How to use libtorch trace model

1

I installed libtorch1.5 and try to use it train model, I want use C++ API to trace the trained model and save it. But I just find the API torch::jit::tracer::trace, how to use it? I can’t get any explain about these API, who can give me some advice?

1 Like
2 
# An instance of your model.
model = torch.load('my_model.pth')
model.eval()
# An example input you would normally provide to your model's forward() method.
example = torch.rand(1, 3, 224, 224)
 
# Use torch.jit.trace to generate a torch.jit.ScriptModule via tracing.
traced_script_module = torch.jit.trace(model.cuda(), example.cuda())

# save it
traced_script_module.save("model_cpp.pt")
4

First of all, thank you for your answer, I actually ues libtorch(c++ API), so I want to train and save finally model in the same platform, but I can’t find any instructions about how to trace or scrpt model in C++ platform. You may not get my problems.