This is a simple C++ demo application that uses the ExecuTorch library and KissFFT for SPEECHCOMMANDS CNN model inference.

0. Export the model (model.ipynb) or use a PTE model already provided.

1. The ExecuTorch repository and KissFFT repository need to be downloaded to third-party directory.

    mkdir third-party
    cd third-party
    git clone https://github.com/pytorch/executorch.git
    git clone https://github.com/mborgerding/kissfft.git

   Tested versions: 39c9781 for kissfft and 8c84780 for executorch.

2. Install dev requirements for ExecuTorch (venv advised):

   pip install -r requirements.txt

3. Build the project (python source must have executorch installed):

   ./build.sh

4. Run the demo application:

   ./build/bin/speechcommands_app cat_test.wav

• CMake 3.29 or higher
• C++17 compatible compiler
• ExecuTorch library (release/1.0)
• KissFFT library

• Make sure you have the correct model file (.pte) compatible with ExecuTorch.
• This app can be tested with wav files (from SPEECHCOMMANDS etc.) - it uses first second of the recording (must be sorted wav) to infer the word.