Connect via SSH

Once your deployment reaches the Running state, you can connect via SSH. The connection details are displayed in the Deployments page.

Connection details

Each running deployment shows:

Host — The IP address or hostname to connect to
Port — The SSH port (check your deployment details — it may not be 22)
User — The SSH user (e.g., root, ubuntu)

Connecting

Use your SSH key to connect:

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

Replace <port>, <user>, and <host> with the values from your deployment details.

You can add this to your ~/.ssh/config for convenience:

Host mygpu
  HostName <host>
  Port <port>
  User <user>
  IdentityFile ~/.ssh/your_key

Then just run ssh mygpu.

Which key to use

Platform-generated key — A default key is created for your account at signup. If you generated this key through the platform, the private key was downloaded as a .pem file when you created it.
Your own key — If you uploaded your own public key in SSH Keys, use the corresponding private key.

Make sure to use the key that was selected during deployment.

Jupyter access

If you deployed with a Jupyter-enabled template (e.g., PyTorch + Jupyter, TensorFlow + Jupyter), a Jupyter URL is displayed alongside the SSH connection details. Click it to open JupyterLab directly in your browser.

Transfer files

Use scp or rsync to move files to and from your instance:

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

# Download a file
scp -i ~/.ssh/your_key -P <port> <user>@<host>:/workspace/model.pt ./

# Sync a directory
rsync -avz -e "ssh -i ~/.ssh/your_key -p <port>" ./data/ <user>@<host>:/workspace/data/

Port forwarding

Forward remote ports to your local machine for accessing web UIs like TensorBoard:

# Forward TensorBoard (remote port 6006 → local port 6006)
ssh -i ~/.ssh/your_key -p <port> -L 6006:localhost:6006 <user>@<host>

# Forward multiple ports
ssh -i ~/.ssh/your_key -p <port> -L 6006:localhost:6006 -L 8888:localhost:8888 <user>@<host>

Then open http://localhost:6006 in your browser.

VS Code Remote SSH

You can use VS Code’s Remote - SSH extension to develop directly on your GPU instance:

Install the Remote - SSH extension in VS Code
Add your instance to ~/.ssh/config (see the config example above)
Open the Command Palette (Ctrl+Shift+P) → Remote-SSH: Connect to Host → select your instance
VS Code opens a new window connected to your GPU instance — edit files, run terminals, and debug as if it were local

Troubleshooting

Connection refused

The instance may still be booting. Wait 1–2 minutes after the status shows Running and try again.

Permission denied (publickey)

Verify you’re using the correct private key (the one matching the public key selected during deployment).
Check file permissions: chmod 600 ~/.ssh/your_key
Ensure you’re using the correct SSH user shown in the deployment details.

Connection timed out

Confirm the host and port from your deployment details.
Check that your local firewall or network allows outbound connections on the specified port.

Lost your platform-generated private key

The private key for a platform-generated SSH key is only available for download once. If you’ve lost it, generate a new key pair in SSH Keys or upload your own public key, then use that key for future deployments.

Getting Started

Help

Connection details

Connecting

Which key to use

Jupyter access

Transfer files

Port forwarding

VS Code Remote SSH

Troubleshooting

​Connection details

​Connecting

​Which key to use

​Jupyter access

​Transfer files

​Port forwarding

​VS Code Remote SSH

​Troubleshooting

Connection details

Connecting

Which key to use

Jupyter access

Transfer files

Port forwarding

VS Code Remote SSH

Troubleshooting