使用python连接caffe中的relu

Question

我想用caffe在python中实现“concatenatedrelu”。 它定义为ConcRelu=Concat（Relu（x），-Relu（-x））

贝娄是一个天真的实现，但不工作！我会感激任何帮助！你知道吗

def prelu(self,bottom,args):
    'prelu layer'
    return L.PReLU(bottom, in_place=True)

def crelu(self,bottom,args):
    'crelu layer'
    negative_slope = 0
    if len(args) == 1:
        negative_slope = float(args[0])
    rp = {'negative_slope': negative_slope}

    sp = {  'bias_term': False,
            'filler': {'value': -1}
    }

    neg_side = L.Scale(bottom, scale_param=sp)

    a = [L.ReLU(neg_side, relu_param=rp, in_place=True), L.ReLU(bottom, relu_param=rp, in_place=True)]
    cp = { 'axis' : 1 }

    return L.Concat(*a, concat_param=cp)

===我得到的输出（有错误）：===

WARNING: Logging before InitGoogleLogging() is written to STDERR
I0130 17:15:01.251324 145276 upgrade_proto.cpp:67] Attempting to upgrade input file specified using deprecated input fields: res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_tmp/CNN_020_deploy.txt

I0130 17:15:01.251384 145276 upgrade_proto.cpp:70] Successfully upgraded file specified using deprecated input fields.
W0130 17:15:01.251389 145276 upgrade_proto.cpp:72] Note that future Caffe releases will only support input layers and not input fields.
I0130 17:15:01.657894 145276 net.cpp:51] Initializing net from parameters:
name: "CNN_020"
state {
  phase: TEST
  level: 0
}
layer {
  name: "input"
  type: "Input"
  top: "data"
  input_param {
    shape {
      dim: 1
      dim: 2
      dim: 1
      dim: 2048
    }
  }
}
layer {
  name: "conv1"
  type: "Convolution"
  bottom: "data"
  top: "conv1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  convolution_param {
    num_output: 16
    bias_term: false
    pad: 0
    pad: 0
    kernel_size: 1
    kernel_size: 3
    group: 1
    stride: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    axis: 1
  }
}
layer {
  name: "Scale1"
  type: "Scale"
  bottom: "conv1"
  top: "Scale1"
  scale_param {
    filler {
      value: -1
    }
    bias_term: false
  }
}
layer {
  name: "ReLU1"
  type: "ReLU"
  bottom: "Scale1"
  top: "Scale1"
  relu_param {
    negative_slope: 0
  }
}
layer {
  name: "ReLU2"
  type: "ReLU"
  bottom: "conv1"
  top: "conv1"
  relu_param {
    negative_slope: 0
  }
}
layer {
  name: "crelu1"
  type: "Concat"
  bottom: "Scale1"
  bottom: "conv1"
  top: "crelu1"
  concat_param {
    axis: 1
  }
}
layer {
  name: "ampl"
  type: "InnerProduct"
  bottom: "crelu1"
  top: "ampl"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  inner_product_param {
    num_output: 20
    bias_term: false
    weight_filler {
      type: "gaussian"
      std: 0.01
    }
    bias_filler {
      type: "constant"
      value: 0.2
    }
  }
}
I0130 17:15:01.658027 145276 layer_factory.hpp:77] Creating layer input
I0130 17:15:01.658046 145276 net.cpp:84] Creating Layer input
I0130 17:15:01.658058 145276 net.cpp:380] input -> data
I0130 17:15:01.658095 145276 net.cpp:122] Setting up input
I0130 17:15:01.658108 145276 net.cpp:129] Top shape: 1 2 1 2048 (4096)
I0130 17:15:01.658113 145276 net.cpp:137] Memory required for data: 16384
I0130 17:15:01.658118 145276 layer_factory.hpp:77] Creating layer conv1
I0130 17:15:01.658131 145276 net.cpp:84] Creating Layer conv1
I0130 17:15:01.658138 145276 net.cpp:406] conv1 <- data
I0130 17:15:01.658144 145276 net.cpp:380] conv1 -> conv1
I0130 17:15:01.658269 145276 net.cpp:122] Setting up conv1
I0130 17:15:01.658282 145276 net.cpp:129] Top shape: 1 16 1 2046 (32736)
I0130 17:15:01.658287 145276 net.cpp:137] Memory required for data: 147328
I0130 17:15:01.658298 145276 layer_factory.hpp:77] Creating layer conv1_conv1_0_split
I0130 17:15:01.658313 145276 net.cpp:84] Creating Layer conv1_conv1_0_split
