Caffe的LeNet实现

最新推荐文章于 2024-07-09 21:10:27 发布
最新推荐文章于 2024-07-09 21:10:27 发布
阅读量346 收藏 2
点赞数
分类专栏： Caffe 文章标签： Caffe lenet
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/weixin_42717395/article/details/88946814
版权

lenet_solver.prototxt：

net: "/home/zhaoys/myf/lenet/lenet_caffe/lenet_train_test.prototxt"
# 测试的时候，要迭代的次数。
# 应该满足：test_iter * test_batchsize = testset_images，这样可以覆盖整个测试集。
test_iter: 100
# 训练的时候，经过多少次迭代，进行一次测试。
# 应该满足：test_interval * train_batchsize = trainset_images，这样可以覆盖整个训练集。
test_interval: 500
# 基础的学习率，不宜太大，也不宜太小。
base_lr: 0.01
# 动量，使得模型平稳快速收敛。
momentum: 0.9
# 权值衰减，惩罚项，防止过拟合
weight_decay: 0.0005
# 学习率变化的策略
lr_policy: "inv"
# 学习率的衰减
gamma: 0.0001
# 指数
power: 0.75
# 进过多少次显示一次loss数值
display: 100
# 最大迭代次数
max_iter: 10000
# 保存模型数据
snapshot: 5000
snapshot_prefix: "/home/zhaoys/myf/lenet/lenet_caffe/snapshot/train"
# 训练模式
solver_mode: GPU

lenet_train_test.prototxt：

name: "LeNet"
layer {
	name: "mnist"
	type: "Data"
	top: "data"
	top: "label"
	include {
		phase: TRAIN
	}
	transform_param {
		scale: 0.00390625
	}
	data_param {
		source: "/home/zhaoys/myf/dataset/mnist/mnist_train_lmdb"
		batch_size: 64
		backend: LMDB
	}
}
layer {
	name: "mnist"
	type: "Data"
	top: "data"
	top: "label"
	include {
		phase: TEST
	}
	transform_param {
		scale: 0.00390625
	}
	data_param {
		source: "/home/zhaoys/myf/dataset/mnist/mnist_test_lmdb"
		batch_size: 100
		backend: LMDB
	}
}
layer {
	name: "conv1"
	type: "Convolution"
	bottom: "data"
	top: "conv1"
	param {
		lr_mult: 1
	}
	param {
		lr_mult: 2
	}
	convolution_param {
		num_output: 20
		kernel_size: 5
		stride: 1
		weight_filler {
			type: "xavier"
		}
		bias_filler {
			type: "constant"
		}
	}
}
layer {
	name: "pool1"
	type: "Pooling"
	bottom: "conv1"
	top: "pool1"
	pooling_param {
		pool: MAX
		kernel_size: 2
		stride: 2
	}
}
layer {
	name: "conv2"
	type: "Convolution"
	bottom: "pool1"
	top: "conv2"
	param {
		lr_mult: 1
	}
	param {
		lr_mult: 2
	}
	convolution_param {
		num_output: 50
		kernel_size: 5
		stride: 1
		weight_filler {
			type: "xavier"
		}
		bias_filler {
			type: "constant"
		}
	}
}

layer {
	name: "pool2"
	type: "Pooling"
	bottom: "conv2"
	top: "pool2"
	pooling_param {
		pool: MAX
		kernel_size: 2
		stride: 2
	}
}
layer {
	name: "ip1"
	type: "InnerProduct"
	bottom: "pool2"
	top: "ip1"
	param {
		lr_mult: 1
	}
	param {
		lr_mult: 2
	}
	inner_product_param {
		num_output: 500
		weight_filler {
			type: "xavier"
		}
		bias_filler {
			type: "constant"
		}
	}
}
layer {
	name: "relu1"
	type: "ReLU"
	bottom: "ip1"
	top: "ip1"
}
layer {
	name: "ip2"
	type: "InnerProduct"
	bottom: "ip1"
	top: "ip2"
	param {
		lr_mult: 1
	}
	param {
		lr_mult: 2
	}
	inner_product_param {
		num_output: 10
		weight_filler {
			type: "xavier"
		}
		bias_filler {
			type: "constant"
		}
	}
}
layer {
	name: "accuracy"
	type: "Accuracy"
	bottom: "ip2"
	bottom: "label"
	top: "accuracy"
	include {
		phase: TEST
	}
}
layer {
	name: "loss"
	type: "SoftmaxWithLoss"
	bottom: "ip2"
	bottom: "label"
	top: "loss"
}

train_lenet.sh：

#!/usr/bin/env/sh
set -e
GLOG_logtostderr=0 \
    GLOG_log_dir='/home/zhaoys/myf/lenet/lenet_caffe/log' \
    /home/panyun/sourcecode/caffe/cmake_build/tools/caffe \
    train --solver=lenet_solver.prototxt

lenet.png：

info：

Log file created at: 2018/12/31 05:14:03
Running on machine: sophie
Log line format: [IWEF]mmdd hh:mm:ss.uuuuuu threadid file:line] msg
I1231 05:14:03.284312 112226 caffe.cpp:218] Using GPUs 0
I1231 05:14:03.325053 112226 caffe.cpp:223] GPU 0: GeForce GTX 1080 Ti
I1231 05:14:03.616346 112226 solver.cpp:44] Initializing solver from parameters: 
test_iter: 100
test_interval: 500
base_lr: 0.01
display: 100
max_iter: 10000
lr_policy: "inv"
gamma: 0.0001
power: 0.75
momentum: 0.9
weight_decay: 0.0005
snapshot: 5000
snapshot_prefix: "/home/zhaoys/myf/lenet/lenet_caffe/snapshot/train"
solver_mode: GPU
device_id: 0
net: "/home/zhaoys/myf/lenet/lenet_caffe/lenet_train_test.prototxt"
train_state {
  level: 0
  stage: ""
}
I1231 05:14:03.616485 112226 solver.cpp:87] Creating training net from net file: /home/zhaoys/myf/lenet/lenet_caffe/lenet_train_test.prototxt
I1231 05:14:03.616652 112226 net.cpp:294] The NetState phase (0) differed from the phase (1) specified by a rule in layer mnist
I1231 05:14:03.616668 112226 net.cpp:294] The NetState phase (0) differed from the phase (1) specified by a rule in layer accuracy
I1231 05:14:03.616734 112226 net.cpp:51] Initializing net from parameters: 
name: "LeNet"
state {
  phase: TRAIN
  level: 0
  stage: ""
}
layer {
  name: "mnist"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TRAIN
  }
  transform_param {
    scale: 0.00390625
  }
  data_param {
    source: "/home/zhaoys/myf/dataset/mnist/mnist_train_lmdb"
    batch_size: 64
    backend: LMDB
  }
}
layer {
  name: "conv1"
  type: "Convolution"
  bottom: "data"
  top: "conv1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 20
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "pool1"
  type: "Pooling"
  bottom: "conv1"
  top: "pool1"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layer {
  name: "conv2"
  type: "Convolution"
  bottom: "pool1"
  top: "conv2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 50
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "pool2"
