Skip to content

How to Launch Text Lab

Text Lab runs as an interactive app on the University of Bern's High Performance Computing cluster (UBELIX) via the Open OnDemand portal.

IMPORTANT:
You must activate your UBELIX account before you can use Text Lab.
See the UBELIX account activation guide for details.

Step-by-Step Guide

  1. Log in to the UBELIX Open OnDemand Portal.
  2. Navigate to Interactive Apps or Data Science Lab Services and select Text Lab.
  3. You will see a configuration form.

Basic Configuration

By default, only the essential options are shown:

Parameter Recommended Setting Description
Account gratis or your available account Select the account under which the session should run.
Job Time (hours) 1 to 4 Total time you plan to use Text Lab. If the session reaches the selected limit, it will stop automatically.
Advanced Slurm Options Unchecked for most users Leave unchecked for standard usage. Enable it only if you need to manually choose GPU and other Slurm settings.

Advanced Slurm Options

If you check Advanced Slurm Options, additional parameters will appear.

Parameter Recommended Setting Description
SLURM Partition gpu Standard GPU partition for most users. Use gpu-invest only if your account and project require it.
Quality of Service (QoS) Depends on your account Choose a QoS that matches your account and partition. For many standard users, job_gratis is the correct choice.
GPU Type rtx4090 for most tasks Suitable for most transcription and OCR tasks. Use A100, H100, or H200 only if you need more GPU memory or want to run larger LLMs.
Number of GPU(s) requested 1 One GPU is enough for most use cases. Request more than one only for large LLM workloads.
wckey Optional Provide a valid project identifier for accounting if needed. This is usually optional when using the gratis account.
Reservation Optional Only fill this in if you have been given an active Slurm reservation name, for example for a workshop or course.

Notes on QoS and GPU Selection

  • job_gratis is typically the default choice for users launching Text Lab with a gratis account.
  • job_gpu or job_gpu_preemptable may be relevant if you are using a specific project allocation.
  • A100 or H100 are recommended only for larger LLM workloads and usually together with job_gpu_preemptable, depending on your account setup.
  • If you are unsure, start with the default settings and only enable advanced options when necessary.

Starting the Session

  1. Click Launch.
  2. Wait for the job to start. The status will change from Queued to Running.
  3. Click Connect to Text Lab to open the interface in your browser.

More Information

For more information about Slurm parameters and job configuration on UBELIX, see the UBELIX documentation.