We currently have three local machines named Bleen, Dolores, and Quadcorn.
|Dolores||Ubuntu||i9-7920X @ 2.90 GHz - 12 cores||3x GTX 1080 Ti, 1x RTX 2080 Ti||94 Gb|
|Bleen||Ubuntu||i7-6850k @ 3.60GHz - 6 cores||4x GTX 1080 Ti||64 GB|
|Quadcorn||Mint||i7-5960X @ 3.00GHz - 8 Cores||3x GTX 1080 Ti, 1x RTX 2080 Ti||64 Gb|
Getting an account
Setting Up Anaconda
The first thing you should do once you have an account is to set up Anaconda. Anaconda is a python envrionment manage. This allows seperate users to easily have different versions of python as well as different versions of packages. Individual users can also create seperate environments if different projects have conflicting version needs.
The screen function is used to manage virtual terminals that can run in the background. These virtual terminals continue to persist after you disconnect from the ssh terminal. That allows you to run long programs without needing to stay connected, as well as run jupyter notebook sessiosn that won’t crash if you disconnect. Here is a page with extensive documentation.
In the screen documentation, “C-“ refers to clicking “Ctrl+c”. For example, the command “C-a” means “Click ‘Ctrl+c’ then click ‘a’”.
Basics of Screen:
screen -S <name of screen>
The -S option allows you to create a new screen, and give it a name. This makes it easy to reconnect afterwards.
C-a is the command to detatch from the screen and return to your base terminal.
screen -r <name of screen>
The -r command allows you to reconnect to a screen with the name you have given it. To get a list of currently open and running screens, type “screen -r” without a name.
Occasionally a screen will be listed as “attached” while you are actually not connected to it. In that case use the -d command to attach to screens that for some reason remain listed as attached.
I recommend running notebook sessions within a virtual terminal from the screen function. That way if you momentarily lose connection, your session won’t be closed.
To start a jupyter notebook or lab session type:
jupyter notebook --no-browser --port=<XXXX> or jupyter lab --no-browser --port=<XXXX>
where XXXX is an unused port number.
Next on your local machine’s terminal type:
ssh -N -f -L localhost:YYYY:localhost:XXXX email@example.com
Use the same XXXX and above. YYYY is any port number available on your local machine.
Finally, open your browser and connect to ‘localhost:YYYY’. You may be prompted to input a password, which is found on the screen in which you launched the jupyter session.
Please be mindful of the storage you are using, especially in the home folder. Most machines have seperate hardrives, generally under a folder named /data
df -h to get a list of all drives and their available space.