How can i get access to first and second Tensor from Tuple, returned from forward() method in Libtorch C\C++

jit
1

Hello Team, it is my code.
I’m using Jit generated model with Libtorch in C\C++ in Ubuntu 18.04, with CUDA 10.2

I get correct result form my first call result = mE.forward(x);

    torch::Tensor tensor_image = ...
    torch::IValue result;   // it is standard return of forward() in JIT traced model

    std::vector<torch::jit::IValue> x;
    x.push_back(tensor_image);
    result  = mE.forward(x);
    std::cout << sp_and_gl_output_CUDA << std::endl;
    std::vector<torch::IValue> in_to_G;
    in_to_G.push_back(result);
    img_tensor = mG.forward(in_to_G);

On the image print of result variable, returned by forward() method.

std::cout << sp_and_gl_output_CUDA << std::endl;

The result contained, two Tensors, but i can’t to pass it in next forward(), in right form.

изображение
изображение1757×1174 164 KB

The result of execution:

mG.forward load spatial and global:

forward() Expected a value of type 'Tensor' for argument 'v' but instead found type 'Tuple[Tensor, Tensor]'.
Position: 1
Declaration: forward(__torch__.models.networks.generator.StyleGAN2ResnetGenerator self, Tensor v, Tensor v0) -> (Tensor)
Exception raised from checkArg at ../aten/src/ATen/core/function_schema_inl.h:184 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x6b (0x7f1c5d0290db in /home/interceptor/Документы/Git_Medium_repo/Binary_search_engine_CUDA/torch_to_cpp/libtorch/lib/libc10.so)
frame #1: c10::detail::torchCheckFail(char const*, char const*, unsigned int, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) + 0xce (0x7f1c5d024d2e in /home/interceptor/Документы/Git_Medium_repo/Binary_search_engine_CUDA/torch_to_cpp/libtorch/lib/libc10.so)
frame #2: <unknown function> + 0x128d542 (0x7f1c473af542 in /home/interceptor/Документы/Git_Medium_repo/Binary_search_engine_CUDA/torch_to_cpp/libtorch/lib/libtorch_cpu.so)
frame #3: <unknown function> + 0x12918c1 (0x7f1c473b38c1 in /home/interceptor/Документы/Git_Medium_repo/Binary_search_engine_CUDA/torch_to_cpp/libtorch/lib/libtorch_cpu.so)
frame #4: torch::jit::GraphFunction::operator()(std::vector<c10::IValue, std::allocator<c10::IValue> >, std::unordered_map<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, c10::IValue, std::hash<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::equal_to<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::allocator<std::pair<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const, c10::IValue> > > const&) + 0x2d (0x7f1c49c97b2d in /home/interceptor/Документы/Git_Medium_repo/Binary_search_engine_CUDA/torch_to_cpp/libtorch/lib/libtorch_cpu.so)
frame #5: torch::jit::Method::operator()(std::vector<c10::IValue, std::allocator<c10::IValue> >, std::unordered_map<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, c10::IValue, std::hash<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::equal_to<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::allocator<std::pair<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const, c10::IValue> > > const&) const + 0x175 (0x7f1c49ca7235 in /home/interceptor/Документы/Git_Medium_repo/Binary_search_engine_CUDA/torch_to_cpp/libtorch/lib/libtorch_cpu.so)
frame #6: <unknown function> + 0x6266 (0x564143038266 in ./particle_cuda_test)
frame #7: <unknown function> + 0x47af (0x5641430367af in ./particle_cuda_test)
frame #8: __libc_start_main + 0xe7 (0x7f1bf9682bf7 in /lib/x86_64-linux-gnu/libc.so.6)
frame #9: <unknown function> + 0x4daa (0x564143036daa in ./particle_cuda_test)

Main module:  torch_and_cuda_call.cpp  completed. 
ok!
1 Like
2

This placeholder for second call:

   in_to_G.push_back(torch::ones({1, 8,8,8}).to(at::kCUDA));
   in_to_G.push_back(torch::ones({1, 2048}).to(at::kCUDA));
   img_tensor = mG.forward(in_to_G);

worked!
How can i get two Tensors for create new input vector from it?
i need to parse this result, to get Tensor in it:
torch::IValue result;

There are somethink like this issues on the stack and github:

  1. c++ - How to get values return by Tuple Object in Maskcrnn libtorch - Stack Overflow
  2. C++ Convert c10::IValue to tensor or tuple or to TensorList failed · Issue #48691 · pytorch/pytorch · GitHub
3

Seems it WORKED

    auto output1_sp = sp_and_gl_output_CUDA.toTuple()->elements()[0].toTensor();
    auto output2_gl = sp_and_gl_output_CUDA.toTuple()->elements()[1].toTensor();
    //-------------------------------------------------------------------------------------------------------------------
    in_to_G.push_back(output1_sp.to(at::kCUDA));
    in_to_G.push_back(output2_gl.to(at::kCUDA));
    //-------------------------------------------------------------------------------------------------------------------
    img_tensor = mG.forward(in_to_G);
1 Like