Darknet训练模型错误

按照github文档做完，在测试时报错（执行检测的命令可以运行成功）

请问这个cuDNN无法找到FWD算法怎么解决。

问题已经解决，详细可见github相同问题https://github.com/AlexeyAB/darknet/issues/7669

总的来说就是训练超过GPU显存，修改batches和宽高即可

我看到opencv是3.4版本的，cuDNN是8。cuda版本感觉是高了，建议按照下面配置检查一下，别的目前从报错中看不到其他错误，如有帮助请采纳：

• CMake >= 3.18: https://cmake.org/download/
• Powershell (already installed on windows): https://docs.microsoft.com/en-us/powershell/scripting/install/installing-powershell
• CUDA >= 10.2: https://developer.nvidia.com/cuda-toolkit-archive (on Linux do Post-installation Actions)
• OpenCV >= 2.4: use your preferred package manager (brew, apt), build from source using vcpkg or download from OpenCV official site (on Windows set system variable OpenCV_DIR = C:\opencv\build - where are the include and x64 folders image)
• cuDNN >= 8.0.2 https://developer.nvidia.com/rdp/cudnn-archive (on Linux copy cudnn.h,libcudnn.so... as described here https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html#installlinux-tar , on Windows copy cudnn.h,cudnn64_7.dll, cudnn64_7.lib as described here https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html#installwindows )
• GPU with CC >= 3.0: https://en.wikipedia.org/wiki/CUDA#GPUs_supported