Before accessing the SLAC JupyterLab, please apply for a SLAC computing account.
You need to use your “SLAC ID” (aka SLAC Windows account) to login to SLAC JupyterLab. For more information regarding “SLAC ID”, please refer to “SDF: New SLAC…”.
SLAC JupyterLab will put you on a new 25GB home directory /sdf/home/<username_initial>/<username>
. If you have GPFS or AFS spaces at SLAC, you will find that the JupyterLab can access GPFS spaces but can not access AFS space. You will need to manually copy your files from AFS to the new home. In JupyterLab’s terminal, you can run scp/sftp to copy files.
Once you login, click “Interactive Apps” from the top menu bar. Then choose “Jupyter”. You will need to make a few choices:
The ATLAS instance we built may not satisfy your need. If you have your own Jupyter environment that is accessible from SLAC (on SLAC disk or in CVMFS), you may be able to run it on SLAC’s Jupyter infrastracture. To do so:
jupyter notebook
or jupyter lab
(depend on whether you choose Jupyter Notebook or Jupyter Lab) to launch your Jupyter environment.export SINGULARITY_IMAGE=/gpfs/slac/.../my_singularity_image.sif
function jupyter() { singularity exec --nv -B /gpfs,/scratch,/nfs,/gpfs ${SINGULARITY_IMAGE} jupyter $@ }
or from a Conda environment (assuming Anaconda 3 is installed at ~/anaconda3):
source ~/anaconda3/etc/profile.d/conda.sh
conda activate
The above atlas-jupyter/20200502 instance resides in a Singularity image. You can use it at anywhere as long as the host can access the following CVMFS file. For example, on cent7a.slac.stanford.edu, you can run this command by hand:
singularity run -B /cvmfs,/gpfs,/scratch,/nfs,/afs /cvmfs/atlas.sdcc.bnl.gov/jupyter/t3s/slac/singularity/atlas-slac-w-slurm.sif
(add --nv
after "run"
if the host supports Nvidia CUDA GPUs). When you see it prints out a line like the following,
http://localhost:8888/?token=ec4d404fe69d2ff760d611c0509a9e8ac770c7f46ac32860
then use ssh -L 8888:localhost:8888 cent7a.slac.stanford.edu
to create a SSH tunnel. After this, paste the above URL in your browser to access your jupyter instance.
Note centos7.slac.stanford.edu
is a DNS alias of several machines cent7[a-d].slac.stanford.edu
. Do not use the DNS alias centos7
in the above case, use cent7[a-d]
instead.
The Jupyter environment provides several kernels and extensions. This includes:
Python’s pip module allows users to add packages to the JupyterLab environment as they need. For example, one can use PYCUDA and DASK distributed scheduling with SLURM. Check out this doc on how to do that.
Please use the following e-mail addresses to get help. The division below is not strict. Questions will be routed to appropriate staff members.