How to split the libtorch forward method's output?

donkey007

hello, im new to this… i need help

this is my c++ code

// [0]Define the input tensor
torch::Tensor veh_input2 = torch::tensor({
{{5.006278301940284, 0.0786753188073942, 4.411626434845282, 0.09598564608132709, 4.52527292334112, 0.1367866935617143},
{4.344146508992793, 0.034440922157513695, 4.472077232158242, 0.04202762418462361, 4.399348505369012, 0.05233431582289311},
{5.800142409494782, 0.11272599320992382, 5.238293982154694, 0.15335229366842879, 4.269588880340237, 0.22455373622631702},
{4.18455135464626, 0.03927927361417172, 4.165869740437117, 0.045566222570332576, 4.237983947536861, 0.055740630185644596}}
});

// [1]Run the model
auto output = veh_model.forward({veh_input2});

std::cout << "==========[1]============";
std::cout << typeid(output).name() << std::endl;
std::cout << output << std::endl;

============ program die here !! ===================

// [2]Extract the outputs
torch::Tensor mu_veh = output.toTuple()->elements()[0].toTensor();
torch::Tensor sigma_veh = output.toTuple()->elements()[1].toTensor();

@@@@@@@@@@@@ output @@@@@@@@@@@@@@@@

and this is my output

==========[1]============
typeid(output).name() ==> struct c10::IValue

[(1,.,.) =
5.3061 15.4691
4.2534 15.5291
5.9763 13.6928
4.2134 15.4633
[ CPUFloatType{1,4,2} ], (1,.,.) =
2.5224 3.6053
2.0312 4.4862
2.6079 4.7206
1.9234 4.2305
[ CPUFloatType{1,4,2} ]]

i want to access each value of tensor vector like “5.3061”
why this c++ code didnt work?

// [2]Extract the outputs
torch::Tensor mu_veh = output.toTuple()->elements()[0].toTensor();
torch::Tensor sigma_veh = output.toTuple()->elements()[1].toTensor();
if you know, let me know thanks

JonB

@donkey007 said in How to split the libtorch forward method's output?:

why this c++ code didnt work?

What "doesn't work"? How should anyone in a Qt forum know what "the libtorch forward method's output" produces from

auto output = veh_model.forward({veh_input2});
std::cout << output << std::endl;

?

donkey007

@JonB
thanks for your answer!

[(1,.,.) =
5.3061 15.4691
4.2534 15.5291
5.9763 13.6928
4.2134 15.4633
[ CPUFloatType{1,4,2} ], (1,.,.) =
2.5224 3.6053
2.0312 4.4862
2.6079 4.7206
1.9234 4.2305
[ CPUFloatType{1,4,2} ]]

this is my " std::cout << output << std::endl; " output

and i try slicing output like this

torch::Tensor mu_veh = output.toTuple()->elements()[0].toTensor();
torch::Tensor sigma_veh = output.toTuple()->elements()[1].toTensor();

but this c++ code dosen't work...

do you know why?

JonB

@donkey007 No. Look at the torch::Tensor documentation.