入门篇-build_on_x86_cuda.sh执行出错

hmchen

[ 20%] Built target sv_gimbal
make: *** [Makefile:130：all] 错误 2



Ubuntu版本：Ubuntu 20.04.6 LTS / Gnome版本：3.36.8 / RTX 4060 Ti

以下依赖环境已配齐：
GLIBC (2.34 or later)
CMake (version 3.0 or later)
MinGw (Windows下需使用MinGw进行构建及编译)





NVCC已安装，但CUDNN按1.2节“1.2 CUDNN8.9.7+TensorRT8.6.1.6安装”一直失败，提示权限不足和无法找到文件地址，如下：
#CUDNN897+TensorRT8616安装chmod +x ubuntu2004-cudnn-12-0-for40series.sh && ./ubuntu2004-cudnn-12-0-for40series.sh#安装完成后需要重启

hmchen

CuDNN报错如下：



TensorRT报错如下：


还请协助debug，谢谢～

hmchen

10/31更新：验证是make -j4行出错，提示是：
[ 20%] Building NVCC (Device) object CMakeFiles/sv_yoloplugins.dir/algorithm/common_det/cuda/yolov7/sv_yoloplugins_generated_yololayer.cu.o
In file included from /home/hmchen/SpireCV/algorithm/common_det/cuda/yolov7/yololayer.h:4,
                 from /home/hmchen/SpireCV/algorithm/common_det/cuda/yolov7/yololayer.cu:1:
/home/hmchen/SpireCV/algorithm/common_det/cuda/yolov7/macros.h:4:10: fatal error: NvInfer.h: 没有那个文件或目录
    4 | #include <NvInfer.h>
      |          ^~~~~~~~~~~
compilation terminated.
CMake Error at sv_yoloplugins_generated_yololayer.cu.o.Release.cmake:220 (message):
  Error generating
  /home/hmchen/SpireCV/build/CMakeFiles/sv_yoloplugins.dir/algorithm/common_det/cuda/yolov7/./sv_yoloplugins_generated_yololayer.cu.o


make[2]: *** [CMakeFiles/sv_yoloplugins.dir/build.make:65：CMakeFiles/sv_yoloplugins.dir/algorithm/common_det/cuda/yolov7/sv_yoloplugins_generated_yololayer.cu.o] 错误 1
make[1]: *** [CMakeFiles/Makefile2:449：CMakeFiles/sv_yoloplugins.dir/all] 错误 2
make: *** [Makefile:130：all] 错误 2

hmchen

上述问题已解决。新问题：https://download.amovlab.com/model/install/veri.onnx报404了，请尽快上线

hmchen

【bug】以下三个资源无法下载：
https://download.amovlab.com/model/install/veri.onnx
https://download.amovlab.com/model/install/MonDepthEst_In.engine
https://dl.amovlab.cn/model/install/MonDepthEst_Out.engine

对应.sh文件代码：

#veri_model1="veri.onnx"
#veri_model1_fn=${root_dir}/${veri_model1}

#mde_model1="MonDepthEst_In.engine"
#mde_model1_fn=${root_dir}/${mde_model1}

#mde_model2="MonDepthEst_Out.engine"
#mde_model2_fn=${root_dir}/${mde_model2}