Connect and Verify
Log in, confirm the GPU is actually there, and don't start a long run until it is.
Connect with the details the dashboard gives you
Once the instance turns green (active), the dashboard hands you the login — a username and a host. Use them:
ssh <username>@<instance-ip>Use the private key that matches the public key you added when you created the instance. Heads up: never email a private key to ANK, and never commit it to a repo. One leak and someone's running instances on your dime.
Check the GPU is real
nvidia-smi doesn't lie. On a GPU image, run it and look for your card and the driver version:
nvidia-smiNothing shows up? Before you point a workload at it:
- Did you actually pick a GPU tier?
- Did you pick
ubuntu-gpuorubuntu-torch? - Is provisioning done (not still "starting")?
If yes to all and still nothing, stop and ping support before a long run turns into a wasted afternoon.
Save the recipe
Write down the tier, config, image, billing plan, and every dependency you installed. Next time you — or a teammate — need to reproduce a working setup, or support needs to figure out why something broke, you'll thank future-you.