I0130 17:15:01.658318 145276 net.cpp:406] conv1_conv1_0_split <- conv1
I0130 17:15:01.658324 145276 net.cpp:380] conv1_conv1_0_split -> conv1_conv1_0_split_0
I0130 17:15:01.658332 145276 net.cpp:380] conv1_conv1_0_split -> conv1_conv1_0_split_1
I0130 17:15:01.658344 145276 net.cpp:122] Setting up conv1_conv1_0_split
I0130 17:15:01.658350 145276 net.cpp:129] Top shape: 1 16 1 2046 (32736)
I0130 17:15:01.658355 145276 net.cpp:129] Top shape: 1 16 1 2046 (32736)
I0130 17:15:01.658360 145276 net.cpp:137] Memory required for data: 409216
I0130 17:15:01.658365 145276 layer_factory.hpp:77] Creating layer Scale1
I0130 17:15:01.658375 145276 net.cpp:84] Creating Layer Scale1
I0130 17:15:01.658380 145276 net.cpp:406] Scale1 <- conv1_conv1_0_split_0
I0130 17:15:01.658386 145276 net.cpp:380] Scale1 -> Scale1
I0130 17:15:01.658408 145276 net.cpp:122] Setting up Scale1
I0130 17:15:01.658416 145276 net.cpp:129] Top shape: 1 16 1 2046 (32736)
I0130 17:15:01.658419 145276 net.cpp:137] Memory required for data: 540160
I0130 17:15:01.658427 145276 layer_factory.hpp:77] Creating layer ReLU1
I0130 17:15:01.658435 145276 net.cpp:84] Creating Layer ReLU1
I0130 17:15:01.658440 145276 net.cpp:406] ReLU1 <- Scale1
I0130 17:15:01.658447 145276 net.cpp:367] ReLU1 -> Scale1 (in-place)
I0130 17:15:01.658455 145276 net.cpp:122] Setting up ReLU1
I0130 17:15:01.658462 145276 net.cpp:129] Top shape: 1 16 1 2046 (32736)
I0130 17:15:01.658465 145276 net.cpp:137] Memory required for data: 671104
I0130 17:15:01.658469 145276 layer_factory.hpp:77] Creating layer ReLU2
I0130 17:15:01.658474 145276 net.cpp:84] Creating Layer ReLU2
I0130 17:15:01.658479 145276 net.cpp:406] ReLU2 <- conv1_conv1_0_split_1
F0130 17:15:01.658486 145276 net.cpp:375] Top blob 'conv1' produced by multiple sources.
*** Check failure stack trace: ***
/afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/trainAndTest: line 99: 145276 Aborted                 $pydir/dumpLayersSize.py ${tmp_root}_deploy.txt ${oroot}
Tue Jan 30 17:15:01 CET 2018
/usr/bin/time -v caffe -gpu 0 --log_dir=res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_tmp train -solver res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_tmp/CNN_020_solver.txt
Tue Jan 30 17:15:20 CET 2018
cp: missing destination file operand after �~@~Xres/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel�~@~Y
Try 'cp --help' for more information.
tail: cannot open �~@~Xres/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_solver.log.train�~@~Y for reading: No such file or directory
/afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/trainAndTest: line 204: gnuplot: command not found
/afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/trainAndTest: line 223: gnuplot: command not found
/usr/bin/time -v -o res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_time_train.txt python /afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/python/testdataset.py -eg hist -l label -mf=data/2048_1e5_0.00/2048_1e5_0.00_s_met.txt res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_deploy.txt res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel data/2048_1e5_0.00/2048_1e5_0.00_s_train.h5 res/2048_1e5_0.00_s/CNN_020_bs10/restest/CNN_020_train
WARNING: Logging before InitGoogleLogging() is written to STDERR
W0130 17:15:26.614568 145537 _caffe.cpp:139] DEPRECATION WARNING - deprecated use of Python interface
W0130 17:15:26.614625 145537 _caffe.cpp:140] Use this instead (with the named "weights" parameter):
W0130 17:15:26.614630 145537 _caffe.cpp:142] Net('res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_deploy.txt', 1, weights='res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel')
