tensorrt engine-＞serialize()报错调试

１．打开log信息打印

/******************************/
// TensorRT utility
/******************************/
class Logger : public ILogger
{
    void log(Severity severity, const char *msg) override
    {
        if (severity != Severity::kINFO)
            std::cout << msg << std::endl;
    }
};

２．查看如下报错

cuda/cudaFusedConvActLayer.cpp (287) - Cuda Error in executeFused: 9
cuda/cudaFusedConvActLayer.cpp (287) - Cuda Error in executeFused: 9

可在gedit /usr/local/cuda/include/driver_types.h中找到相关报错信息

３.得到结论

一维卷积单个维度太大，更换为二维卷积