coreReadArchive.h (271) - Serialization Error in readExternal: 0 (Type mismatch)

./yolox ../model_trt.engine -i ../../../../assets/dog.jpg

[05/13/2022-15:07:49] [E] [TRT] coreReadArchive.h (271) - Serialization Error in readExternal: 0 (Type mismatch)
[05/13/2022-15:07:49] [E] [TRT] INVALID_STATE: std::exception
[05/13/2022-15:07:49] [E] [TRT] INVALID_CONFIG: Deserialize the cuda engine failed.
Segmentation fault (core dumped)

运行依赖库
ldd yolox

linux-vdso.so.1 (0x00007ffcd9b4c000)
libnvinfer.so.7 => /home/oem/pgf/deploy/YOLOX/TensorRT-7.2.3.4/lib/libnvinfer.so.7 (0x00007f1743c3b000)
libcudart.so.10.2 => /usr/local/cuda/targets/x86_64-linux/lib/libcudart.so.10.2 (0x00007f17439bd000)
libopencv_imgcodecs.so.405 => /home/oem/opencv/opencv-4.5.5/build/lib/libopencv_imgcodecs.so.405 (0x00007f17435de000)
libopencv_imgproc.so.405 => /home/oem/opencv/opencv-4.5.5/build/lib/libopencv_imgproc.so.405 (0x00007f1741a5d000)
libopencv_core.so.405 => /home/oem/opencv/opencv-4.5.5/build/lib/libopencv_core.so.405 (0x00007f1740a34000)
libstdc++.so.6 => /lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007f174083a000)
libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f17406eb000)
libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007f17406d0000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f17404de000)
libcudnn.so.8 => /usr/local/cuda/targets/x86_64-linux/lib/libcudnn.so.8 (0x00007f17402b4000)
libmyelin.so.1 => /home/oem/pgf/deploy/YOLOX/TensorRT-7.2.3.4/lib/libmyelin.so.1 (0x00007f173fa38000)
libnvrtc.so.10.2 => /usr/local/cuda/targets/x86_64-linux/lib/libnvrtc.so.10.2 (0x00007f173e289000)
librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007f173e27f000)
libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f173e279000)
/lib64/ld-linux-x86-64.so.2 (0x00007f1754f6e000)
libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f173e256000)
libcublas.so.10 => /lib/x86_64-linux-gnu/libcublas.so.10 (0x00007f1739fa0000)
libcublasLt.so.10 => /lib/x86_64-linux-gnu/libcublasLt.so.10 (0x00007f173810b000)

模型转换环境：
python:3.7.11
pytorch:1.7.0
torchvision:0.8.1
tensorrt:7.2.3.4

Start by reproducing ./yolox ../model_trt.engine -i ../../../../assets/dog.jpg with the reported Python, PyTorch, TensorRT, CUDA, and OpenCV versions, then inspect the linked runtime libraries from the ldd yolox output. Determine whether the engine and TensorRT runtime are compatible; done means the engine deserializes and inference completes without the type-mismatch error or segmentation fault.