Connect to GPU Instance

SSH access

1. Save your SSH public key

Before your first deployment, add your SSH public key in Account Settings → Security. Paste your public key (e.g. the contents of ~/.ssh/id_ed25519.pub) into the SSH Public Key field and click Save Key.

Your key is injected into every instance you deploy after saving it. Instances deployed before the key was saved will not have it — you will need to redeploy.

2. Find your connection details

Once your instance is RUNNING, click on it in the My Instances tab to open the details panel. Under Connection, you'll find the SSH host and port.

3. Connect

Use your local terminal to connect:

ssh root@<host> -p <port> -i ~/.ssh/your_key

Replace <host> and <port> with the values from the connection panel, and point -i to the private key matching the public key you saved.

You log in as root with full access. The default shell is bash. Your working directory is /workspace.

File transfer (SSH)

You can transfer files to and from your instance securely using standard SSH tools like scp or rsync.

Using SCP

To upload a file from your local machine to the instance:

scp -P <port> -i ~/.ssh/your_key ./local-file.tar.gz root@<host>:/workspace/

To download a file from the instance to your local machine:

scp -P <port> -i ~/.ssh/your_key root@<host>:/workspace/data.tar.gz ./local-backup/

Using rsync

For large directories or incremental syncing, rsync is recommended. The following command uploads a directory, skipping files that haven't changed:

rsync -avz -e "ssh -p <port> -i ~/.ssh/your_key" ./local-directory/ root@<host>:/workspace/remote-directory/

Troubleshooting

• Connection refused: Ensure your instance is RUNNING and you have correctly saved your SSH public key in Account Settings before deploying the instance.
• Permission denied: Verify that the path to your private key following the -i flag is correct, and that your private key has the proper file permissions (chmod 600 ~/.ssh/your_key).

JupyterLab access

Every GPU instance with JupyterLab enabled runs a JupyterLab server accessible from your browser. No installation or configuration is needed.

1. Get the connection URL

In the instance details panel, look for the Jupyter connection link. Click it to open JupyterLab in a new browser tab.

2. Authenticate

JupyterLab is password-protected. The auto-generated password is shown in the instance details panel under Jupyter Password. Enter it when prompted.

3. Use the IDE

JupyterLab opens in the /workspace directory. From here you can create notebooks, open a terminal, upload files, and run code — all in your browser. Files saved to /workspace persist through pause/resume cycles (and permanently if you're using a Network Volume).