  type: "Pooling"
  bottom: "conv2"
  top: "pool2"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layer {
  name: "ip1"
  type: "InnerProduct"
  bottom: "pool2"
  top: "ip1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  inner_product_param {
    num_output: 500
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "relu1"
  type: "ReLU"
  bottom: "ip1"
  top: "ip1"
}
layer {
  name: "ip2"
  type: "InnerProduct"
  bottom: "ip1"
  top: "ip2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  inner_product_param {
    num_output: 10
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "loss"
  type: "SoftmaxWithLoss"
  bottom: "ip2"
  bottom: "label"
  top: "loss"
}
I1231 05:14:03.616830 112226 layer_factory.hpp:77] Creating layer mnist
I1231 05:14:03.616914 112226 db_lmdb.cpp:35] Opened lmdb /home/zhaoys/myf/dataset/mnist/mnist_train_lmdb
I1231 05:14:03.616936 112226 net.cpp:84] Creating Layer mnist
I1231 05:14:03.616943 112226 net.cpp:380] mnist -> data
I1231 05:14:03.616960 112226 net.cpp:380] mnist -> label
I1231 05:14:03.617887 112226 data_layer.cpp:45] output data size: 64,1,28,28
I1231 05:14:03.619812 112226 net.cpp:122] Setting up mnist
I1231 05:14:03.619823 112226 net.cpp:129] Top shape: 64 1 28 28 (50176)
I1231 05:14:03.619828 112226 net.cpp:129] Top shape: 64 (64)
I1231 05:14:03.619832 112226 net.cpp:137] Memory required for data: 200960
I1231 05:14:03.619838 112226 layer_factory.hpp:77] Creating layer conv1
I1231 05:14:03.619874 112226 net.cpp:84] Creating Layer conv1
I1231 05:14:03.619879 112226 net.cpp:406] conv1 <- data
I1231 05:14:03.619892 112226 net.cpp:380] conv1 -> conv1
I1231 05:14:04.082386 112226 net.cpp:122] Setting up conv1
I1231 05:14:04.082428 112226 net.cpp:129] Top shape: 64 20 24 24 (737280)
I1231 05:14:04.082432 112226 net.cpp:137] Memory required for data: 3150080
I1231 05:14:04.082495 112226 layer_factory.hpp:77] Creating layer pool1
I1231 05:14:04.082514 112226 net.cpp:84] Creating Layer pool1
I1231 05:14:04.082518 112226 net.cpp:406] pool1 <- conv1
I1231 05:14:04.082530 112226 net.cpp:380] pool1 -> pool1
I1231 05:14:04.082574 112226 net.cpp:122] Setting up pool1
I1231 05:14:04.082581 112226 net.cpp:129] Top shape: 64 20 12 12 (184320)
I1231 05:14:04.082584 112226 net.cpp:137] Memory required for data: 3887360
I1231 05:14:04.082587 112226 layer_factory.hpp:77] Creating layer conv2
I1231 05:14:04.082600 112226 net.cpp:84] Creating Layer conv2
I1231 05:14:04.082604 112226 net.cpp:406] conv2 <- pool1
I1231 05:14:04.082609 112226 net.cpp:380] conv2 -> conv2
I1231 05:14:04.084203 112226 net.cpp:122] Setting up conv2
I1231 05:14:04.084215 112226 net.cpp:129] Top shape: 64 50 8 8 (204800)
I1231 05:14:04.084218 112226 net.cpp:137] Memory required for data: 4706560
I1231 05:14:04.084225 112226 layer_factory.hpp:77] Creating layer pool2
I1231 05:14:04.084233 112226 net.cpp:84] Creating Layer pool2
I1231 05:14:04.084237 112226 net.cpp:406] pool2 <- conv2
I1231 05:14:04.084242 112226 net.cpp:380] pool2 -> pool2
I1231 05:14:04.084270 112226 net.cpp:122] Setting up pool2
I1231 05:14:04.084275 112226 net.cpp:129] Top shape: 64 50 4 4 (51200)
I1231 05:14:04.084281 112226 net.cpp:137] Memory required for data: 4911360
I1231 05:14:04.084285 112226 layer_factory.hpp:77] Creating layer ip1
I1231 05:14:04.084292 112226 net.cpp:84] Creating Layer ip1
I1231 05:14:04.084295 112226 net.cpp:406] ip1 <- pool2
I1231 05:14:04.084300 112226 net.cpp:380] ip1 -> ip1
I1231 05:14:04.086802 112226 net.cpp:122] Setting up ip1
I1231 05:14:04.086815 112226 net.cpp:129] Top shape: 64 500 (32000)
I1231 05:14:04.086818 112226 net.cpp:137] Memory required for data: 5039360
I1231 05:14:04.086827 112226 layer_factory.hpp:77] Creating layer relu1
I1231 05:14:04.086833 112226 net.cpp:84] Creating Layer relu1
I1231 05:14:04.086838 112226 net.cpp:406] relu1 <- ip1
I1231 05:14:04.086843 112226 net.cpp:367] relu1 -> ip1 (in-place)
I1231 05:14:04.087512 112226 net.cpp:122] Setting up relu1
I1231 05:14:04.087523 112226 net.cpp:129] Top shape: 64 500 (32000)
I1231 05:14:04.087527 112226 net.cpp:137] Memory required for data: 5167360
I1231 05:14:04.087529 112226 layer_factory.hpp:77] Creating layer ip2
I1231 05:14:04.087536 112226 net.cpp:84] Creating Layer ip2
I1231 05:14:04.087540 112226 net.cpp:406] ip2 <- ip1
I1231 05:14:04.087546 112226 net.cpp:380] ip2 -> ip2
I1231 05:14:04.088604 112226 net.cpp:122] Setting up ip2
I1231 05:14:04.088616 112226 net.cpp:129] Top shape: 64 10 (640)
I1231 05:14:04.088619 112226 net.cpp:137] Memory required for data: 5169920
I1231 05:14:04.088625 112226 layer_factory.hpp:77] Creating layer loss
I1231 05:14:04.088634 112226 net.cpp:84] Creating Layer loss
I1231 05:14:04.088637 112226 net.cpp:406] loss <- ip2
I1231 05:14:04.088641 112226 net.cpp:406] loss <- label
I1231 05:14:04.088649 112226 net.cpp:380] loss -> loss
I1231 05:14:04.094878 112226 layer_factory.hpp:77] Creating layer loss
I1231 05:14:04.095325 112226 net.cpp:122] Setting up loss
I1231 05:14:04.095335 112226 net.cpp:129] Top shape: (1)
I1231 05:14:04.095340 112226 net.cpp:132]     with loss weight 1
I1231 05:14:04.095353 112226 net.cpp:137] Memory required for data: 5169924
I1231 05:14:04.095356 112226 net.cpp:198] loss needs backward computation.
