Invalid argument(s): Failed to lookup symbol 'load_ml_model': dlsym(RTLD_DEFAULT, load_ml_model): symbol not found

[tnghieu]

When using await PytorchLite.loadClassificationModel( "assets/models/best.pt", 640, 480, 9, labelPath: "assets/labels/labels.txt", );

[abdelaziz-mahdy]

What platform are you testing on,

Also did you close the app and rebuild it after adding the package?

[tnghieu]

iOS

Yes, I also tried cleaning and removing the podfile.

[abdelaziz-mahdy]

I will try to test, but can you provide an example repo? Or configuration since my integration tests work

[tnghieu]

https://github.com/tnghieu/sharkeye

[abdelaziz-mahdy]

Hello, thank you for the repo, sadly I am not available this week

[tnghieu]

Hi @zezo357, I've made updates and used the export to torchscript and I am still seeing the same error. Is this compatible with iOS?

[abdelaziz-mahdy]

Using my GitHub actions, yes it works on both iOS and Android

I will need more details, like the version you are using in pubspec

Flutter doctor -v

[tnghieu]

[✓] Flutter (Channel stable, 3.13.1, on macOS 13.5 22G74 darwin-arm64, locale en-US) [✓] Android toolchain - develop for Android devices (Android SDK version 34.0.0) [✓] Xcode - develop for iOS and macOS (Xcode 14.3.1) [✓] Chrome - develop for the web [✓] Android Studio (version 2022.3) [✓] VS Code (version 1.80.2) [✓] Connected device (3 available) [✓] Network resources

pytorch_lite: ^4.0.0

My repo has latest changes with the model weights.

[abdelaziz-mahdy]

Thank you, can you wait for next weekend? Since I can't try now as I mentioned

[tnghieu]

Yes I can wait. Thank you for your time.

[abdelaziz-mahdy]

been trying for a while the only solution i found is to add the cpp file in Compile Source in xcode

which is a very bad workaround since whenever i update the package you will have to update it

so i am thinking of going back to native (java and objc)

• Think this will fix the large size on both android and ios
• Having good speed since it will use native ways to handle images

i will try to maintain new interface and just change the back logic

sadly it may take sometime to completely migrate back and add the ios logic since i dont understand obj c or swift

sorry for the inconvenience and if i can help with something else let me know

btw loved the idea of your project simple and does the job very good

[tnghieu]

Thank you. I look forward to the updates.

[tnghieu]

Hi @zezo357 , I see you've made a new update/PR. Thank you for the work!

I do have a question - does this library support object detection models? I ask because google-ml-object-detection library in Flutter can only use custom image classification models.

[abdelaziz-mahdy]

Hi @zezo357 , I see you've made a new update/PR. Thank you for the work!

I do have a question - does this library support object detection models? I ask because google-ml-object-detection library in Flutter can only use custom image classification models.

yolov5 and yolov8 are supported yes

[abdelaziz-mahdy]

please check version https://pub.dev/packages/pytorch_lite/versions/4.1.0-dev

camera code is tested on android but not tested on ios hopefully it works (will need someone to test it and report its status)

[abdelaziz-mahdy]

Fixed in 4.2.0