Sam in osx c++ mps mode.

[bemoregt]

https://youtu.be/m--nSlcjUBU?si=UXsuH7hv2nd_kOGf

Sam in c++, qt gui, Mac osx, mps mode.

[ryouchinsa]

Thanks for your information.
You might mean that LibTorch is better than ONNX Runtime to run SAM models in C++.

1. Loading an ONNX Model with LibTorch (C++ Deployment):
   To deploy SAM in a C++ environment using LibTorch, you would typically convert the PyTorch SAM model to ONNX format first.
   Export to ONNX: In Python, export the trained SAM model to ONNX using torch.onnx.export.
   Load in C++: In your C++ application, use an ONNX Runtime C++ API to load and run the ONNX model. LibTorch itself is primarily for TorchScript models, not directly for ONNX. You would use ONNX Runtime in conjunction with LibTorch if you also have other parts of your pipeline in C++ that leverage LibTorch.
2. Loading a Torch JIT Model with LibTorch (C++ Deployment):
   If you want to use Torch JIT for C++ deployment, you would serialize the SAM model to TorchScript format.
   Trace/Script the Model: In Python, use torch.jit.trace or torch.jit.script to convert your SAM model into a TorchScript ScriptModule.
   Save the ScriptModule: Save the ScriptModule to a file.
   Load in C++: In your C++ application, use torch::jit::load() from the LibTorch library to load the serialized TorchScript model. You can then execute the model using its forward method.