues" on DICE

Some of the people in mlpractical have been affected by a recent update to the numpy and numercial library pushed to DICE last week. The symptom was restarting notebook kernel when someone tried to run the exercise involving numpy usage.

The reason of this most likely affected people who either 1) ended up with default atlas libraries (which has been updated in the meantime) or 2) re-compiled numpy with new DICE OpenBLAS already available, but LD_LIBRARY_PATH pointed to the version compiled last time - which could introduce some unexepcted behaviours.

Fix

Follow the below setps before you activate the old virtual environment. The fix basically involves rebuilding the virtual environments. But the whole process is now much simpler (due to the fact OpenBLAS is now a deafult numerical library on DICE).

1. Comment out (or remove) export=$LD_LIBRARY_PATH... line in your ~/.bashrc script. Then type unset LD_LIBRARY_PATH in the terminal. To make sure this variable is not set, type export and check visually in the printed list of variables

2. Go to ~/mlpractical/repos-3rd/virtualenv and install the new virtual environment by typing:

./virtualenv.py --python /usr/bin/python2.7 --no-site-packages $MLP_WDIR/venv2

3. Activate it by typing: source $MLP_WDIR/venv2/bin/activate and install the usual for the course packages using pip:

   • pip install pip --upgrade
   • pip install numpy
   • pip install ipython
   • pip install notebook
   • pip install matplotlib

4. Now enter ~/mlpractical/repo-mlp and see whether numpy has been linked to DICE-standard OpenBLAS (and works) by starting python notebook:

ipython notebook

and running two first interactive examples from 00_Introduction.py. If they run, you can simply modify activate_mlp alias in ./bashrcto point to venv2 instead of venv

5. You can also remove both old venv amd other not needed anymore directories with numpy and OpenBLAS sources in ~/mlpractical/repos-3rd directory.