User Guides

To provide comprehensive guide for HPC users and native cloud users.

Danger

This guide is still under active development, and we make no promises about the reliability of content. Your feedback and contribution help improving document.

Get started

• Home
  • Getting Started
  • How-To: Contribute to this Document
• Quick Start
  • Prelude Instructions
• Software Guide
  • RNA and ChIP-sequencing analysis
  • Single Particle Analysis:
  • Software Define Radio and Cloud Radio Access Network:
• Macromolecular Modeling and Visualization:
  • Amber:
  • VMD:
• GROMACS: High Performance Molecular Dynamics
  • Example GROMACS on 4 GPUs 128CPU MPIs
  • Download Benchmark Dataset
  • Run on Singularity:
  • Performance x 2 with Direct GPUs NVLINK
• Pilot Test: Cluster Access and Performance Awareness:
  • Given Test at:
  • Features:
  • Principal investigators (PIs) or Project Owners:
  • HPC system administrators:
  • Center directors:
• SLURM Job Submit Guide
  • Basic SLURM Script
  • คำอธิบาย SBATCH Options
  • การใช้ /scratch
  • Multi-GPU Job
  • Multi-Node Job
  • คำสั่งที่ใช้บ่อย
• SLURM Array Jobs Guide
  • Quick Start
  • Why Array Jobs?
    • 1. Parallel Execution
    • 2. QOS Limit Workaround
    • 3. Fairness & Resource Utilization
  • FFT Frequency Analysis Example
    • Task Distribution
    • scripts/fft_analysis.py
    • slurm/fft-example.sh
  • Smoke Test (No Real Data)
    • scripts/fft_smoke_test.py
    • slurm/fft-smoke-test.sh
    • Run Smoke Test
  • QOS Limits Explained
    • What You See
    • Meaning
    • Why QOS Limits Exist
    • How Array Jobs Help
    • Check Your Cluster’s QOS
  • Commands Reference
  • Resource Efficiency
    • Task Sizing Guide
    • CPU-Oriented Tasks
    • Node Packing (CPU-Heavy)
    • When to Increase CPUs Per Task
    • Memory-Bound vs CPU-Bound
    • Right-Sizing Matters
  • Output Format
  • Aggregating Results
  • Best Practices
  • Summary

Math\/Physics tools

• Mathematical modeling in Physics Engineering, Biological and Social Science with QwQ LLM from China
  • Prompt:
  • Response:
  • Interpretation:
  • References:
• Quantum Machine Learning
  • Quantum-classical hybrid paradigm and variational quantum algorithms
  • Covalent
  • Quantum computing with Dask Parallel:
• Mathematics and Statistics Tools
  • MATLAB singularity with License server enabled
  • RStudio on Jupyter with The Rocker Project
  • RStudio Server(Optin II)
  • Troubleshouting Rstudio::
  • Clean Stal files::
  • Start container ::
  • RStudio Customization Library

Bioinformatics tools

• Biomedical Application
  • BLAST
  • Allocate resource for testing
  • Making example file
  • Blasting
• Reproducting research results with Python and Conda Package management
  • How to
• Nextflow:Reproducible scientific workflows
  • Run Nextflow core or nf-core AlphaFold2 workflow
• AlphaFold 3 : The inference pipeline
  • Dataset
  • ALPHAFOLD 3 MODEL PARAMETERS TERMS OF USE
  • Step 0: Clone Alphafold 3 source code
  • Step 1: Create Alphafold 3 input
  • Step 2: Run Alphafold 3 in container, Singularity
  • Step 3: View oupput
  • Reference:
• NVIDIA Parabricks
  • Test on a compute node
  • EXAMPLE RUN
• CellBender Quick-Start Tutorial
  • Introduction
  • Installation
  • Quick-Start Tutorial: remove-background
  • Downstream Analysis
  • Usage Guide: Running on Real Data
  • Examining Output H5 Files
  • Running on HPC (Example: Biowulf/Slurm)
  • Terra / WDL Workflow
  • Command-Line Reference
  • Restarting from Checkpoints
  • Troubleshooting
  • Additional Resources

