Getting irrelevant content on IOS. App has been run on device successfully, but giving wrong response for each frame. FYI, we are using iPhone 15 Pro Max.

[Deswaluday]

[shoumikhin]

@Deswaluday which model are you using? How do you export it? Which image/prompt used?

[Deswaluday]

We are using llava-1.5-7b-hf and we have followed the below command to export the files for model.

python -m executorch.examples.models.llava.export_llava --pte-name llava.pte --with-artifacts

Below is the screenshot of loaded Llava model and tokenizer files:

Here is the screen shot of an image with irrelevant content :

Here is the screenshot with content where model did not found the image

We have followed all the steps from the documentation given by you.