('res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_deploy.txt', 'res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel', 'data/2048_1e5_0.00/2048_1e5_0.00_s_train.h5', 'data/2048_1e5_0.00/2048_1e5_0.00_s_met.txt')
Traceback (most recent call last):
  File "/afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/python/testdataset.py", line 395, in <module>
    aE = predict(model, weights, data)
  File "/afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/python/testdataset.py", line 200, in predict
    net = caffe.Net(modelFile, weightsFile, caffe.TEST)
RuntimeError: Could not open file res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel
/usr/bin/time -v -o res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_time_val.txt python /afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/python/testdataset.py -eg hist -l label -mf=data/2048_1e5_0.00/2048_1e5_0.00_s_met.txt res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_deploy.txt res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel data/2048_1e5_0.00/2048_1e5_0.00_s_val.h5 res/2048_1e5_0.00_s/CNN_020_bs10/restest/CNN_020_val
WARNING: Logging before InitGoogleLogging() is written to STDERR
W0130 17:15:32.969283 145545 _caffe.cpp:139] DEPRECATION WARNING - deprecated use of Python interface
W0130 17:15:32.969337 145545 _caffe.cpp:140] Use this instead (with the named "weights" parameter):
W0130 17:15:32.969341 145545 _caffe.cpp:142] Net('res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_deploy.txt', 1, weights='res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel')
('res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_deploy.txt', 'res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel', 'data/2048_1e5_0.00/2048_1e5_0.00_s_val.h5', 'data/2048_1e5_0.00/2048_1e5_0.00_s_met.txt')
Traceback (most recent call last):
  File "/afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/python/testdataset.py", line 395, in <module>
    aE = predict(model, weights, data)
  File "/afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/python/testdataset.py", line 200, in predict
    net = caffe.Net(modelFile, weightsFile, caffe.TEST)
RuntimeError: Could not open file res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel
/usr/bin/time -v -o res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_time_test.txt python /afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/python/testdataset.py -eg hist -l label -mf=data/2048_1e5_0.00/2048_1e5_0.00_s_met.txt res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_deploy.txt res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel data/2048_1e5_0.00/2048_1e5_0.00_s_test.h5 res/2048_1e5_0.00_s/CNN_020_bs10/restest/CNN_020_test
WARNING: Logging before InitGoogleLogging() is written to STDERR
W0130 17:15:38.393525 145552 _caffe.cpp:139] DEPRECATION WARNING - deprecated use of Python interface
W0130 17:15:38.393584 145552 _caffe.cpp:140] Use this instead (with the named "weights" parameter):
W0130 17:15:38.393587 145552 _caffe.cpp:142] Net('res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_deploy.txt', 1, weights='res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel')
('res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020_deploy.txt', 'res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel', 'data/2048_1e5_0.00/2048_1e5_0.00_s_test.h5', 'data/2048_1e5_0.00/2048_1e5_0.00_s_met.txt')
Traceback (most recent call last):
  File "/afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/python/testdataset.py", line 395, in <module>
    aE = predict(model, weights, data)
  File "/afs/in2p3.fr/home/n/nhatami/sps/spectroML/src/python/testdataset.py", line 200, in predict
    net = caffe.Net(modelFile, weightsFile, caffe.TEST)
RuntimeError: Could not open file res/2048_1e5_0.00_s/CNN_020_bs10/CNN_020.caffemodel