I1231 05:14:04.095363 112226 net.cpp:198] ip2 needs backward computation.
I1231 05:14:04.095367 112226 net.cpp:198] relu1 needs backward computation.
I1231 05:14:04.095371 112226 net.cpp:198] ip1 needs backward computation.
I1231 05:14:04.095374 112226 net.cpp:198] pool2 needs backward computation.
I1231 05:14:04.095377 112226 net.cpp:198] conv2 needs backward computation.
I1231 05:14:04.095381 112226 net.cpp:198] pool1 needs backward computation.
I1231 05:14:04.095384 112226 net.cpp:198] conv1 needs backward computation.
I1231 05:14:04.095388 112226 net.cpp:200] mnist does not need backward computation.
I1231 05:14:04.095392 112226 net.cpp:242] This network produces output loss
I1231 05:14:04.095417 112226 net.cpp:255] Network initialization done.
I1231 05:14:04.095551 112226 solver.cpp:172] Creating test net (#0) specified by net file: /home/zhaoys/myf/lenet/lenet_caffe/lenet_train_test.prototxt
I1231 05:14:04.095569 112226 net.cpp:294] The NetState phase (1) differed from the phase (0) specified by a rule in layer mnist
I1231 05:14:04.095638 112226 net.cpp:51] Initializing net from parameters: 
name: "LeNet"
state {
  phase: TEST
}
layer {
  name: "mnist"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TEST
  }
  transform_param {
    scale: 0.00390625
  }
  data_param {
    source: "/home/zhaoys/myf/dataset/mnist/mnist_test_lmdb"
    batch_size: 100
    backend: LMDB
  }
}
layer {
  name: "conv1"
  type: "Convolution"
  bottom: "data"
  top: "conv1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 20
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "pool1"
  type: "Pooling"
  bottom: "conv1"
  top: "pool1"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layer {
  name: "conv2"
  type: "Convolution"
  bottom: "pool1"
  top: "conv2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 50
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "pool2"
  type: "Pooling"
  bottom: "conv2"
  top: "pool2"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layer {
  name: "ip1"
  type: "InnerProduct"
  bottom: "pool2"
  top: "ip1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  inner_product_param {
    num_output: 500
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "relu1"
  type: "ReLU"
  bottom: "ip1"
  top: "ip1"
}
layer {
  name: "ip2"
  type: "InnerProduct"
  bottom: "ip1"
  top: "ip2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  inner_product_param {
    num_output: 10
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "accuracy"
  type: "Accuracy"
  bottom: "ip2"
  bottom: "label"
  top: "accuracy"
  include {
    phase: TEST
  }
}
layer {
  name: "loss"
  type: "SoftmaxWithLoss"
  bottom: "ip2"
  bottom: "label"
  top: "loss"
}
I1231 05:14:04.095754 112226 layer_factory.hpp:77] Creating layer mnist
I1231 05:14:04.095804 112226 db_lmdb.cpp:35] Opened lmdb /home/zhaoys/myf/dataset/mnist/mnist_test_lmdb
I1231 05:14:04.095819 112226 net.cpp:84] Creating Layer mnist
I1231 05:14:04.095824 112226 net.cpp:380] mnist -> data
I1231 05:14:04.095830 112226 net.cpp:380] mnist -> label
I1231 05:14:04.095901 112226 data_layer.cpp:45] output data size: 100,1,28,28
I1231 05:14:04.097389 112226 net.cpp:122] Setting up mnist
I1231 05:14:04.097401 112226 net.cpp:129] Top shape: 100 1 28 28 (78400)
I1231 05:14:04.097405 112226 net.cpp:129] Top shape: 100 (100)
I1231 05:14:04.097407 112226 net.cpp:137] Memory required for data: 314000
I1231 05:14:04.097410 112226 layer_factory.hpp:77] Creating layer label_mnist_1_split
I1231 05:14:04.097416 112226 net.cpp:84] Creating Layer label_mnist_1_split
I1231 05:14:04.097421 112226 net.cpp:406] label_mnist_1_split <- label
I1231 05:14:04.097447 112226 net.cpp:380] label_mnist_1_split -> label_mnist_1_split_0
I1231 05:14:04.097455 112226 net.cpp:380] label_mnist_1_split -> label_mnist_1_split_1
I1231 05:14:04.097487 112226 net.cpp:122] Setting up label_mnist_1_split
I1231 05:14:04.097493 112226 net.cpp:129] Top shape: 100 (100)
I1231 05:14:04.097497 112226 net.cpp:129] Top shape: 100 (100)
I1231 05:14:04.097501 112226 net.cpp:137] Memory required for data: 314800
I1231 05:14:04.097504 112226 layer_factory.hpp:77] Creating layer conv1
I1231 05:14:04.097517 112226 net.cpp:84] Creating Layer conv1
I1231 05:14:04.097522 112226 net.cpp:406] conv1 <- data
I1231 05:14:04.097527 112226 net.cpp:380] conv1 -> conv1
I1231 05:14:04.099556 112226 net.cpp:122] Setting up conv1
I1231 05:14:04.099567 112226 net.cpp:129] Top shape: 100 20 24 24 (1152000)
I1231 05:14:04.099570 112226 net.cpp:137] Memory required for data: 4922800
I1231 05:14:04.099577 112226 layer_factory.hpp:77] Creating layer pool1
