Why is "Layer ___ forced to im2col_v2, because its output is used in concat" compile mesage given?

idata · ‎10-02-2018

What does this message mean? This portion of the model doesn't even have a Concat layer. Should I be concerned? Prototxt below is a modified version of beginning layers of a tiny_yolo model. The compiler message refers to the final, "conv15/1x1" layer.

name: "tiny_ncsdk_dts"

input: "data"
input_shape {
  dim:1
  dim:1
  dim:224
  dim:224 }

layer {
  name: "conv1"
  type: "Convolution"
  bottom: "data"
  top: "conv1"
  convolution_param {
    num_output: 32
    kernel_size: 3
    pad: 1
    stride: 2
    bias_term: false
    weight_filler {
      type: "xavier"
    }
  }
}
layer {
  name: "bn1"
  type: "BatchNorm"
  bottom: "conv1"
  top: "bn1"
}

layer {
  name: "relu1"
  type: "ReLU"
  bottom: "bn1"
  top: "relu1"
  relu_param{
    negative_slope: 0.1
  }   
}

layer{
  name: "conv2"
  type: "Convolution"
  bottom: "relu1"
  top: "conv2"
  convolution_param {
    num_output: 32
    kernel_size: 3
    pad: 1
    stride: 2
    bias_term: false
    weight_filler {
      type: "xavier"
    }
  }
}
layer {
  name: "bn2"
  type: "BatchNorm"
  bottom: "conv2"
  top: "bn2"
}

layer {
  name: "relu2"
  type: "ReLU"
  bottom: "bn2"
  top: "relu2"
  relu_param{
    negative_slope: 0.1
  }   
}

layer{
  name: "conv3"
  type: "Convolution"
  bottom: "relu2"
  top: "conv3"
  convolution_param {
    num_output: 64
    kernel_size: 3
    pad: 1
    stride: 1
    bias_term: false
    weight_filler {
      type: "xavier"
    }
  }
}
layer {
  name: "bn3"
  type: "BatchNorm"
  bottom: "conv3"
  top: "bn3"
}

layer {
  name: "relu3"
  type: "ReLU"
  bottom: "bn3"
  top: "relu3"
  relu_param{
    negative_slope: 0.1
  }   
}

layer{
  name: "conv5"
  type: "Convolution"
  bottom: "relu3"
  top: "conv5"
  convolution_param {
    num_output: 64
    kernel_size: 3
    pad: 1
    stride: 2
    bias_term: false
    weight_filler {
      type: "xavier"
    }
  }
}
layer {
  name: "bn5"
  type: "BatchNorm"
  bottom: "conv5"
  top: "bn5"
}
layer {
  name: "relu5"
  type: "ReLU"
  bottom: "bn5"
  top: "relu5"
  relu_param{
    negative_slope: 0.1
  }   
}

layer{
  name: "conv6"
  type: "Convolution"
  bottom: "relu5"
  top: "conv6"
  convolution_param {
    num_output: 128
    kernel_size: 3
    pad: 1
    stride: 1
    bias_term: false
    weight_filler {
      type: "xavier"
    }
  }
}
layer {
  name: "bn6"
  type: "BatchNorm"
  bottom: "conv6"
  top: "bn6"
}
layer {
  name: "relu6"
  type: "ReLU"
  bottom: "bn6"
  top: "relu6"
  relu_param{
    negative_slope: 0.1
  }   
}

layer{
  name: "conv8"
  type: "Convolution"
  bottom: "bn6"
  top: "conv8"
  convolution_param {
    num_output: 128
    kernel_size: 3
    pad: 1
    stride: 1
    bias_term: false
    weight_filler {
      type: "xavier"
    }
  }
}
layer {
  name: "bn8"
  type: "BatchNorm"
  bottom: "conv8"
  top: "bn8"
}
layer {
  name: "relu8"
  type: "ReLU"
  bottom: "bn8"
  top: "relu8"
  relu_param{
    negative_slope: 0.1
  }   
}
layer {
  name: "pool8"
  type: "Pooling"
  bottom: "relu8"
  top: "pool8"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}

layer{
  name: "conv9"
  type: "Convolution"
  bottom: "pool8"
  top: "conv9"
  convolution_param {
    num_output: 256
    kernel_size: 3
    pad: 1
    stride: 1
    bias_term: false
    weight_filler {
      type: "xavier"
    }
  }
}
layer {
  name: "bn9"
  type: "BatchNorm"
  bottom: "conv9"
  top: "bn9"
}
layer {
  name: "relu9"
  type: "ReLU"
  bottom: "bn9"
  top: "relu9"
  relu_param{
    negative_slope: 0.1
  }   
}

