AUP Learning Cloud

AUP Learning Cloud is a tailored JupyterHub deployment designed to provide an intuitive and hands-on AI learning experience. It features a comprehensive suite of AI toolkits running on AMD hardware acceleration, enabling users to learn and experiment with ease.

Quick Start

The simplest way to deploy AUP Learning Cloud on a single machine in a development or demo environment.

Prerequisites

• Hardware: AMD Ryzen™ AI Halo Device (e.g., AI Max+ 395, AI Max 390)
• Memory: 32GB+ RAM (64GB recommended)
• Storage: 500GB+ SSD
• OS: Ubuntu 24.04.3 LTS
• Docker: Install Docker and configure for non-root access

# Install Docker
curl -fsSL https://get.docker.com | sh

# Add current user to docker group
sudo usermod -aG docker $USER

# Apply group changes without logout (or logout/login instead)
newgrp docker

# Install Build Tools
sudo apt install build-essential

Note: See Docker Post-installation Steps and Install Docker Engine on Ubuntu for details.

Installation

git clone https://github.com/AMDResearch/aup-learning-cloud.git
cd aup-learning-cloud/deploy/
sudo ./single-node.sh install

After installation completes, open http://localhost:30890 in your browser. No login credentials are required - you will be automatically logged in.

Script Commands

Command	Description
install	Full installation (K3s, tools, GPU plugin, images, JupyterHub)
uninstall	Complete removal of all components
upgrade-runtime	Upgrade JupyterHub deployment
build-images	Build and import container images
pull-images	Pull external images for offline use
install-tools	Install Helm and K9s only
install-runtime	Deploy JupyterHub only
remove-runtime	Remove JupyterHub only

Example:

# Upgrade JupyterHub after configuration changes
sudo ./single-node.sh upgrade-runtime

# Rebuild images after modifying Dockerfiles
sudo ./single-node.sh build-images

💡 Tip: If you need to use alternative container registries or package mirrors, see Mirror Configuration.

Manual Installation

For users who prefer step-by-step manual installation or need more control over the deployment process:

• Single-Node Manual Deployment - Detailed manual setup for development and demo environments
• Multi-Node Cluster Deployment - Production deployment with Ansible playbooks

Learning Solution

AUP Learning Cloud offers the following Learning Toolkits:

Important

Only Deep Learning and Large Language Model from Scratch are available in the v1.0 release.

• Computer Vision
  Includes 10 hands-on labs covering common computer vision concepts and techniques.

• Deep Learning
  Includes 12 hands-on labs covering common deep learning concepts and techniques.

• Large Language Model from Scratch
  Includes 9 hands-on labs designed to teach LLM development from scratch.

• Physical Simulation
  Includes 4 hands-on labs on building a virtual robotic arm and grasping objects.

Key Features

Hardware Acceleration

AUP Learning Cloud provides a multi-user Jupyter notebook environment with the following hardware acceleration:

• AMD GPU: Leverage ROCm for high-performance deep learning and AI workloads.
• AMD NPU: Utilize Ryzen™ AI for efficient neural processing unit tasks.
• AMD CPU: Support for general-purpose CPU-based computations.

Flexible Deployment

Kubernetes provides a robust infrastructure for deploying and managing JupyterHub. We support both single-node and multi-node K3s cluster deployments.

Authentication

Seamless integration with GitHub Single Sign-On (SSO) and Native Authenticator for secure and efficient user authentication.

• Auto-admin on install: Initial admin created automatically with random password
• Dual login: GitHub OAuth + Native accounts on single login page
• Batch user management: CSV/Excel-based bulk operations via scripts

Storage Management and Security

Dynamic NFS provisioning ensures scalable and persistent storage for user data, while end-to-end TLS encryption with automated certificate management guarantees secure and reliable communication.

Available Notebook Environments

Current environments are set up as RESOURCE_IMAGES in runtime/jupyterhub/files/hub. These settings should be consistent with Prepullers in runtime/values.yaml.

Environment	Image	Version	Hardware
Base CPU	ghcr.io/amdresearch/auplc-default	v1.0	CPU
CV COURSE	ghcr.io/amdresearch/auplc-cv	v1.0	GPU (Strix-Halo)
DL COURSE	ghcr.io/amdresearch/auplc-dl	v1.0	GPU (Strix-Halo)
LLM COURSE	ghcr.io/amdresearch/auplc-llm	v1.0	GPU (Strix-Halo)
PhySim COURSE	ghcr.io/amdresearch/auplc-physim	v1.0	GPU (Strix-Halo)

Documentation

• JupyterHub Configuration - Detailed JupyterHub settings
• Authentication Guide - GitHub OAuth and native authentication
• User Management Guide - Batch user operations with scripts
• User Quota System - Resource usage tracking and quota management
• GitHub OAuth Setup - OAuth configuration
• Maintenance Manual - Operations guide

Contributing

Please refer to CONTRIBUTING.md for details on how to contribute to the project.

Acknowledgments and Credits

AUP would like to thank the following universities and professors. This learning solution was made possible through the joint efforts of these partners.

University	Professors and Labs	Toolkits
National Taiwan University	Prof. Chun-Yi Lee, ELSA Lab	DL, CV
Nanjing University	Prof. Jingwei Xu, NJUDeepEngine	LLM

The following repositories are used in AUP Learning Cloud, either in close to original form or as an inspiration:

• Genesis