In the Tensorflow Lite repo, I’m building the Tensorflow Lite binary optimized for specific models like so:

bash tensorflow/lite/ios/build_frameworks.sh \
  --input_models=model1.tflite,model2.tflite \
  --target_archs=arm64

This generates an optimized TensorFlowLiteC binary containing only the ops needed for model1.tflite, model2.tflite. Instead, I need it to build the TensorFlowLiteCMetal binary to get GPU accelerated inferencing.

The script and documentation do not mention TensorflowLiteCMetal.

How do I get the script to generate the TensorflowLiteCMetal binary instead?