I1231 05:14:04.099584 112226 net.cpp:84] Creating Layer pool1
I1231 05:14:04.099588 112226 net.cpp:406] pool1 <- conv1
I1231 05:14:04.099594 112226 net.cpp:380] pool1 -> pool1
I1231 05:14:04.099628 112226 net.cpp:122] Setting up pool1
I1231 05:14:04.099633 112226 net.cpp:129] Top shape: 100 20 12 12 (288000)
I1231 05:14:04.099637 112226 net.cpp:137] Memory required for data: 6074800
I1231 05:14:04.099640 112226 layer_factory.hpp:77] Creating layer conv2
I1231 05:14:04.099649 112226 net.cpp:84] Creating Layer conv2
I1231 05:14:04.099654 112226 net.cpp:406] conv2 <- pool1
I1231 05:14:04.099660 112226 net.cpp:380] conv2 -> conv2
I1231 05:14:04.101852 112226 net.cpp:122] Setting up conv2
I1231 05:14:04.101866 112226 net.cpp:129] Top shape: 100 50 8 8 (320000)
I1231 05:14:04.101871 112226 net.cpp:137] Memory required for data: 7354800
I1231 05:14:04.101881 112226 layer_factory.hpp:77] Creating layer pool2
I1231 05:14:04.101889 112226 net.cpp:84] Creating Layer pool2
I1231 05:14:04.101893 112226 net.cpp:406] pool2 <- conv2
I1231 05:14:04.101899 112226 net.cpp:380] pool2 -> pool2
I1231 05:14:04.101932 112226 net.cpp:122] Setting up pool2
I1231 05:14:04.101939 112226 net.cpp:129] Top shape: 100 50 4 4 (80000)
I1231 05:14:04.101958 112226 net.cpp:137] Memory required for data: 7674800
I1231 05:14:04.101964 112226 layer_factory.hpp:77] Creating layer ip1
I1231 05:14:04.101974 112226 net.cpp:84] Creating Layer ip1
I1231 05:14:04.101990 112226 net.cpp:406] ip1 <- pool2
I1231 05:14:04.102005 112226 net.cpp:380] ip1 -> ip1
I1231 05:14:04.104290 112226 net.cpp:122] Setting up ip1
I1231 05:14:04.104302 112226 net.cpp:129] Top shape: 100 500 (50000)
I1231 05:14:04.104305 112226 net.cpp:137] Memory required for data: 7874800
I1231 05:14:04.104311 112226 layer_factory.hpp:77] Creating layer relu1
I1231 05:14:04.104317 112226 net.cpp:84] Creating Layer relu1
I1231 05:14:04.104321 112226 net.cpp:406] relu1 <- ip1
I1231 05:14:04.104326 112226 net.cpp:367] relu1 -> ip1 (in-place)
I1231 05:14:04.104709 112226 net.cpp:122] Setting up relu1
I1231 05:14:04.104718 112226 net.cpp:129] Top shape: 100 500 (50000)
I1231 05:14:04.104722 112226 net.cpp:137] Memory required for data: 8074800
I1231 05:14:04.104725 112226 layer_factory.hpp:77] Creating layer ip2
I1231 05:14:04.104733 112226 net.cpp:84] Creating Layer ip2
I1231 05:14:04.104737 112226 net.cpp:406] ip2 <- ip1
I1231 05:14:04.104741 112226 net.cpp:380] ip2 -> ip2
I1231 05:14:04.104845 112226 net.cpp:122] Setting up ip2
I1231 05:14:04.104851 112226 net.cpp:129] Top shape: 100 10 (1000)
I1231 05:14:04.104854 112226 net.cpp:137] Memory required for data: 8078800
I1231 05:14:04.104859 112226 layer_factory.hpp:77] Creating layer ip2_ip2_0_split
I1231 05:14:04.104866 112226 net.cpp:84] Creating Layer ip2_ip2_0_split
I1231 05:14:04.104869 112226 net.cpp:406] ip2_ip2_0_split <- ip2
I1231 05:14:04.104873 112226 net.cpp:380] ip2_ip2_0_split -> ip2_ip2_0_split_0
I1231 05:14:04.104879 112226 net.cpp:380] ip2_ip2_0_split -> ip2_ip2_0_split_1
I1231 05:14:04.104907 112226 net.cpp:122] Setting up ip2_ip2_0_split
I1231 05:14:04.104912 112226 net.cpp:129] Top shape: 100 10 (1000)
I1231 05:14:04.104916 112226 net.cpp:129] Top shape: 100 10 (1000)
I1231 05:14:04.104919 112226 net.cpp:137] Memory required for data: 8086800
I1231 05:14:04.104923 112226 layer_factory.hpp:77] Creating layer accuracy
I1231 05:14:04.104928 112226 net.cpp:84] Creating Layer accuracy
I1231 05:14:04.104933 112226 net.cpp:406] accuracy <- ip2_ip2_0_split_0
I1231 05:14:04.104938 112226 net.cpp:406] accuracy <- label_mnist_1_split_0
I1231 05:14:04.104943 112226 net.cpp:380] accuracy -> accuracy
I1231 05:14:04.104950 112226 net.cpp:122] Setting up accuracy
I1231 05:14:04.104954 112226 net.cpp:129] Top shape: (1)
I1231 05:14:04.104957 112226 net.cpp:137] Memory required for data: 8086804
I1231 05:14:04.104975 112226 layer_factory.hpp:77] Creating layer loss
I1231 05:14:04.104982 112226 net.cpp:84] Creating Layer loss
I1231 05:14:04.104985 112226 net.cpp:406] loss <- ip2_ip2_0_split_1
I1231 05:14:04.104990 112226 net.cpp:406] loss <- label_mnist_1_split_1
I1231 05:14:04.104995 112226 net.cpp:380] loss -> loss
I1231 05:14:04.105001 112226 layer_factory.hpp:77] Creating layer loss
I1231 05:14:04.105728 112226 net.cpp:122] Setting up loss
I1231 05:14:04.105739 112226 net.cpp:129] Top shape: (1)
I1231 05:14:04.105741 112226 net.cpp:132]     with loss weight 1
I1231 05:14:04.105749 112226 net.cpp:137] Memory required for data: 8086808
I1231 05:14:04.105752 112226 net.cpp:198] loss needs backward computation.
