hailomz parse fails with AttributeError: 'NoneType' object has no attribute 'save_har' on YOLOv8n ONNX (Model Zoo v2.14.0)

General
1

Hello Hailo team,

I am encountering a consistent error when running hailomz parse on a YOLOv8n ONNX model.
Although the parsing process appears to complete successfully, it always ends with a Python exception:

AttributeError: 'NoneType' object has no attribute 'save_har'

I would like to confirm whether this is a known issue or expected behavior in the current Model Zoo version, and whether hailomz parse is intended to be used standalone in this setup.

Command Used

hailomz parse yolov8n --ckpt ./best.onnx

:page_facing_up: Full Error Log (Relevant Part)

[info] NMS structure of yolov8 (or equivalent architecture) was detected.
[info] In order to use HailoRT post-processing capabilities, these end node names should be used:
  /model.22/cv2.0/cv2.0.2/Conv
  /model.22/cv3.0/cv3.0.2/Conv
  /model.22/cv2.1/cv2.1.2/Conv
  /model.22/cv3.1/cv3.1.2/Conv
  /model.22/cv2.2/cv2.2.2/Conv
  /model.22/cv3.2/cv3.2.2/Conv
[info] Translation completed on ONNX model yolov8n
[info] Appending model script commands to yolov8n from string
[info] Added nms postprocess command to model script.
Traceback (most recent call last):
  File ".../hailomz", line 8, in <module>
    sys.exit(main())
  ...
  File ".../main_utils.py", line 146, in parse_model
    runner.save_har(results_dir / f"{network_info.network.network_name}.har")
AttributeError: 'NoneType' object has no attribute 'save_har'

:white_check_mark: Observations

  • The ONNX model is valid and loads correctly with onnx.load()

  • YOLOv8 NMS structure is correctly detected

  • Start nodes and end nodes are correctly inferred

  • The error occurs only at the final save_har() step

  • Despite this error, running hailomz compile directly successfully produces a .hef file

This suggests that parsing itself succeeds, but the runner object becomes None before saving the .har.

:test_tube: Environment Information

Software Versions

Hailo Model Zoo:        v2.14.0
HailoRT:               v4.21.0
Hailo Dataflow Compiler: v3.31.0
Python:                3.10.12
OS:                    Ubuntu x86_64

Installation Method

pip install hailo_dataflow_compiler-3.31.0-py3-none-linux_x86_64.whl
pip install hailo_model_zoo-2.14.0-py3-none-any.whl
sudo dpkg -i hailort_4.21.0_amd64.deb
pip install hailort-4.21.0-cp310-cp310-linux_x86_64.whl

:red_question_mark: Questions

  1. Is this NoneType.save_har error in hailomz parse a known issue in Model Zoo v2.14.0?

  2. Is hailomz parse expected to work standalone for custom ONNX models, or is it recommended to skip parse/optimize and use hailomz compile directly?

  3. Is there a recommended way to generate a valid .har file for inspection/debugging when using custom YOLOv8 ONNX models?

Any clarification or guidance would be greatly appreciated.

Thank you for your support.

Additional note:

I would like to add that the same NoneType.save_har issue also occurs when using hailomz compile, not only hailomz parse.

Since hailomz compile internally invokes the same parsing flow (parse_model()), the failure appears to be shared between both commands. In my environment, both parse and compile reach successful YOLOv8 structure detection and NMS inference, but fail when attempting to save the intermediate .har due to the runner object being None.

This suggests the issue is not limited to standalone parsing, but affects the full compile pipeline as well.