Any plane to support GPU acceleration using TensorRT?

[AgentSMLA]

as title, i love this project

[guyzoler]

Download the source and edit, if you have properly installed ONNX with tensorrt just edit line 38 at YoloCore.cs
change: ? new InferenceSession(onnxModel, SessionOptions.MakeSessionOptionWithCudaProvider(gpuId))
to: ? new InferenceSession(onnxModel, SessionOptions.MakeSessionOptionWithTensorrtProvider(gpuId))

I have it running, but it has the same performance for some reason, and it takes longer to initialize.

[snjs000111]

do i need to convert yolo8pose.onnx to yolo8pose.engine by tool trtexec ?
when i use tensorrt program abort at this line but why?
// Run inference on the OrtIoBinding and bind allocated GPU-memory.
session.RunWithBinding(runOptions, ortIoBinding);

[guyzoler]

You will need to convert your model, I followed this example:
https://docs.ultralytics.com/integrations/tensorrt/#usage

[Eve-88]

There is a error when using "SessionOptions.MakeSessionOptionWithTensorrtProvider(gpuId)"
onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 "" when trying to load "path\onnxruntime_providers_tensorrt.dll"

my packeage：
YoloDotNet v2.2
cuda 12.4
cudnn v9.1
tensorrt 8.6.1.6
onnxruntime_gpu 1.20.1
The environment variable configuration was successful, and the test went without any issues.
how I can solve this problem