I1231 05:14:04.105758 112226 net.cpp:200] accuracy does not need backward computation.
I1231 05:14:04.105762 112226 net.cpp:198] ip2_ip2_0_split needs backward computation.
I1231 05:14:04.105765 112226 net.cpp:198] ip2 needs backward computation.
I1231 05:14:04.105768 112226 net.cpp:198] relu1 needs backward computation.
I1231 05:14:04.105772 112226 net.cpp:198] ip1 needs backward computation.
I1231 05:14:04.105775 112226 net.cpp:198] pool2 needs backward computation.
I1231 05:14:04.105780 112226 net.cpp:198] conv2 needs backward computation.
I1231 05:14:04.105783 112226 net.cpp:198] pool1 needs backward computation.
I1231 05:14:04.105787 112226 net.cpp:198] conv1 needs backward computation.
I1231 05:14:04.105790 112226 net.cpp:200] label_mnist_1_split does not need backward computation.
I1231 05:14:04.105793 112226 net.cpp:200] mnist does not need backward computation.
I1231 05:14:04.105796 112226 net.cpp:242] This network produces output accuracy
I1231 05:14:04.105801 112226 net.cpp:242] This network produces output loss
I1231 05:14:04.105811 112226 net.cpp:255] Network initialization done.
I1231 05:14:04.105844 112226 solver.cpp:56] Solver scaffolding done.
I1231 05:14:04.106065 112226 caffe.cpp:248] Starting Optimization
I1231 05:14:04.106072 112226 solver.cpp:272] Solving LeNet
I1231 05:14:04.106076 112226 solver.cpp:273] Learning Rate Policy: inv
I1231 05:14:04.106482 112226 solver.cpp:330] Iteration 0, Testing net (#0)
I1231 05:14:04.113574 112226 blocking_queue.cpp:49] Waiting for data
I1231 05:14:04.161089 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:04.161952 112226 solver.cpp:397]     Test net output #0: accuracy = 0.1022
I1231 05:14:04.161972 112226 solver.cpp:397]     Test net output #1: loss = 2.39148 (* 1 = 2.39148 loss)
I1231 05:14:04.164427 112226 solver.cpp:218] Iteration 0 (0 iter/s, 0.0583279s/100 iters), loss = 2.46638
I1231 05:14:04.164444 112226 solver.cpp:237]     Train net output #0: loss = 2.46638 (* 1 = 2.46638 loss)
I1231 05:14:04.164453 112226 sgd_solver.cpp:105] Iteration 0, lr = 0.01
I1231 05:14:04.297490 112226 solver.cpp:218] Iteration 100 (751.7 iter/s, 0.133032s/100 iters), loss = 0.230898
I1231 05:14:04.297534 112226 solver.cpp:237]     Train net output #0: loss = 0.230898 (* 1 = 0.230898 loss)
I1231 05:14:04.297541 112226 sgd_solver.cpp:105] Iteration 100, lr = 0.00992565
I1231 05:14:04.422021 112226 solver.cpp:218] Iteration 200 (803.339 iter/s, 0.124481s/100 iters), loss = 0.165111
I1231 05:14:04.422052 112226 solver.cpp:237]     Train net output #0: loss = 0.165111 (* 1 = 0.165111 loss)
I1231 05:14:04.422060 112226 sgd_solver.cpp:105] Iteration 200, lr = 0.00985258
I1231 05:14:04.544306 112226 solver.cpp:218] Iteration 300 (818.064 iter/s, 0.12224s/100 iters), loss = 0.157251
I1231 05:14:04.544351 112226 solver.cpp:237]     Train net output #0: loss = 0.157251 (* 1 = 0.157251 loss)
I1231 05:14:04.544366 112226 sgd_solver.cpp:105] Iteration 300, lr = 0.00978075
I1231 05:14:04.662237 112226 solver.cpp:218] Iteration 400 (848.329 iter/s, 0.117879s/100 iters), loss = 0.0765152
I1231 05:14:04.662271 112226 solver.cpp:237]     Train net output #0: loss = 0.0765152 (* 1 = 0.0765152 loss)
I1231 05:14:04.662276 112226 sgd_solver.cpp:105] Iteration 400, lr = 0.00971013
I1231 05:14:04.775310 112226 solver.cpp:330] Iteration 500, Testing net (#0)
I1231 05:14:04.831015 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:04.831811 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9706
I1231 05:14:04.831848 112226 solver.cpp:397]     Test net output #1: loss = 0.090288 (* 1 = 0.090288 loss)
I1231 05:14:04.832991 112226 solver.cpp:218] Iteration 500 (585.778 iter/s, 0.170713s/100 iters), loss = 0.116756
I1231 05:14:04.833009 112226 solver.cpp:237]     Train net output #0: loss = 0.116756 (* 1 = 0.116756 loss)
I1231 05:14:04.833017 112226 sgd_solver.cpp:105] Iteration 500, lr = 0.00964069
I1231 05:14:04.949153 112226 solver.cpp:218] Iteration 600 (861.143 iter/s, 0.116125s/100 iters), loss = 0.078895
I1231 05:14:04.949213 112226 solver.cpp:237]     Train net output #0: loss = 0.078895 (* 1 = 0.078895 loss)
I1231 05:14:04.949223 112226 sgd_solver.cpp:105] Iteration 600, lr = 0.0095724
I1231 05:14:05.069152 112226 solver.cpp:218] Iteration 700 (833.828 iter/s, 0.119929s/100 iters), loss = 0.142484
I1231 05:14:05.069192 112226 solver.cpp:237]     Train net output #0: loss = 0.142484 (* 1 = 0.142484 loss)
I1231 05:14:05.069200 112226 sgd_solver.cpp:105] Iteration 700, lr = 0.00950522
I1231 05:14:05.184397 112226 solver.cpp:218] Iteration 800 (868.055 iter/s, 0.1152s/100 iters), loss = 0.211653
I1231 05:14:05.184422 112226 solver.cpp:237]     Train net output #0: loss = 0.211653 (* 1 = 0.211653 loss)