layer{
  name: "conv11"
  type: "Convolution"
  bottom: "relu9"
  top: "conv11"
  convolution_param {
    num_output: 256
    kernel_size: 3
    pad: 1
    stride: 1
    bias_term: false
    weight_filler {
      type: "xavier"
    }
  }
  param {  
    name: "conv11_w"
  }
}
layer {
  name: "relu11"
  type: "ReLU"
  bottom: "conv11"
  top: "relu11"
  relu_param{
    negative_slope: 0.1
  }   
}

layer{
  name: "conv11_2"
  type: "Convolution"
  bottom: "relu11"
  top: "conv11_2"
  convolution_param {
    num_output: 256
    kernel_size: 3
    pad: 1
    stride: 1
    bias_term: false
    weight_filler {
      type: "xavier"
    }
  }
  param {
    name: "conv11_w"
  }
}
layer {
  name: "bn11_2"
  type: "BatchNorm"
  bottom: "conv11_2"
  top: "bn11_2"
}
layer {
  name: "relu11_2"
  type: "ReLU"
  bottom: "bn11_2"
  top: "relu11_2"
  relu_param{
    negative_slope: 0.1
  }   
}

#####
layer {
  name: "conv14/3x3"
  type: "Convolution"
  bottom: "relu11_2"
  top: "conv14/3x3"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  convolution_param {
    num_output: 512
    pad: 1
    kernel_size: 3
    stride: 1
    group: 256
    weight_filler {
      type: "msra"
    }
    bias_term: false
    engine: CAFFE
  }
}
layer {
  name: "conv14/3x3/bn"
  type: "BatchNorm"
  bottom: "conv14/3x3"
  top: "conv14/3x3/bn"
}
layer {
  name: "conv14/3x3/scale"
  type: "Scale"
  bottom: "conv14/3x3/bn"
  top: "conv14/3x3/scale"
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv14/relu1"
  type: "ReLU"
  bottom: "conv14/3x3/scale"
  top: "conv14/relu1"
  relu_param{
    negative_slope: 0.1
  }  
}
layer {
  name: "conv14/1x1"
  type: "Convolution"
  bottom: "conv14/relu1"
  top: "conv14/1x1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  convolution_param {
    num_output: 512
    pad: 0
    kernel_size: 1
    stride: 1
    weight_filler {
      type: "msra"
    }
  bias_term: false
  }
}
layer {
  name: "conv14/1x1/bn"
  type: "BatchNorm"
  bottom: "conv14/1x1"
  top: "conv14/1x1/bn"
}
layer {
  name: "conv14/1x1/scale"
  type: "Scale"
  bottom: "conv14/1x1/bn"
  top: "conv14/1x1/scale"
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv14/relu2"
  type: "ReLU"
  bottom: "conv14/1x1/scale"
  top: "conv14/1x1/relu"
  relu_param{
    negative_slope: 0.1
  }  
}
#####
#  Had typo below, in which layer name had cap X: "conv15/3X3"
layer {
  name: "conv15/3x3"
  type: "Convolution"
  bottom: "conv14/1x1/relu"
  top: "conv15/3x3"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  convolution_param {
    num_output: 512
    pad: 1
    kernel_size: 3
    stride: 1
    group: 512
    #dilation: 2
    weight_filler {
      type: "msra"
    }
    bias_term: false
    engine: CAFFE
  }
}
layer {
  name: "conv15/3x3/bn"
  type: "BatchNorm"
  bottom: "conv15/3x3"
  top: "conv15/3x3/bn"
}
layer {
  name: "conv15/3x3/scale"
  type: "Scale"
  bottom: "conv15/3x3/bn"
  top: "conv15/3x3/scale"
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv15/relu1"
  type: "ReLU"
  bottom: "conv15/3x3/scale"
  top: "conv15/relu1"
  relu_param{
    negative_slope: 0.1
  }  
}
layer {
  name: "conv15/1x1"
  type: "Convolution"
  bottom: "conv15/relu1"
  top: "conv15/1x1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  convolution_param {
    num_output: 512
    pad: 0
    kernel_size: 1
    stride: 1
    weight_filler {
      type: "msra"
    }
  bias_term: false
  }
}

root@synapse:~/Code/DeepTextSpotter/models# mvNCCompile -s 12 -o temp.graph junk.prototxt 
mvNCCompile v02.00, Copyright @ Intel Corporation 2017

****** WARNING: using empty weights ******
Layer  conv15/1x1  forced to im2col_v2, because its output is used in concat
/usr/local/bin/ncsdk/Controllers/FileIO.py:65: UserWarning: You are using a large type. Consider reducing your data sizes for best performance
Blob generated