when I converted the ONNX model to rtr engine , core dumped

[DeepBehavier]

I have retrained the two-stage model of uniad and successfully exported the ONNX model. when I converted the ONNX model to rtr engine, core dumped. Using gdb to locate the "std::unique_ptr serializedEngine{builder.buildSerializedNetwork(*env.network, *config)};" in networkToSerializedEngine function.

The gdp stack message is as follows：

Thread 1 "trtexec" received signal SIGSEGV, Segmentation fault.
0x00007ffff78e6328 in ?? () from /lib/x86_64-linux-gnu/libgcc_s.so.1
(gdb) bt
#0 0x00007ffff78e6328 in ?? () from /lib/x86_64-linux-gnu/libgcc_s.so.1
#1 0x00007ffff78e6fd1 in _Unwind_Find_FDE () from /lib/x86_64-linux-gnu/libgcc_s.so.1
#2 0x00007ffff78e260a in ?? () from /lib/x86_64-linux-gnu/libgcc_s.so.1
#3 0x00007ffff78e407d in _Unwind_RaiseException () from /lib/x86_64-linux-gnu/libgcc_s.so.1
#4 0x00007ffff7af524b in __cxa_throw () from /lib/x86_64-linux-gnu/libstdc++.so.6
#5 0x00007fff9e153375 in ?? () from /lib/x86_64-linux-gnu/libnvinfer.so.8
#6 0x00007fff9dbf12d3 in ?? () from /lib/x86_64-linux-gnu/libnvinfer.so.8
#7 0x00007fff9dbf314a in ?? () from /lib/x86_64-linux-gnu/libnvinfer.so.8
#8 0x00007fff9dc5d56e in ?? () from /lib/x86_64-linux-gnu/libnvinfer.so.8
#9 0x00007fff9e0078ae in ?? () from /lib/x86_64-linux-gnu/libnvinfer.so.8
#10 0x00007fff9de684e6 in ?? () from /lib/x86_64-linux-gnu/libnvinfer.so.8
#11 0x00007fff9de6d4ef in ?? () from /lib/x86_64-linux-gnu/libnvinfer.so.8
#12 0x00007fff9de6dfb4 in ?? () from /lib/x86_64-linux-gnu/libnvinfer.so.8
#13 0x00005555555fce34 in nvinfer1::IBuilder::buildSerializedNetwork(nvinfer1::INetworkDefinition&, nvinfer1::IBuilderConfig&) ()
#14 0x00005555555f76df in sample::networkToSerializedEngine(sample::BuildOptions const&, sample::SystemOptions const&, nvinfer1::IBuilder&, sample::BuildEnvironment&, std::ostream&) ()
#15 0x00005555555f7f8f in sample::modelToBuildEnv(sample::ModelOptions const&, sample::BuildOptions const&, sample::SystemOptions&, sample::BuildEnvironment&, std::ostream&) ()
#16 0x00005555555f8fb5 in sample::getEngineBuildEnv(sample::ModelOptions const&, sample::BuildOptions const&, sample::SystemOptions&, sample::BuildEnvironment&, std::ostream&) ()
#17 0x000055555555f9be in main ()

My environment is as follows:
GPU: 3090
TensorRT: 8.6.1.6
Cuda: 11.8
Torch 1.12

when convert onnx to trt engine, the instructions are as follows:

/usr/src/tensorrt/bin/trtexec
--onnx=/media/oem2/data2/zhh/zhh/end2end/trt/DL4AGX/AV-Solutions/uniad-trt/UniAD/onnx/uniad_tiny_imgx0.25_cp.repaired.onnx
--saveEngine=./save_engine
--plugins=./build/libuniad_plugin.so
--verbose
--dumpLayerInfo
--dumpProfile
--separateProfileRun
--profilingVerbosity=detailed
--useCudaGraph
--minShapes=${SHAPES//${MIN}/${MIN}}
--optShapes=${SHAPES//${MIN}/${OPT}}
--maxShapes=${SHAPES//${MIN}/${MAX}}
--loadInputs=${INPUTS}

On the nvidia Tensorrt website, the highest version of the 8.6 series is 8.6.1.6.

How can I solve it?