I1231 05:14:05.184430 112226 sgd_solver.cpp:105] Iteration 800, lr = 0.00943913
I1231 05:14:05.299252 112226 solver.cpp:218] Iteration 900 (870.904 iter/s, 0.114823s/100 iters), loss = 0.174603
I1231 05:14:05.299284 112226 solver.cpp:237]     Train net output #0: loss = 0.174603 (* 1 = 0.174603 loss)
I1231 05:14:05.299291 112226 sgd_solver.cpp:105] Iteration 900, lr = 0.00937411
I1231 05:14:05.337515 112271 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:05.413573 112226 solver.cpp:330] Iteration 1000, Testing net (#0)
I1231 05:14:05.470590 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:05.472153 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9809
I1231 05:14:05.472216 112226 solver.cpp:397]     Test net output #1: loss = 0.0572349 (* 1 = 0.0572349 loss)
I1231 05:14:05.473395 112226 solver.cpp:218] Iteration 1000 (574.397 iter/s, 0.174095s/100 iters), loss = 0.096166
I1231 05:14:05.473451 112226 solver.cpp:237]     Train net output #0: loss = 0.0961661 (* 1 = 0.0961661 loss)
I1231 05:14:05.473464 112226 sgd_solver.cpp:105] Iteration 1000, lr = 0.00931012
I1231 05:14:05.590864 112226 solver.cpp:218] Iteration 1100 (851.755 iter/s, 0.117405s/100 iters), loss = 0.00682953
I1231 05:14:05.590906 112226 solver.cpp:237]     Train net output #0: loss = 0.00682963 (* 1 = 0.00682963 loss)
I1231 05:14:05.590914 112226 sgd_solver.cpp:105] Iteration 1100, lr = 0.00924715
I1231 05:14:05.705976 112226 solver.cpp:218] Iteration 1200 (869.114 iter/s, 0.11506s/100 iters), loss = 0.0163695
I1231 05:14:05.706015 112226 solver.cpp:237]     Train net output #0: loss = 0.0163696 (* 1 = 0.0163696 loss)
I1231 05:14:05.706025 112226 sgd_solver.cpp:105] Iteration 1200, lr = 0.00918515
I1231 05:14:05.821393 112226 solver.cpp:218] Iteration 1300 (866.807 iter/s, 0.115366s/100 iters), loss = 0.0149529
I1231 05:14:05.821424 112226 solver.cpp:237]     Train net output #0: loss = 0.014953 (* 1 = 0.014953 loss)
I1231 05:14:05.821432 112226 sgd_solver.cpp:105] Iteration 1300, lr = 0.00912412
I1231 05:14:05.935075 112226 solver.cpp:218] Iteration 1400 (880.008 iter/s, 0.113635s/100 iters), loss = 0.0102752
I1231 05:14:05.935109 112226 solver.cpp:237]     Train net output #0: loss = 0.0102753 (* 1 = 0.0102753 loss)
I1231 05:14:05.935118 112226 sgd_solver.cpp:105] Iteration 1400, lr = 0.00906403
I1231 05:14:06.049176 112226 solver.cpp:330] Iteration 1500, Testing net (#0)
I1231 05:14:06.100338 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:06.100885 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9849
I1231 05:14:06.100903 112226 solver.cpp:397]     Test net output #1: loss = 0.050994 (* 1 = 0.050994 loss)
I1231 05:14:06.101977 112226 solver.cpp:218] Iteration 1500 (599.298 iter/s, 0.166862s/100 iters), loss = 0.0748195
I1231 05:14:06.102027 112226 solver.cpp:237]     Train net output #0: loss = 0.0748197 (* 1 = 0.0748197 loss)
I1231 05:14:06.102036 112226 sgd_solver.cpp:105] Iteration 1500, lr = 0.00900485
I1231 05:14:06.215562 112226 solver.cpp:218] Iteration 1600 (880.884 iter/s, 0.113522s/100 iters), loss = 0.148792
I1231 05:14:06.215605 112226 solver.cpp:237]     Train net output #0: loss = 0.148793 (* 1 = 0.148793 loss)
I1231 05:14:06.215622 112226 sgd_solver.cpp:105] Iteration 1600, lr = 0.00894657
I1231 05:14:06.329977 112226 solver.cpp:218] Iteration 1700 (874.374 iter/s, 0.114368s/100 iters), loss = 0.0334666
I1231 05:14:06.330003 112226 solver.cpp:237]     Train net output #0: loss = 0.0334667 (* 1 = 0.0334667 loss)
I1231 05:14:06.330009 112226 sgd_solver.cpp:105] Iteration 1700, lr = 0.00888916
I1231 05:14:06.443758 112226 solver.cpp:218] Iteration 1800 (879.127 iter/s, 0.113749s/100 iters), loss = 0.0173706
I1231 05:14:06.443785 112226 solver.cpp:237]     Train net output #0: loss = 0.0173707 (* 1 = 0.0173707 loss)
I1231 05:14:06.443794 112226 sgd_solver.cpp:105] Iteration 1800, lr = 0.0088326
I1231 05:14:06.524282 112271 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:06.558614 112226 solver.cpp:218] Iteration 1900 (870.899 iter/s, 0.114824s/100 iters), loss = 0.140631
I1231 05:14:06.558640 112226 solver.cpp:237]     Train net output #0: loss = 0.140631 (* 1 = 0.140631 loss)
I1231 05:14:06.558646 112226 sgd_solver.cpp:105] Iteration 1900, lr = 0.00877687
I1231 05:14:06.670892 112226 solver.cpp:330] Iteration 2000, Testing net (#0)
I1231 05:14:06.718638 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:06.719323 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9862
I1231 05:14:06.719348 112226 solver.cpp:397]     Test net output #1: loss = 0.0417459 (* 1 = 0.0417459 loss)
I1231 05:14:06.720479 112226 solver.cpp:218] Iteration 2000 (617.924 iter/s, 0.161832s/100 iters), loss = 0.02471
