ExecuTorch iOS Demo App Tutorial

Welcome to the tutorial on setting up the ExecuTorch iOS Demo App!

This app uses the MobileNet v3 model to process live camera images leveraging three different backends: XNNPACK, Core ML and Metal Performance Shaders (MPS).

Prerequisites

Before we start, make sure you have the following tools installed:

1. Xcode 15 and Command Line Tools

Install Xcode 15 from the Mac App Store and then install the Command Line Tools using the terminal:

xcode-select --install

2. Python 3.10+

Python 3.10 or above, along with pip, should be pre-installed on MacOS 13.5+. If needed, download Python and install it. Verify the Python and pip versions using these commands:

which python3 pip
python3 --version
pip --version

3. Getting Started Tutorial

Before proceeding, follow the Setting Up ExecuTorch tutorial to configure the basic environment.

4. Backend Dependencies

Also, follow the corresponding sections from Core ML and MPS tutorials to install additional dependencies for those backends. Do not build anything just yet!

Models and Labels

Now let's move on to exporting and bundling the MobileNet v3 model.

1. Export Model

Export the MobileNet v3 model with CoreML, MPS and XNNPACK delegates, and move the exported model to a specific location where the Demo App will pick them up:

python3 -m examples.portable.scripts.export --model_name="mv3"
python3 -m examples.xnnpack.aot_compiler --delegate --model_name="mv3"
python3 -m examples.apple.coreml.scripts.export_and_delegate --model_name="mv3"
python3 -m examples.apple.mps.scripts.mps_example --model_name="mv3"

mkdir -p examples/demo-apps/apple_ios/ExecuTorchDemo/ExecuTorchDemo/Resources/Models/MobileNet/
mv mv3*.pte examples/demo-apps/apple_ios/ExecuTorchDemo/ExecuTorchDemo/Resources/Models/MobileNet/

2. Download Labels

Download the MobileNet model labels required for image classification:

curl https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt \
  -o examples/demo-apps/apple_ios/ExecuTorchDemo/ExecuTorchDemo/Resources/Models/MobileNet/imagenet_classes.txt

Build Runtime and Backends

Next, we will build the necessary frameworks for ExecuTorch and move them over for app linking.

1. Build Frameworks

./build/build_apple_frameworks.sh --Release --coreml --mps --xnnpack

2. Move Frameworks for App Linking

Make sure to have all the .xcframework bundles generated at the previous step at a specific location where the Demo App will pick them up:

mv cmake-out examples/demo-apps/apple_ios/ExecuTorchDemo/ExecuTorchDemo/Frameworks

Final Steps

We're almost done! Now, we just need to open the project in Xcode, run the tests, and finally run the app.

1. Open Project in Xcode

Double-click on the project file under examples/demo-apps/apple_ios/ExecuTorchDemo or run the command:

open examples/demo-apps/apple_ios/ExecuTorchDemo/ExecuTorchDemo.xcodeproj

2. Run Tests

You can run tests on Simulaltor directly in Xcode with Cmd + U or use the command line:

xcrun simctl create executorch "iPhone 15"
xcodebuild clean test \
     -project examples/demo-apps/apple_ios/ExecuTorchDemo/ExecuTorchDemo.xcodeproj \
     -scheme App \
     -destination name=executorch
xcrun simctl delete executorch

3. Run App

Finally, set up Code Signing in Xcode, and then run the app using Cmd + R.

Congratulations! You've successfully set up the ExecuTorch iOS Demo App. Now, you can explore and enjoy the power of ExecuTorch on your iOS device!