darknet.dll 在C#中调用出错 cuDNN Error cudnn_convolutional_setup()

Question

我采用的时darknet V4版本，在使用CUDA10.2 CUDNN8.0.5 OPENCV4.5.3编译完毕darknet之后，做darknet.exe测试，可以检测到dog图片，采用调用摄像头检测也是可以，并且可以看到GPU已经被使用。但是在使用darknet.dll时，出现错误。加载网络到一定时就报错，具体错误代码如下：

151 route  147                                    ->   38 x  38 x 256 
 152 conv    512       3 x 3/ 2     38 x  38 x 256 ->   19 x  19 x 512 0.852 BF
 153 route  152 116                                ->   19 x  19 x1024 
 154 conv    512       1 x 1/ 1     19 x  19 x1024 ->   19 x  19 x 512 0.379 BF
 155 conv   1024       3 x 3/ 1     19 x  19 x 512 ->   19 x  19 x1024 3.407 BF
 156 conv    512       1 x 1/ 1     19 x  19 x1024 ->   19 x  19 x 512 0.379 BF
 157 conv   1024       3 x 3/ 1     19 x  19 x 512 ->   19 x  19 x1024 3.407 BF
 158 conv    512       1 x 1/ 1     19 x  19 x1024 ->   19 x  19 x 512 0.379 BF
 159 conv   1024       3 x 3/ 1     19 x  19 x 512 ->   19 x  19 x1024 3.407 BF
 160 conv    255       1 x 1/ 1     19 x  19 x1024 ->   19 x  19 x 255 0.189 BF
 161 yolo
[yolo] params: iou loss: ciou (4), iou_norm: 0.07, obj_norm: 1.00, cls_norm: 1.00, delta_norm: 1.00, scale_x_y: 1.05
Total BFLOPS 128.459 
avg_outputs = 1068395 
 Allocate additional workspace_size = 81.03 MB 
Loading weights from .\yolov4.weights... Used GPU 0 
net.optimized_memory = 0 
mini_batch = 1, batch = 16, time_steps = 1, train = 1 
Create CUDA-stream - 0 
 Create cudnn-handle 0 
nms_kind: greedynms (1), beta = 0.600000 
nms_kind: greedynms (1), beta = 0.600000 
nms_kind: greedynms (1), beta = 0.600000 
Done! Loaded 162 layers from weights-file 
Darknet error location: E:\TRY\darknet-master\src\dark_cuda.c, cudnn_check_error, line #205
cuDNN Error: CUDNN_STATUS_BAD_PARAM: Bad file descriptor
 seen 64, trained: 32032 K-images (500 Kilo-batches_64) 
 cuDNN status Error in: file: E:\TRY\darknet-master\src\convolutional_layer.c : cudnn_convolutional_setup() : line: 253 : build time: Aug 17 2021 - 20:41:18 
 cuDNN Error: CUDNN_STATUS_BAD_PARAM。

我调用darknet.dll采用的C# wrapper。其调用的代码为：

  [DllImport(YoloLibraryGpu, EntryPoint = "init")]
        internal static extern int InitializeYoloGpu(string configurationFilename, string weightsFilename, int gpuIndex);

        [DllImport(YoloLibraryGpu, EntryPoint = "detect_image")]
        internal static extern int DetectImageGpu(string filename, ref BboxContainer container);

        [DllImport(YoloLibraryGpu, EntryPoint = "detect_mat")]
        internal static extern int DetectImageGpu(IntPtr pArray, int nSize, ref BboxContainer container);

        [DllImport(YoloLibraryGpu, EntryPoint = "dispose")]
        internal static extern int DisposeYoloGpu();

        [DllImport(YoloLibraryGpu, EntryPoint = "get_device_count")]
        internal static extern int GetDeviceCount();

        [DllImport(YoloLibraryGpu, EntryPoint = "get_device_name")]
        internal static extern int GetDeviceName(int gpu, StringBuilder deviceName);
        //每次程序都在初始化darknet网络的时候报错，即在下面这一行代码：
     this._yoloWrapper = new YoloWrapper(config.ConfigFile, config.WeightsFile, config.NamesFile, gpuConfig);

后面又换到cuda11.0，cudnn换到8.2.0，问题还是照旧出现！！！！
请码友帮忙参谋一下，请收下我的膝盖！

Answer 1

暂时不知道什么原因。你可以把测试打包代码发一下，不然光看也不知道原因。我猜测是权重初始化出了问题，dark net我也常常遇到各种莫名其妙的问题。