ios

ONNX Runtime Mobile Question Answering iOS sample application with Ort-Extensions support for pre/post processing

Overview

This is a basic Question Answering example application for ONNX Runtime on iOS with Ort-Extensions support for pre/post processing. The demo app accomplishes the task of basic QA, i.e. gives answers based on user provided context.

The model used here is from source: HuggingFace and accomodated into ONNX version with pre/post processing support.

Requirements

• Install Xcode 13.0 and above (preferably latest version)
• A valid Apple Developer ID
• An iOS device or iOS simulator
• Xcode command line tools xcode-select --install
• Clone the onnxruntime-inference-examples source code repo

Build And Run

1. Install CocoaPods. sudo gem install cocoapods

2. In terminal, run pod install under <onnxruntime-inference-example-root>/mobile/examples/question_answering/ios/ to generate the workspace file and install required pod files.

   Note: At the end of this step, you should get a file called ORTQuestionAnswering.xcworkspace.

3. Prepare the required model used in this sample.

   Use provided prepare_model.py script under mobile/examples/question_answering/android/ to do the model-preparing work. The whole procedure includes four steps.

   1. download model from huggingface.
   2. convert model to onnx.
   3. quantize onnx model.
   4. add pre/post processing to onnx model.

   Note: See more information in mobile/examples/question_answering/android/README.md.

   cd mobile/examples/question_answering/android
   bash prepare_models.sh

   Note: Currently the model is already provided under mobile/examples/question_answering/ios/ORTQuestionAnswering/ORTQuestionAnswering/csarron_mobilebert_uncased_squad_v2_quant_with_pre_post_processing.onnx

4. Open <ONNXRuntime-inference-example-root>/mobile/examples/question_answering/ios/ORTQuestionAnswering.xcworkspace in Xcode and make sure to select your corresponding development team under Target-General-Signing for a proper codesign procedure to run the app (only on device required, if running on iOS simulator can skip this step.)

5. Connect your iOS device/simulator, build and run the app. Click Get Answers! button to see performed answer based on the given context. A default hint message will be displayed on initial screen.

Here's an example screenshot of the app: