Unable to create interpreter while using custom yolov11n model

[micktsai]

Search before asking

• I checked the docs, issues, and discussions and could not find an answer.

Question

I encounter Invalid argument(s): Unable to create interpreter. When I'm trying to use my customize YOLO v11n model(add P2 head and remove P5 head). Model type is tflite.
I had tried v8n model with same settings (add P2 head and remove P5 head) and it works normally.
Is there any setting wrong at export model? or the current version can't support customize YOLOv11n model?
Below is how I export my model.

model.export(
    format="tflite",
    half=True,
    imgsz=1280
)

Additional

No response

[UltralyticsAssistant]

👋 Hello @micktsai, thank you for submitting a ultralytics/yolo-flutter-app 🚀 Issue. This is an automated response to help us triage quickly—an Ultralytics engineer will also assist soon 🛠️🙂. To help us address your concern efficiently, please ensure you've provided the following information:

1. For bug reports:

   • A clear and concise description of the bug
   • A minimum reproducible example MRE that demonstrates the issue (especially important here since this involves a custom YOLO11n architecture + TFLite export)
   • Your environment details (OS, Python version, package versions)
   • Expected behavior vs. actual behavior
   • Any error messages or logs related to the issue (please include the full Flutter console output and any stack trace around “Unable to create interpreter”)

2. For feature requests:

   • A clear and concise description of the proposed feature
   • The problem this feature would solve
   • Any alternative solutions you've considered

3. For questions:

   • Provide as much context as possible about your question
   • Include any research you've already done on the topic
   • Specify which parts of the documentation, if any, you've already consulted

Please make sure you've searched existing issues to avoid duplicates. If you need to add any additional information, please comment on this issue.

Thank you for your contribution to improving our project! 🙏🚀

[pderrenger]

Custom Ultralytics YOLO11n topology changes (like adding a P2 head/removing P5) aren’t guaranteed to produce a valid/compatible TFLite graph, so please first upgrade to the latest ultralytics, re-export without half=True (and ideally with a standard imgsz=640), then verify the .tflite loads in Python via YOLO("your_model.tflite"); if it loads in Python but Flutter still fails at interpreter creation, it’s a TFLite runtime/op support issue—share the full export log plus the result of yolo predict model=your_model.tflite source=... so we can identify the unsupported op (see the Ultralytics TFLite export guide for the expected workflow).

[asabri97]

@micktsai Thanks for reporting this. The ‘Unable to create interpreter’ message usually means the TFLite runtime can’t allocate or map the model.
A few quick checks:

1. try loading with useGpu: false to rule out GPU delegate issues
2. re-export without fused NMS and without fp16, and at a smaller resolution first: model.export(format='tflite', nms=False, half=False, imgsz=640) confirm 640 works before 1280
3. sanity-test the TFLite locally in Python (YOLO('model.tflite')(source='test.jpg')) to ensure the file itself is valid
4. please share the exact logcat/flutter run -v error line, device/OS, and whether the model has P2 only or P2–P4 heads. With that info, we can nail down whether it’s a delegate/output-shape or memory issue and suggest the right fix.

[micktsai]

Custom Ultralytics YOLO11n topology changes (like adding a P2 head/removing P5) aren’t guaranteed to produce a valid/compatible TFLite graph, so please first upgrade to the latest ultralytics, re-export without half=True (and ideally with a standard imgsz=640), then verify the .tflite loads in Python via YOLO("your_model.tflite"); if it loads in Python but Flutter still fails at interpreter creation, it’s a TFLite runtime/op support issue—share the full export log plus the result of yolo predict model=your_model.tflite source=... so we can identify the unsupported op (see the Ultralytics TFLite export guide for the expected workflow).

I had tried loading in Python before loading in flutter. And it works normally.

[micktsai]

@micktsai Thanks for reporting this. The ‘Unable to create interpreter’ message usually means the TFLite runtime can’t allocate or map the model. A few quick checks:

1. try loading with useGpu: false to rule out GPU delegate issues
2. re-export without fused NMS and without fp16, and at a smaller resolution first: model.export(format='tflite', nms=False, half=False, imgsz=640) confirm 640 works before 1280
3. sanity-test the TFLite locally in Python (YOLO('model.tflite')(source='test.jpg')) to ensure the file itself is valid
4. please share the exact logcat/flutter run -v error line, device/OS, and whether the model has P2 only or P2–P4 heads. With that info, we can nail down whether it’s a delegate/output-shape or memory issue and suggest the right fix.

Thanks for your reply,
I have tried 1, 2 before posting this issue but it still fail, and my model works normally in Python no matter imgsz is 640 or 1280
I will try to use tflite model later for detail error message, so far, I use coreML model to bypass this issue

[pderrenger]

If the model loads/runs in Python but tflite_flutter can’t create the interpreter, that’s almost always a mobile TFLite runtime limitation (unsupported op or tensor-allocation failure) triggered by the custom YOLO11n head, so when you get a chance please paste the exact failing line from flutter run -v / logcat (or iOS console) plus your device/OS and tflite_flutter version; also confirm whether a stock exported yolo11n_float32.tflite runs in the same app so we can isolate it to the custom topology (export steps/expected artifacts are in the Ultralytics TFLite export guide for details).

[asabri97]

@micktsai any updates here? Could you please share the logs if the problem still occurs?

[asabri97]

@micktsai Please share the logs, I need the exact error from the app for me to be able to help; run with TFLite again and capture full error line you can run flutter run -v and you can filter for adb logcat filtered for ObjectDetector, YOLO or tflite right when you load the model so we see the native stack trace

[github-actions]

👋 Hello there! We wanted to give you a friendly reminder that this issue has not had any recent activity and may be closed soon, but don't worry - you can always reopen it if needed. If you still have any questions or concerns, please feel free to let us know how we can help.

For additional resources and information, please see the links below:

• Docs: https://docs.ultralytics.com
• HUB: https://hub.ultralytics.com
• Community: https://community.ultralytics.com

Feel free to inform us of any other issues you discover or feature requests that come to mind in the future. Pull Requests (PRs) are also always welcomed!

Thank you for your contributions to YOLO 🚀 and Vision AI ⭐