CSCS Clariden
Quick start for accessing CSCS Clariden and validating SSH, Slurm, Python, GPU jobs, uenv, and the ML Console.
Clariden is the CSCS machine-learning cluster on Alps. FHNW/AISL users can use it when access has been allocated through a course, research project, workshop, event, or other approved activity.
JupyterHub / JupyterLab
For interactive teaching and notebook-based experimentation, use the Clariden JupyterHub.
After signing in, select the project/account, partition, wall time, and runtime environment provided for your course or project. A tested Apertus setup used:
Partition:
DedicatedRuntime image:
nvcr.io/nvidia/pytorch:25.04-py3Advanced options: left empty
Hardware exposed to the notebook: 4 x NVIDIA GH200 120 GB
For the complete notebook workflow, see Training Apertus on CSCS Clariden.
When finished, stop the Jupyter server through the Hub control panel. Logging out of JupyterLab alone does not necessarily terminate the underlying Slurm allocation.
Before you start
You need a CSCS account with MFA, access to a project or course allocation that includes Clariden, and the corresponding Slurm project/account code.
Do not share private SSH keys, passwords, CSCS credentials, Hugging Face tokens, or other secrets.
SSH and Slurm
SSH remains useful for batch jobs, command-line workflows, and debugging outside JupyterLab.
Confirm your allocation at portal.cscs.ch.
Create a local SSH key if needed.
Sign the public key at user-account.cscs.ch.
Use the SSH configuration provided by CSCS.
Connect with
ssh cscs-clariden.Run compute workloads through Slurm, not directly on the login node.
A minimal allocation check is:
CSCS ML Console
The CSCS ML Console is useful for monitoring jobs, viewing StdOut/StdErr, browsing accessible filesystems, and submitting simple jobs. It is a companion to JupyterHub, SSH, and Slurm rather than a separate compute system.
Related resources
Last updated