We plan to use Visual Studio 2019 to load a model that has already been trained in PyTorch with Script conversion into C++ with CUDA 11.3 and LibTorch 1.11 for class classification. Input image is a black and white image, input size is 224x224. Therefore, the tensor size is {1,1,224,224}.
Here is the current code, and as an example, we are currently inputting a sample tensor with no image input.
#include <torch/torch.h>
#include <iostream>
#include <torch/script.h>
#include <opencv2/opencv.hpp>
int main() {
// Load the model
std::string model_path = "model_weights_c++.pt";
torch::jit::script::Module module;
try {
module = torch::jit::load(model_path);
}
catch (const c10::Error& e) {
std::cerr << "Error loading the model: " << e.what() << std::endl;
return -1;
}
// Set the device to GPU if available
torch::Device device(torch::kCPU);
if (torch::cuda::is_available()) {
device = torch::Device(torch::kCUDA); }
}
// Create a vector of inputs.
std::vector<torch::jit::IValue> inputs;
inputs.push_back(torch::ones({ 1, 1, 224, 224 }));
module.to(device);
// Execute the model and turn its output into a tensor.
std::vector<torch::jit::IValue> outputs;
at::Tensor output = module.forward(inputs).toTensor().to(device);
std::cout << output.slice(/*dim=*/1, /*start=*/0, /*end=*/5) << '\n';
return 0;
}
The build succeeds, but while executing module.forward(), the following program stops the effect with an unhandled exception. [return stack.front()].
What is the reason? Please help.
at::IValue operator()(
Stack stack, kwargs& kwargs& kwargs
const Kwargs& kwargs = Kwargs()) {
getSchema().checkAndNormalizeInputs(stack, kwargs);
run(stack);
return stack.front(); } <-------------here
}
The call history is shown in the following image.
