发现 Unstructured 的 “TensorrtExecutionProvider“ 比 “CUDAExecutionProvider“ 慢

修改 unstructured_inference/models/detectron2onnx.py 和 unstructured_inference/models/detectron2onnx.py 的代码，优先使用 “CUDAExecutionProvider”。

        # ordered_providers = [#     "TensorrtExecutionProvider",#     "CUDAExecutionProvider",#     "CPUExecutionProvider",# ]ordered_providers = ["CUDAExecutionProvider","TensorrtExecutionProvider","CPUExecutionProvider",]

完结！