I1231 05:14:06.720502 112226 solver.cpp:237]     Train net output #0: loss = 0.0247102 (* 1 = 0.0247102 loss)
I1231 05:14:06.720510 112226 sgd_solver.cpp:105] Iteration 2000, lr = 0.00872196
I1231 05:14:06.836062 112226 solver.cpp:218] Iteration 2100 (865.44 iter/s, 0.115548s/100 iters), loss = 0.0129391
I1231 05:14:06.836104 112226 solver.cpp:237]     Train net output #0: loss = 0.0129393 (* 1 = 0.0129393 loss)
I1231 05:14:06.836114 112226 sgd_solver.cpp:105] Iteration 2100, lr = 0.00866784
I1231 05:14:06.951599 112226 solver.cpp:218] Iteration 2200 (865.877 iter/s, 0.11549s/100 iters), loss = 0.0219193
I1231 05:14:06.951627 112226 solver.cpp:237]     Train net output #0: loss = 0.0219195 (* 1 = 0.0219195 loss)
I1231 05:14:06.951632 112226 sgd_solver.cpp:105] Iteration 2200, lr = 0.0086145
I1231 05:14:07.072963 112226 solver.cpp:218] Iteration 2300 (824.262 iter/s, 0.121321s/100 iters), loss = 0.0956042
I1231 05:14:07.073019 112226 solver.cpp:237]     Train net output #0: loss = 0.0956043 (* 1 = 0.0956043 loss)
I1231 05:14:07.073026 112226 sgd_solver.cpp:105] Iteration 2300, lr = 0.00856192
I1231 05:14:07.186334 112226 solver.cpp:218] Iteration 2400 (882.527 iter/s, 0.113311s/100 iters), loss = 0.0106698
I1231 05:14:07.186359 112226 solver.cpp:237]     Train net output #0: loss = 0.0106699 (* 1 = 0.0106699 loss)
I1231 05:14:07.186367 112226 sgd_solver.cpp:105] Iteration 2400, lr = 0.00851008
I1231 05:14:07.298082 112226 solver.cpp:330] Iteration 2500, Testing net (#0)
I1231 05:14:07.345970 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:07.346665 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9843
I1231 05:14:07.346683 112226 solver.cpp:397]     Test net output #1: loss = 0.0502878 (* 1 = 0.0502878 loss)
I1231 05:14:07.347754 112226 solver.cpp:218] Iteration 2500 (619.627 iter/s, 0.161387s/100 iters), loss = 0.0344929
I1231 05:14:07.347771 112226 solver.cpp:237]     Train net output #0: loss = 0.034493 (* 1 = 0.034493 loss)
I1231 05:14:07.347805 112226 sgd_solver.cpp:105] Iteration 2500, lr = 0.00845897
I1231 05:14:07.462756 112226 solver.cpp:218] Iteration 2600 (869.82 iter/s, 0.114966s/100 iters), loss = 0.0772851
I1231 05:14:07.462828 112226 solver.cpp:237]     Train net output #0: loss = 0.0772852 (* 1 = 0.0772852 loss)
I1231 05:14:07.462841 112226 sgd_solver.cpp:105] Iteration 2600, lr = 0.00840857
I1231 05:14:07.576750 112226 solver.cpp:218] Iteration 2700 (877.864 iter/s, 0.113913s/100 iters), loss = 0.0620485
I1231 05:14:07.576797 112226 solver.cpp:237]     Train net output #0: loss = 0.0620486 (* 1 = 0.0620486 loss)
I1231 05:14:07.576807 112226 sgd_solver.cpp:105] Iteration 2700, lr = 0.00835886
I1231 05:14:07.690094 112226 solver.cpp:218] Iteration 2800 (882.703 iter/s, 0.113288s/100 iters), loss = 0.00376383
I1231 05:14:07.690132 112226 solver.cpp:237]     Train net output #0: loss = 0.00376396 (* 1 = 0.00376396 loss)
I1231 05:14:07.690140 112226 sgd_solver.cpp:105] Iteration 2800, lr = 0.00830984
I1231 05:14:07.699453 112271 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:07.803148 112226 solver.cpp:218] Iteration 2900 (884.872 iter/s, 0.113011s/100 iters), loss = 0.0225361
I1231 05:14:07.803174 112226 solver.cpp:237]     Train net output #0: loss = 0.0225363 (* 1 = 0.0225363 loss)
I1231 05:14:07.803179 112226 sgd_solver.cpp:105] Iteration 2900, lr = 0.00826148
I1231 05:14:07.914687 112226 solver.cpp:330] Iteration 3000, Testing net (#0)
I1231 05:14:07.971788 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:07.972503 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9878
I1231 05:14:07.972627 112226 solver.cpp:397]     Test net output #1: loss = 0.0385816 (* 1 = 0.0385816 loss)
I1231 05:14:07.974078 112226 solver.cpp:218] Iteration 3000 (585.199 iter/s, 0.170882s/100 iters), loss = 0.00977571
I1231 05:14:07.974169 112226 solver.cpp:237]     Train net output #0: loss = 0.00977591 (* 1 = 0.00977591 loss)
I1231 05:14:07.974200 112226 sgd_solver.cpp:105] Iteration 3000, lr = 0.00821377
I1231 05:14:08.110509 112226 solver.cpp:218] Iteration 3100 (750.428 iter/s, 0.133257s/100 iters), loss = 0.0216236
I1231 05:14:08.110647 112226 solver.cpp:237]     Train net output #0: loss = 0.0216238 (* 1 = 0.0216238 loss)
I1231 05:14:08.110666 112226 sgd_solver.cpp:105] Iteration 3100, lr = 0.0081667
I1231 05:14:08.228408 112226 solver.cpp:218] Iteration 3200 (849.185 iter/s, 0.11776s/100 iters), loss = 0.0130305
I1231 05:14:08.228459 112226 solver.cpp:237]     Train net output #0: loss = 0.0130307 (* 1 = 0.0130307 loss)
I1231 05:14:08.228464 112226 sgd_solver.cpp:105] Iteration 3200, lr = 0.00812025
I1231 05:14:08.341622 112226 solver.cpp:218] Iteration 3300 (883.719 iter/s, 0.113158s/100 iters), loss = 0.0661136
