Connecting
Log in to your machine over SSH once your reservation is live, and keep long jobs running when you disconnect.
On this page
You can connect once your reservation is Live. At the start it shows as Setting up for a few minutes while the machine is prepared; it's ready within about five minutes of the start.
Find your SSH command
While the reservation is live, the SSH command appears in two places:
- at the top of your Dashboard, and on the reservation's page in My Reservations, with a copy button, and
- in the Reservation live email.
It looks like this:
ssh user@my-llm-box.res-4ad0a522.ssh.gputolease.aiThe address combines your instance name and the reservation ID, so it's different for every reservation. Always copy it from your dashboard rather than typing it. The command isn't shown before the reservation starts or after it ends.
Connect
Paste the command into a terminal (Terminal on macOS, any shell on Linux, PowerShell or Windows Terminal on Windows) and press Enter.
The first time you connect to a machine, SSH asks you to confirm its host key:
The authenticity of host 'my-llm-box.res-4ad0a522.ssh.gputolease.ai' can't be established.
ED25519 key fingerprint is SHA256:…
Are you sure you want to continue connecting (yes/no/[fingerprint])?Type yes. SSH remembers the machine, and you won't be asked again for this address.
If your private key isn't the default file, name it with -i:
ssh -i ~/.ssh/my_other_key user@my-llm-box.res-4ad0a522.ssh.gputolease.aiTo avoid retyping these options, add the machine to your SSH config.
Keep jobs running when you disconnect
When your SSH session closes, the programs started in it stop too. For anything long — a download, a benchmark, a model server — run it inside a terminal multiplexer so it keeps going. With tmux:
tmux new -s work # start a session and run your job inside it
# press Ctrl-b then d to detach; the job keeps running
tmux attach -t work # reattach later, after reconnectingIf tmux isn't installed (macOS doesn't include it), screen works the same way where it's available:
screen -S work # start a session and run your job inside it
# press Ctrl-a then d to detach
screen -r work # reattach laterFor a single command, nohup is enough:
nohup python train.py > train.log 2>&1 &
tail -f train.logReaching a server on the machine
To use a web UI or an API running on the machine — for example an OpenAI-compatible server on port 8080 — forward the port over SSH instead of opening it to the internet:
ssh -N -L 8080:localhost:8080 user@my-llm-box.res-4ad0a522.ssh.gputolease.aiWhile that runs, http://localhost:8080 on your computer reaches port 8080 on the machine.
Using another computer
The machine is set up with the one key you chose for the reservation. To connect from a second computer, copy that key pair to it, or once connected, add the second computer's public key to ~/.ssh/authorized_keys on the machine for the rest of the reservation.
Can't connect?
See Troubleshooting for Permission denied (publickey), host key warnings and timeouts.