UMass Boston

Running Jupyter using interactive job submission:

The following command will request an interactive job with shell on the first available node:

srun -n 1 --gres=gpu:GTX980:1 -t 08:40:00 --pty bash

Make note of the node, the shell is receiving on (gibbs18 in the above example)

Load required modules

[john.doe001@gibbs18 ~]$ module load python/3.5.1

Start Jupyter, and look for port where jupyter is started (default is 8888 )

[john.doe001@gibbs18 ~]$ jupyter-notebook --no-browser

Result of the above command

[I 12:32:45.650 NotebookApp] Serving notebooks from local directory: /shared/home/john.doe001
[I 12:32:45.650 NotebookApp] 0 active kernels
[I 12:32:45.650 NotebookApp] The Jupyter Notebook is running at: http://localhost:8888/
[I 12:32:45.650 NotebookApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).

Tunnel the port using your local machine, and use the node name noted above

ssh john.doe001@gibbs.umb.edu  -L  8002:localhost:8001 ssh gibbs18  -L
8001:localhost:8888

Final port (8888) should match with jupyter port, the two middle ports should match with each other and first port (8002) will be the local machine port, which will be used on browser using http://localhost:8002/

Note: Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).