Help needed: Converting .onnx to .hef for yolov8 for Raspberry Pi with AI Hat +

user805 · March 10, 2026, 8:23am

Hi, I’m working on a raspberry pi AI project. I did transfer learning on python to train a yolov8 model and have the .onnx file, however converting this to .hef has been incredibly difficult. I initially managed to get the .hef file in the 15h architecture with my laptop, however my pi said it only supports 8architecture. I’ve tried severally to run the Hailo AI software suite - docker on Ubuntu but I’ve been running into a lot of issues, and my laptop only has 16GB RAM and so it doesn’t really meet the RAM requirement. if I provide the .onnx and .yml , would you be willing to help me convert to a .hef file. I’ve been trying to work on this for over a week but it’s getting incredibly frustrating and it seems like I’m getting nowhere

Michael · March 10, 2026, 8:38am

Hi,

We have an updated and detailed guide here:

github.com/hailo-ai/hailo-apps

doc/developer_guide/retraining_example.md

main

# YOLOv8 Retraining Example

This guide demonstrates how to retrain a YOLOv8 model for barcode detection using the Kaggle barcode-detector dataset. After training, we'll convert the model to HEF format and deploy it on the Raspberry Pi 5 AI Kit (or any x86 platform with a Hailo accelerator).

#### Required compute resources:
- **Training phase**: Any cloud service with GPU resources (including Python notebook-based services like Google Colab)
- **Hailo compilation phase**: Conversion to HEF format is compute-intensive but manageable on standard PCs (e.g., overnight task)

For complete reference implementations, see the Jupyter notebooks (please note some of the stages, including installations, might take time): 
- [`retraining.ipynb`](retraining.ipynb)
- [`compilation.ipynb`](compilation.ipynb)

## Training Setup

### Set up environment

First, create a Python virtual environment:

```bash
python -m venv env

This file has been truncated. show original

Please let me know if that was helpful.

Thanks,

user805 · March 10, 2026, 2:53pm

Hello, thank you for your response. please which .whl files should i download. the 8/8L section does not have a dataflow compiler

Michael · March 10, 2026, 8:44pm

Hi @user805 ,

In Hailo Developer Zone there is a filter on top for Hailo-8.
Current DFC version is 3.33

Thanks,

user805 · March 10, 2026, 9:02pm

theres nothing there for me

user805 · March 10, 2026, 9:03pm

user805 · March 13, 2026, 6:31am

Hi I was able to manage to get it, but everytime i try to run the hailomzz compile step it gives me this error:

File “/usr/local/lib/python3.12/dist-packages/hailo_sdk_client/sdk_backend/script_parser/nms_postprocess_command.py”, line 629, in _get_output_preds
raise AllocatorScriptParserException(
hailo_sdk_client.sdk_backend.sdk_backend_exceptions.AllocatorScriptParserException: Error in the last layers of the model, expected conv but found LayerType.activation layer.

Michael · March 13, 2026, 7:24pm

Hi @user805 ,

This error means the compiler expects the final layers before NMS to be convolution layer, but your model has an activation layer (e.g., SiLU/ReLU) as the last layer instead.

Thanks,

user805 · March 13, 2026, 8:17pm

Hi,

Is there anything I can do to resolve this