Tools on the Desktop¶
Every desktop is a Linux workspace (XFCE) that opens in your browser tab. What is installed depends on the image your eLab administrator has selected for the desktop type you launch. This page lists what each image carries so you know what to expect before you start one.
The full CPU desktop¶
The standard CPU image is a complete data-science workstation:
| Tool | What it is |
|---|---|
| JupyterLab | Launch it from the desktop or menu; it opens in Firefox on the desktop. The base Python environment already has NumPy, pandas, scikit-learn, SciPy, statsmodels, Matplotlib and seaborn. |
| VS Code | The full editor, with the terminal built in. |
| RStudio Desktop and R | R with a CRAN-compatible package source already configured (see Installing Packages). |
| Python with conda | A Miniforge install; conda and pip are on the path in every terminal. Create your own environments in your home directory so they survive restarts. |
| LibreOffice | Writer, Calc, Impress and Draw, in the Office section of the applications menu. It reads and writes Word, Excel and PowerPoint files as well as its own formats. |
| Firefox | For JupyterLab, the git service and anything else your eLab lets you reach. |
| Git | On the command line, plus a Git shortcut on the desktop and in the menu that opens the eLab's git service where one is offered. |
| Terminal | A normal shell. Most command-line tools you would expect on a Linux workstation are present. |
A document written in Microsoft Office may look slightly different here: Arial, Calibri and Cambria are not installed, and LibreOffice substitutes metric-compatible faces of the same widths (Liberation Sans, Carlito, Caladea). Line and page breaks are preserved; the letterforms differ. If a document must be laid out to the character for printing, do that final step where the original fonts are installed.
The lite CPU desktop¶
Where your administrator offers a lite image (roughly half the size, so it starts faster and suits taught sessions), it is deliberately minimal:
- Firefox
- VS Code
- Python 3.12 with
pipandvenv
There is no conda, R, RStudio, JupyterLab or LibreOffice in the lite image. If you need those, ask your administrator which desktop type serves the full image.
GPU desktops¶
A GPU desktop is the full CPU desktop plus one deep-learning framework, ready to use on the GPU. Which one you get depends on the image your administrator has selected:
- PyTorch — a conda environment named
pytorch(PyTorch, torchvision and torchaudio built for CUDA 13), also registered in JupyterLab as the Python (PyTorch) kernel. - TensorFlow — a conda environment named
tensorflow(TensorFlow with bundled CUDA and cuDNN), registered in JupyterLab as the Python (TensorFlow) kernel.
The base Python environment stays as it is on the CPU image. Activate the framework environment
in a terminal (conda activate pytorch or conda activate tensorflow) or pick its kernel in
JupyterLab.
GPU desktops are single-framework by design: a desktop offers PyTorch or TensorFlow, not both. If you need the other one, ask your administrator rather than installing it yourself.
What persists¶
Only your home directory persists between sessions. Anything installed with pip install
--user, into a conda environment under your home, or into R's user library survives a restart.
The desktop runs unprivileged, so system-level installs are not possible, and nothing outside
your home survives anyway. See Your Desktop for the full picture of
where your files live.
Source code and licences¶
The desktop is assembled from separately licensed software. Where a licence entitles you to the source, it is offered here.
| Software | Licence | Source |
|---|---|---|
| RStudio Desktop | AGPL v3 | rstudio/rstudio — the release matching Help → About RStudio |
| LibreOffice | MPL 2.0 | LibreOffice/core — the release matching Help → About LibreOffice; shipped as Ubuntu's own package, unmodified |
| Firefox | MPL 2.0 | mozilla-firefox/firefox |
| Gitea | MIT | go-gitea/gitea |
| Your editor | see Help → About in the editor itself | — |
RStudio is licensed under the AGPL and you use it over a network, which entitles you to its corresponding source. It is shipped unmodified, so the upstream release above is that source — if you would rather receive it another way, ask your administrator.
Python packages, R packages and conda environments you install carry their own licences, set by their authors rather than by this platform.