Pre-annotations not working for video object tracking - 'default.yaml' error

**Describe the bug**
I am trying to use YOLO as a backend model to pre-annotate video frames with a video object tracking template.
I run label studio as a container with the official image and the backend as another container following this [official guide](https://labelstud.io/blog/use-yolo26-with-label-studio-for-fast-bounding-box-pre-annotations/).
Please note that I managed to connect YOLO successfully and get pre-annotations with an image classification task. It just does not work with a video object tracking task.
I get the following error:
'IterableSimpleNamespace' object has no attribute 'model'. This may be caused by a modified or out of date ultralytics 'default.yaml' file.
Here is a docx file containing the full traceback of the error: [logs.docx](https://github.com/user-attachments/files/29461191/logs.docx)
I tried many things, guided by Claude. I tried downgrading ultralytics from 8.4.1 to 8.2.0 in the requirements.txt, also upgrading to 8.4.79 (latest one). I tried to change the default.yaml file as suggested in the logs. Nothing worked.

Here is the template I use:
```

```
The section on video object tracking in the [official doc](https://github.com/HumanSignal/label-studio-ml-backend/tree/master/label_studio_ml/examples/yolo#supported-features) is very confusing as the template in the short video is different from the example written below on this same page. I am not sure the one I use is correct.

**To Reproduce**
Steps to reproduce the behavior:
1. Deploy label studio.
2. Deploy the backend following this [doc](https://labelstud.io/blog/use-yolo26-with-label-studio-for-fast-bounding-box-pre-annotations/).
3. Create a new project with the Video Object Tracking task.
4. Sync at least one video.
5. Go to the project settings and connect the backend model in "Model".
6. Go back to the project and click on a video task. No pre-annotations are displayed.
7. Check the error returned in the logs of the backend container.

**Expected behavior**
I expect annotations to be suggested whenever I click on a video to annotate.

**Screenshots**

**Environment (please complete the following information):**
- OS: MacOS Tahoe 26.5
- Label Studio Version: v1.23.0

Start with the full traceback in logs.docx and compare the backend's requirements.txt and ultralytics default.yaml against the official YOLO video object tracking documentation. Verify whether the supplied Video, VideoRectangle, and Labels template is supported and whether the backend can produce pre-annotations for a video task; done means the reproduced container setup displays suggestions without the model attribute error.

Written by the indexing model from the issue text.