ARE_JupyterLab_Guide.md

Guide to Launch JupyterLab on ARE

Prerequisite

Project membership:

• vp91 (Note: after the request is approved, the system takes around 20-30 mins update your account.)

Step 1. Make User Directory

Create a username directory under /scratch/vp91:

mkdir /scratch/vp91/$USER

This should be the place where you store all your training material.

Step 2. Clone GitHub Repo

The material and data for today’s session are available in our public GitHub repository, clone it to the folder above.

cd /scratch/vp91/$USER    
git clone https://github.com/zora-wuw/llm101.git

Step 3. Prepare JypyterLab in ARE

At your browser go to are.nci.org.au, login with your NCI account. At the interface go to JupyterLab.

Step 4. Configure Job Settings

The parameters depend on the content of the workshop. Below are general use only unless specified by your instructor.

Queue: normal (Note: this is a free text field)
Compute size: large
Project: vp91
Storage: scratch/vp91 (Note: No starting slash)

Step 5. Configure Modules and Environments

Below are examples only. Please check the workshop resources for correct settings.

Click on the Advanced options

Modules: python3/3.10.4 (Note: One space only between modules)
Python or Conda virtual environment base: /scratch/vp91/Training-Venv/llm101_env
Environment variables: HF_HOME="/scratch/vp91/Training-Data/HF_models/"