Large Model tools

• LLaMA
• LoRA: AI at the edge is comming.
• Goolge’s LLM: Gemma
• Responsible AI
  • Inference Gemma on DGX A100:
  • Get up and running with large language models on private cloud
• Get Source repository and make them work on K8S for Ollama
• RestAPI test on host that port forwarded
  • New LLaMa 3.1 NOW suport Thai
• Multi-lingual capabilities in Llama3.1 405B
• API inferencing check available model :
  • Build LLM Apps with Low-code
• 🚀 Deploying Qwen 3.5 122B on Mahidol Cluster
  • High-Performance Coding & Chat via 4-bit AWQ Quantization
  • 📊 Model Specifications
  • 💻 IDE Integration (VS Code)
  • 🌐 Open WebUI Deployment
  • 🧪 Verification & Testing
  • 🛠 Troubleshooting & Maintenance
  • 📝 Usage Notes
• Metasearch for AI-Powered Organizations
  • A Guide to SearXNG Federated Search on the Cluster
  • What is Metasearch?
  • Why Metasearch Matters for Organizations
  • Why Metasearch is Critical for AI Agentic Systems
  • Using SearXNG on This Cluster
  • Troubleshooting
  • Summary
  • References
• OpenClaw — Local Agentic AI Framework
  • For Mahidol University AI Center — Researchers, Students & Staff
  • Executive Summary
  • User Guide — Setting Up OpenClaw with Qwen3.6-27B
  • Connecting Qwen3.6-27B (Full Method)
  • Customizing Your Agent
  • What Can Your OpenClaw Agent Do? (Quick Reference)
  • Security Hardening — User Level
  • Troubleshooting
  • Getting Help

Engineering

• Ansys HFSS
  • Create tmp and design folders.
  • Activate singularity and run hfss image
  • Stop singularity instance
• Gowin FPGA AI Workflow — User Guide
  • Mahidol University AI Center · Design House R&D Laboratory
  • Executive Summary
  • System Architecture
  • Technical Perspective
  • Quick Start
  • FPGA Hardware Setup
  • MCP Tools Reference
  • Workflow Examples
  • Obsidian Integration
  • Slack Integration
  • Gowin Toolchain Reference
  • Security & Operations
  • Troubleshooting
  • Quick Reference Card

DevOps - AI

• DevOps Engineering HPC & AI
  • Claude Code on Windows + Remote SSH to HPC
  • Part 1: Local Windows Setup
  • Part 2: How the Remote SSH Connection Works
  • Part 3: Connect VS Code to Your HPC
  • Part 4: Install Node.js Locally on HPC (No Module System)
  • Part 5: Install Claude Code on the HPC
  • Part 6: Install the VS Code Extension (Remote Side)
  • Part 7: Authenticate Claude Code
  • Part 8: The “Teleport” Feature
  • Important HPC Considerations
  • Quick Reference: Full Setup Checklist
  • Architecture Diagram

Caution!

• HPC User Guide: Avoiding /tmp Full Errors on Compute Nodes
  • Why This Matters
  • 1. Use $SLURM_TMPDIR for Temporary Files
  • 2. Singularity / Apptainer
  • 3. Enroot / Pyxis (NVIDIA Containers)
  • 4. Python and pip
  • 5. PyTorch / Deep Learning Frameworks
  • 6. Recommended ~/.bashrc Block
  • 7. Quick Reference: What NOT To Do
  • 8. Check Your Disk Usage
  • Contact
• Common Warnings — HPC Cluster (Mahidol DGX)
  • WARNING: Do NOT run tmux on compute nodes via SLURM
• Zeta A100 SXM4 80GB — Singularity Precaution Guide
  • Use --nvccli (Preferred) on Zeta
  • Pre-flight Check Script
  • Why --nvccli is Preferred on Zeta
  • Portable Flag Detection (use in all sbatch scripts)
  • Node Reference Table (Tested 2026-03-26)
  • Why --nvccli is Recommended
  • User Trip — Cancelling a Live GPU Job (Learned 2026-03-21)
  • Symptoms of Wrong Flag