Work in JupyterLab

When the instance is ready, click Open Notebook.

JupyterLab opens in a new tab. This is the default way into an instance and gives you a full development environment in the browser.

What’s in there

Terminal — a real Linux shell. Click + at the top, then Terminal, or use the Terminal tile in the Launcher. Use it to install packages, download weights, and start services.

Notebooks — .ipynb documents mixing code and output, the usual way to explore interactively.

File browser — the panel on the left. Use the upload button (↑) to bring in your own notebooks and data.

Check the GPU

From a terminal:

rocm-smi

This lists the Radeon GPUs attached to the instance along with memory use and temperature. To confirm PyTorch sees them:

python -c "import torch; print(torch.cuda.is_available(), torch.cuda.device_count())"

On ROCm builds of PyTorch, the CUDA API names are the ones to use — torch.cuda.is_available() returning True means ROCm is working.

Keeping work alive

Long jobs die if the terminal closes. Launch them detached:

nohup python train.py > train.log 2>&1 &

Files only survive the instance being destroyed if the template used Persistent (PVC) storage. Otherwise, download anything you need first.

Exposing a service

To make an HTTP service running inside the notebook reachable from the internet — a Gradio app, an API you’re developing — see Expose a service.