Skip to content

dsl-unibe-ch/text_lab

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Text Lab

Text Lab Logo

This repository contains the code for the Text Lab – an in-development suite of NLP and text processing tools. Currently, the app includes:

  • Transcribe: A tool for processing audio input and converting it to text.
  • Chat: An interactive chat interface powered by LLMs.

More NLP/text processing functionalities will be added over time.


Overview

The Text Lab Interactive App is designed to be an interactive app, meaning that:

  • Users interact with the app via a web interface.
  • The app provides real-time responses, such as streaming outputs.
  • All tasks (transcription, chatting, etc.) occur within a unified, user-friendly interface.

Note: This app is still under active development and may have some issues or incomplete features.


Launching the App

When launching the app, you can specify several parameters to tailor its execution:

Job Time (hours)

  • Job Time (hours): Specify the maximum duration for which the app will be active. After this time, the app will automatically stop. This parameter helps ensure fair allocation of compute resources.

GPU Type

  • GPU Type: You can request a specific GPU type for running your job.
  • Preemptable Option in Quality of Service (QoS): For example, if you request an A100 GPU, note that it is available as a preemptable resource.
    • A preemptable GPU resource can be reclaimed by the system if needed, which might result in interruptions.
    • This option allows efficient sharing of high-end GPUs among many users.

For more detailed guidelines on job submission and GPU options, please refer to the HPC Documentation.


How to Launch the App

To launch the app, you typically submit a job with parameters such as:

  • Job Time (hours): How long the app should remain active.
  • GPU Selection: Which GPU to use.
  • SLURM Partition: GPU or GPU-invest
  • Number of GPU(s) requested: If you plan to use the chat option with a very larg model, then you might need to request more than 1 GPU.please refer to the HPC Documentation.

The app runs only for the specified duration, ensuring effective resource usage on the HPC system.

The app is accessible via OnDemand HPC and is available only within the Unibe internal network.


Support & Further Information

For support with the app or related NLP services, please contact:
[email protected]

For additional details on job submissions and HPC resource usage, check the HPC Documentation.


About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Packages

No packages published