Dynamic input shape model

Environment

FastDeploy version: 1.0.0
OS Platform: Windows x64
Hardware: CPU noavx
Program Language: C++
CMake: 3.18, 3.25
Visual Studio: 2019

Code

const std::string& model_file = "C:/modnet-resnet50_vd/model.pdmodel";
const std::string& params_file = "C:/modnet-resnet50_vd/model.pdiparams";
const std::string& config_file = "C:/modnet-resnet50_vd/deploy.yaml";

auto option = fastdeploy::RuntimeOption();
option.UsePaddleInferBackend();
option.UseCpu();
option.SetPaddleMKLDNN(false);
    
auto model = fastdeploy::vision::matting::PPMatting(model_file, params_file, config_file, option);

Problem description

console output:
1[WARNING] fastdeploy/vision/matting/ppmatting/ppmatting.cc(78)::fastdeploy::vision::matting::PPMatting::BuildPreprocessPipelineFromConfig
Detected dynamic input shape of your model, only Paddle Inference / OpenVINO support this model now.

I specified option.UsePaddleInferBackend();
Why is this model not supported. Should I use a different library?

Start with the warning reported at fastdeploy/vision/matting/ppmatting/ppmatting.cc:78 and review how the PPMatting model is constructed with RuntimeOption and UsePaddleInferBackend(). Reproduce the Windows x64 CPU case using the model and configuration shown; done means determining why the selected backend reports the dynamic input shape as unsupported or documenting the limitation.