Erroe while converting onnx model to hef

General
1

Sir , i have install hailo dataflow compiler successfully and create a custom model using roboflow yolov8 and conver model to onnx using google colab but in converrtion process when i use command hailomz compile yolov8s --ckpt= best.onnx --hw-arch hailo8l --calib-path train/images --classes 4–performance i have error youir model ir_version 10 is higher than the checkers(9) please any one suggest suggestion it argent thank

2

Hi @iotpbtrap ,

In colab I used this code to change the ir_version of my onnx file:

# Export the model in ONNX format IR_version 9 (adjust IR version if needed)

import onnx

# Load the ONNX model
model = onnx.load("/content/runs/detect/train/weights/best.onnx")

# Change the IR version to 9
model.ir_version = 9

# Save the modified model
onnx.save(model, "/content/runs/detect/train/weights/best9.onnx")

print("Model IR version changed to 9 and saved as best9.onnx")

maybe it also helps for your model.

3

Thank you so much.

Before trying this solution,

I just tried to cover the onnx with opset==9, but it didn’t work.

But, this solution works for me. it maybe ir_version and opset has different meaning…

1 Like
4

Hi @happistday ,

Thanks for pointing this out!

I did a quick search on the official ONNX docs, and here’s a link that explains both ir_version and opset in detail: ONNX Versioning - ONNX 1.18.0 documentation

2 Likes