ONNX model from YOLOv9s fails in Unity Sentis with "IndexError: axis 3 is out of bounds"

ONNX model from YOLOv9s fails in Unity Sentis with “IndexError: axis 3 is out of bounds”

Body:

## 🐛 Bug Report

### Description

I'm encountering an issue running an ONNX model exported from a YOLOv9s `.pt` file in Unity Sentis (using HOLO DNN packages).

- The original `.pt` model works as expected in local PyTorch inference.
- After exporting to ONNX (opset 15), the model fails to run in Unity using Sentis.

### Error Output

Exception: [error] can not add post process to model for some reason.
(IndexError: axis 3 is out of bounds shape of rank, 3
Assertion failure. Value was False
Expected: True)

### Steps to Reproduce

1. Train a YOLOv9s model with custom dataset.
2. Export using: `yolo export format=onnx opset=15`
3. Import into Unity via Sentis with HOLO DNN.
4. Observe runtime failure during post-process attachment.

### Expected Behavior

Model should run in Sentis just like it does in local PyTorch inference.

### Environment

- Model: YOLOv9s
- Format: ONNX (opset 15)
- Unity Version: 2023.x
- Sentis: `com.unity.sentis` 1.3.0-pre (HOLO DNN)
- Platform: Windows
- Input Shape: `(1, 3, 448, 640)`
- Output Shape: `(1, 7, 7581)` from custom dataset

### Additional Notes

Other Ultralytics models with default output shapes like `(1, 84, 8400)` run fine in Sentis. This might be an incompatibility with the custom output shape and post-processing logic in Sentis.

Looking for guidance on adapting this model for compatibility or configuring post-process steps manually if needed.

I’m not much of an AI guy, but I think you should have posted this in the AI section and not in the “News and General Discussions” section