-
2 cases :
- you have a fix PC with a fix IP -> you can connect directly to occigen, jupyter-notebooks will render smoothly
- you have a portable PC with a variable IP -> you must connect to cal1 first then to occigen
-
Make sure to add the options
-CX
(-CXY
on MAC) when you ssh to cal1 and occigen, in order to have the pop-ups rendered (not only for notebooks but also ncview or display) or you can open a VNC interactive session for the pop up. -
When you are on occigen, you must know that
conda install
is not possible (the http server is blocked) and onlypip install
is available. There are 2 solutions:- Ask svp@cines.fr to install you a
conda
environment with a list of libraries, and each time you want a new library installed (make sure to ask forjupyter
orjupyterlab
!) - Use
conda-pack
to install aconda
environment on your occigen account. Forconda-pack
to work,conda
must be installed and be on your path. In order to set this up, first scpminiconda
for a Linux system to occigen and bash installminiconda
. Now, pack your desiredconda
environment in a local Linux environment usingconda-pack
and then scp the packed tar file to$SCRATCHDIR
on occigen. Once you've expanded the tar file, you can activate theconda
environment running$ source $SCRATCHDIR/my_env/bin/activate
. Make sure the path of the Python packages points to where yourconda
environment is ($SCRATCHDIR/my_env/bin/python
).
- Ask svp@cines.fr to install you a
-
You need firefox to render your jupyter notebook/lab: so in your .bashrc put the line
module load firefox
-
Look for the file
/home/albert7a/.jupyter/jupyter_notebook_config.py
in your occigen account and modify the lines:- Comment the line
c.NotebookApp.password = ...
. - Uncomment the line
c.NotebookApp.open_browser=True
. - Uncomment the line
c.NotebookApp.browser =
and choose firefox as the browser. - Uncomment the line
c.NotebookApp.port = ...
and set a port number that you prefer so that the notebook always first tries to use the same port number.
- Comment the line
-
If you want to use
dask-jobqueue
(to submit a SLURM job to get computational resources):- Install the
dask-jobqueue
library. - Copy the file
/home/albert7a/.config/dask/jobqueue.yaml
in your account. - The syntax of the commands to put in the notebook is described here.
- Install the