The model slows down on GPU #2429

Environment

• Pythonnet version: 3.1.0-preview2024-07-03
• Python version:3.11.5
• Operating System:Windows 11
• .NET Runtime: .NET 8.0

Details

• I wrote a python script which run a yolov5 model on a RTX3090.When I run the model on GPU with python only,it takes only 0.02s to process one image.However,when I use C# to call the script ,it takes about 1.5s to process one image. This phenonmenon only happens on GPU,when I use CPU to run the model, the process time with python and C# is the same.
• Sorry,I am not a native speaker,so I am not good at English.

device = torch.device("cuda")
face_detector = torch.hub.load(yolo_path,"custom",path = yolo_model_path ,source = "local")
face_detector.to(device)
face_detector(image)

dynamic a = dd.drowsyDetect(res_array);