Raspberry Pi AI HAT+ 2

Contents

Raspberry Pi AI HAT+ 2 User Manual

Enhancing Generative AI (LLM/VLM) with Hailo-10H Processing Engine

1. Introduction

The Raspberry Pi AI HAT+ 2 is an advanced add-on board designed to bring generative AI capabilities to the Raspberry Pi 5. It integrates a powerful Hailo-10H AI accelerator with 8GB of dedicated on-board RAM, enabling efficient processing of Large Language Models (LLMs) and Vision-Language Models (VLMs) locally, without requiring internet access.

This HAT+ 2 offers similar computer vision performance to its predecessor, the Hailo-8-based Raspberry Pi AI HAT+, while significantly expanding its utility to generative AI tasks. It is ideal for applications such as offline process control, secure data analysis, facilities management, and robotics.

Top-down view of the Raspberry Pi AI HAT+ 2 board, highlighting the Hailo-10H AI accelerator chip and other components.
Figure 1: Top-down view of the Raspberry Pi AI HAT+ 2 board, showing the Hailo-10H AI accelerator.

2. Setup and Installation

To begin using your Raspberry Pi AI HAT+ 2, follow these steps:

  1. Prepare your Raspberry Pi 5: Ensure your Raspberry Pi 5 is running an up-to-date Raspberry Pi OS image. This is crucial for the system to automatically detect and integrate the Hailo-10H accelerator.
  2. Physical Installation: The AI HAT+ 2 ships with 16mm stacking headers, spacers, and screws. These are designed for easy installation on your Raspberry Pi 5, even when the active cooler is in place. Carefully align the HAT+ 2 with the GPIO pins and the PCIe FPC connector on your Raspberry Pi 5.
  3. Secure the HAT+ 2: Use the provided spacers and screws to securely fasten the AI HAT+ 2 to your Raspberry Pi 5. Ensure all connections are firm.
  4. Software Detection: Once powered on, the Raspberry Pi OS should automatically detect the Hailo-10H accelerator. It is fully integrated into Raspberry Pi’s camera software stack, including rpicam-apps camera applications, libcamera, and Picamera2.
The Raspberry Pi AI HAT+ 2 board installed on top of a Raspberry Pi 5.
Figure 2: Raspberry Pi AI HAT+ 2 installed on a Raspberry Pi 5.

3. Operating Instructions

The Raspberry Pi AI HAT+ 2 is designed for advanced AI inference tasks, particularly with generative AI models.

3.1. Accessing AI Models

  • Generative AI models compatible with the Hailo-10H accelerator can be found on the official Hailo website and on GitHub.
  • Look for Hailo-Ollama, an Ollama-compatible API written in C++ on top of HailoRT, which facilitates interaction with the AI accelerator.

3.2. Getting Started with Development

For detailed development guides and examples, it is highly recommended to consult the relevant documentation on the official Raspberry Pi website. This resource provides comprehensive information on integrating the AI HAT+ 2 into your projects and leveraging its capabilities.

4. Maintenance

The Raspberry Pi AI HAT+ 2 is a robust electronic component designed for long-term operation. Minimal maintenance is required:

  • Keep Clean: Ensure the board is free from dust and debris. Use a soft, dry brush or compressed air for cleaning.
  • Temperature Management: Operate the board within its specified temperature range (0°C to 50°C). If using an optional heatsink, ensure it is properly installed for optimal thermal performance.
  • Handle with Care: Avoid physical shock or excessive force when handling the board to prevent damage to components or connectors.

5. Troubleshooting

If you encounter issues during the use of your Raspberry Pi AI HAT+ 2, consider the following:

  • Software Detection Issues: Verify that your Raspberry Pi OS is fully updated. The system should automatically detect the Hailo-10H accelerator. If not, check the PCIe connection and ensure the HAT+ 2 is seated correctly.
  • Performance Problems: Ensure your power supply is adequate for both the Raspberry Pi 5 and the AI HAT+ 2. High-demand AI tasks may require stable power delivery.
  • General Inquiries: For specific questions regarding wiring diagrams, driver files, or technical documentation, please contact customer service. They can provide full technical support and guidance for your specific model.

6. Specifications

Feature Detail
AI Accelerator Hailo-10H
Inferencing Performance 40 TOPS (INT4)
On-board RAM 8GB dedicated
Computer Vision Performance Comparable to Raspberry Pi AI HAT+ (26 TOPS)
Host Interface PCIe Gen3 x1 FPC connector to Raspberry Pi 5
GPIO Header 40-pin (extends Pi's GPIO, no signal used by Hailo-10H)
Included Accessories 16mm stacking headers, spacers, screws
Optional Features Heatsink
Dimensions 64.1 x 56.7 x 5.5mm (Raspberry Pi HAT+ compatible)
Temperature Range 0°C to 50°C
Life Cycle In production until at least January 2036
Battery Included No
Typical Application Fields Education and Learning
High-concerned chemical None
Origin Mainland China

7. Warranty and Support

This product is a professional electronic module/development board. Due to the variety of application scenarios and wiring methods, a unified printed manual cannot be provided with the product package.

For any questions during use, or if you require wiring diagrams, driver files, or additional technical documentation, please feel free to contact our customer service. We are committed to providing full technical support and guidance tailored to your specific model.

An additional user manual in PDF format is available for download: Raspberry Pi AI HAT+ 2 User Manual (PDF).

8. User Tips

  • Always ensure your Raspberry Pi OS is up-to-date to benefit from the latest drivers and software integrations for the AI HAT+ 2.
  • When experimenting with new AI models, start with smaller models to understand the performance characteristics before scaling up to larger, more complex generative AI tasks.
  • Consider joining online Raspberry Pi and AI development communities for shared knowledge and troubleshooting tips.

Ask a question about this manual

Ask about setup, troubleshooting, compatibility, parts, safety, or missing instructions. Manuals+ will review the question and use this page’s manual context to help answer it.

After a successful submission, we securely hash the supplied account or form email before sending it to Microsoft Advertising for conversion attribution. Privacy details.