• Here I am assuming the remote server with Deep learning setup on Nvidia GPU is on Ubuntu.
  • Install Openssh server on Ubuntu machine is not present already.
    • sudo apt-get install openssh-server
    • Password required of remote machine
  • SSH from your local computer to remote Ubuntu machine
    • ssh remoteuser@remoteIP
  • Setup conda virtual envs in remote system if not already resent as per requirement
    • conda activate <env_name>
  • Fire up a jupyter lab session with below command
    • remoteuser@remotehost: jupyter lab --no-browser --port=XXXX
    • XXXX - any port in your remote you wish to use, by default its 8888 or 8000
    • you would get a link in this form - http://localhost:8888/?token=<token>
  • Forward port XXXX (on remote machine) to YYYY (on local machine) and listen to it.
    • localuser@localhost: ssh -N -f -L localhost:YYYY:localhost:XXXX remoteuser@remotehost
    • Password required of local machine
    • In case the YYYY port is busy, run below commands to kill the process
      • lsof -n -i4TCP:YYYY : to get the PID for this port
      • kill <PID>
  • Open the link localhost:YYYY in browser in local machine.