Skip to content

FAQ – ML TUKE

Frequently asked questions about the machine learning platform at TUKE.

Basic information

What is ml.cloud.tuke.sk?

University platform for machine learning built on Kubernetes + Kubeflow.

Users get:

  • Jupyter Notebooks
  • RStudio
  • VS Code (code-server)
  • 4 CPU, 4 GB RAM, a 60 GB storage quota
Who can use the service?

Every student and employee of TUKE. Just log in with your TUKE login.

Do I need to request anything?

No. Your own namespace is created automatically on first sign-in. A storage volume is created when you create a notebook.

Limits and resources

What are the limits for my notebook?
Parameter Value
CPU 4 vCPU
RAM 4 GiB
Storage 60 GiB across all volumes
GPU Not available
Can I use GPU?

Currently no. GPUs are not available in ml.cloud.tuke.sk.

Access to GPUs will be addressed via separate HPC/AI infrastructure.

What if I need more CPU/RAM?

Contact support:

vcloud@helpdesk.tuke.sk

Capacities are allocated individually for courses, research, and projects.

Notebooks

How do I start a Jupyter Notebook?
  1. Log in to ml.cloud.tuke.sk
  2. In the menu, click Notebooks
  3. Press New Notebook
  4. Select image (pytorch, tensorflow)
  5. Set CPU, RAM, Volume
  6. Click Launch
  7. After status Running, click Connect
Can I use a custom Docker image?

Yes. In the Custom Image field, enter:

repository.cloud.tuke.sk/ml/jupyter-pytorch-plus:v1.0.0
What if the notebook does not start or stays in Pending?

Most common reasons:

  • Reached CPU/RAM limits
  • Incorrectly entered Docker image
  • No free capacity

Solution: Reduce requested CPU/RAM, use official Kubeflow image.

Can I shut down the notebook and continue later?

Yes. The notebook can be stopped at any time, data remains saved on PVC.

Data and files

Can I save my data?

Yes. Files are stored on the notebook's workspace volume:

  • It is persistent
  • Remains preserved after shutting down notebook
  • Is accessible in all notebooks in namespace
How do I download files from the notebook?
  • Upload/Download directly in JupyterLab
  • Create ZIP and download via browser
  • Use Git (most recommended for code)

Images and environments

What is the difference between TensorFlow, PyTorch, and Generic image?
Image Use Case
PyTorch Neural networks, NLP, CV
TensorFlow TensorFlow/Keras models
Generic Lightweight notebooks, data analysis
Where do I find shared datasets for lab exercises?

When creating a notebook, tick Globálny zdieľaný dataset under Configurations. The folder is shared by all users, has a 200 GB capacity and its contents are wiped once a year.

Terms of use

Does my data survive deleting the notebook?

Yes. The workspace volume exists independently of the notebook and stays in the Volumes section. Delete it separately if you no longer need it.

Storage and files

How do I reach my files without a running notebook?

In the Volumes section click the blue folder icon next to a volume. A web file browser opens where you can upload and download files.

My PyTorch dataloader reports a shared memory error

When creating the notebook, turn on Enable Shared Memory under Miscellaneous Settings.

Can I work in VS Code or RStudio?

Yes. The very first step of creating a notebook offers three environments: JupyterLab, Visual Studio Code and RStudio. No custom image is needed.

Creating a notebook

Reproducibility

How do I ensure reproducible code?
  • Use requirements.txt or conda environment.yml
  • Save projects to Git repository
  • Use fixed package versions
  • Do not work in default system env

Support

Where are technical problems reported?

vcloud@helpdesk.tuke.sk

When reporting, include:

  • Time of problem
  • Notebook name
  • Selected image
  • Screenshot of error