AI

Dragon provides first-class integration with popular AI/ML frameworks and workflows. These examples demonstrate how to use Dragon’s capabilities for distributed training, scalable data loading, LLM inference, and multi-agent orchestration on HPC clusters.

• AI-in-the-loop: Tightly couple simulation and ML inference in an HPC workflow.

• AI Agents: Build and run LLM-powered multi-agent pipelines with tools, memory, and tracing.

• Distributed Data Loading: Use Dragon-backed PyTorch datasets for scalable, high-performance training.

• DNA/RNA DataLoader: Example of domain-specific distributed data loading.

• AI-in-the-loop Workflow
  • Installation
  • Description of the system used
  • How to run
• Inference Service Examples
  • Example 1 - Single Prompt Service
  • Example 2 - Shared Chat Proxy
  • Example 3 - Structured Output
  • Example 4 - Pre-Batched Prompt List
  • Example 5 - Two Services in One Allocation
  • Example 6 - Guardrails Enabled
• AI Agent Examples
  • 01 — Single Agent
  • 02 — Multi-Agent DAG
  • 03 — Human-in-the-Loop
  • 04 — Memory Management
  • 05 — MCP Tools
  • 06 — Full Pipeline
• PyTorch Dataset Usage with Dragon Distributed Dictionary
  • Installation
  • Description of the system used
  • How to run
• A Custom Dataloader Benchmark for Biological Sequences Using DragonDataset and Dragon DDict
  • Installation
  • Description of the system used
  • How to run
  • Dragon Timings for Use Case