I1231 05:14:08.341646 112226 solver.cpp:237]     Train net output #0: loss = 0.0661139 (* 1 = 0.0661139 loss)
I1231 05:14:08.341655 112226 sgd_solver.cpp:105] Iteration 3300, lr = 0.00807442
I1231 05:14:08.454787 112226 solver.cpp:218] Iteration 3400 (883.897 iter/s, 0.113135s/100 iters), loss = 0.00975509
I1231 05:14:08.454813 112226 solver.cpp:237]     Train net output #0: loss = 0.00975529 (* 1 = 0.00975529 loss)
I1231 05:14:08.454819 112226 sgd_solver.cpp:105] Iteration 3400, lr = 0.00802918
I1231 05:14:08.567121 112226 solver.cpp:330] Iteration 3500, Testing net (#0)
I1231 05:14:08.617847 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:08.618598 112226 solver.cpp:397]     Test net output #0: accuracy = 0.987
I1231 05:14:08.618615 112226 solver.cpp:397]     Test net output #1: loss = 0.0409665 (* 1 = 0.0409665 loss)
I1231 05:14:08.619675 112226 solver.cpp:218] Iteration 3500 (606.587 iter/s, 0.164857s/100 iters), loss = 0.00479197
I1231 05:14:08.619693 112226 solver.cpp:237]     Train net output #0: loss = 0.00479215 (* 1 = 0.00479215 loss)
I1231 05:14:08.619700 112226 sgd_solver.cpp:105] Iteration 3500, lr = 0.00798454
I1231 05:14:08.734412 112226 solver.cpp:218] Iteration 3600 (871.782 iter/s, 0.114708s/100 iters), loss = 0.0306457
I1231 05:14:08.734477 112226 solver.cpp:237]     Train net output #0: loss = 0.0306458 (* 1 = 0.0306458 loss)
I1231 05:14:08.734484 112226 sgd_solver.cpp:105] Iteration 3600, lr = 0.00794046
I1231 05:14:08.848691 112226 solver.cpp:218] Iteration 3700 (875.626 iter/s, 0.114204s/100 iters), loss = 0.0209788
I1231 05:14:08.848737 112226 solver.cpp:237]     Train net output #0: loss = 0.020979 (* 1 = 0.020979 loss)
I1231 05:14:08.848744 112226 sgd_solver.cpp:105] Iteration 3700, lr = 0.00789695
I1231 05:14:08.900290 112271 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:08.965936 112226 solver.cpp:218] Iteration 3800 (853.504 iter/s, 0.117164s/100 iters), loss = 0.00585029
I1231 05:14:08.966074 112226 solver.cpp:237]     Train net output #0: loss = 0.00585046 (* 1 = 0.00585046 loss)
I1231 05:14:08.966104 112226 sgd_solver.cpp:105] Iteration 3800, lr = 0.007854
I1231 05:14:09.091514 112226 solver.cpp:218] Iteration 3900 (797.207 iter/s, 0.125438s/100 iters), loss = 0.0287976
I1231 05:14:09.091565 112226 solver.cpp:237]     Train net output #0: loss = 0.0287977 (* 1 = 0.0287977 loss)
I1231 05:14:09.091573 112226 sgd_solver.cpp:105] Iteration 3900, lr = 0.00781158
I1231 05:14:09.204576 112226 solver.cpp:330] Iteration 4000, Testing net (#0)
I1231 05:14:09.255736 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:09.256489 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9894
I1231 05:14:09.256515 112226 solver.cpp:397]     Test net output #1: loss = 0.0315164 (* 1 = 0.0315164 loss)
I1231 05:14:09.257594 112226 solver.cpp:218] Iteration 4000 (602.322 iter/s, 0.166024s/100 iters), loss = 0.020385
I1231 05:14:09.257612 112226 solver.cpp:237]     Train net output #0: loss = 0.0203852 (* 1 = 0.0203852 loss)
I1231 05:14:09.257618 112226 sgd_solver.cpp:105] Iteration 4000, lr = 0.0077697
I1231 05:14:09.371057 112226 solver.cpp:218] Iteration 4100 (881.541 iter/s, 0.113438s/100 iters), loss = 0.0153677
I1231 05:14:09.371074 112226 solver.cpp:237]     Train net output #0: loss = 0.0153679 (* 1 = 0.0153679 loss)
I1231 05:14:09.371078 112226 sgd_solver.cpp:105] Iteration 4100, lr = 0.00772833
I1231 05:14:09.485442 112226 solver.cpp:218] Iteration 4200 (874.46 iter/s, 0.114356s/100 iters), loss = 0.0126334
I1231 05:14:09.485476 112226 solver.cpp:237]     Train net output #0: loss = 0.0126335 (* 1 = 0.0126335 loss)
I1231 05:14:09.485482 112226 sgd_solver.cpp:105] Iteration 4200, lr = 0.00768748
I1231 05:14:09.601124 112226 solver.cpp:218] Iteration 4300 (864.796 iter/s, 0.115634s/100 iters), loss = 0.038657
I1231 05:14:09.601166 112226 solver.cpp:237]     Train net output #0: loss = 0.0386571 (* 1 = 0.0386571 loss)
I1231 05:14:09.601173 112226 sgd_solver.cpp:105] Iteration 4300, lr = 0.00764712
I1231 05:14:09.714756 112226 solver.cpp:218] Iteration 4400 (880.405 iter/s, 0.113584s/100 iters), loss = 0.0174188
I1231 05:14:09.714783 112226 solver.cpp:237]     Train net output #0: loss = 0.017419 (* 1 = 0.017419 loss)
I1231 05:14:09.714790 112226 sgd_solver.cpp:105] Iteration 4400, lr = 0.00760726
I1231 05:14:09.826902 112226 solver.cpp:330] Iteration 4500, Testing net (#0)
I1231 05:14:09.876176 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:09.876894 112226 solver.cpp:397]     Test net output #0: accuracy = 0.988
I1231 05:14:09.876917 112226 solver.cpp:397]     Test net output #1: loss = 0.0381423 (* 1 = 0.0381423 loss)
I1231 05:14:09.877930 112226 solver.cpp:218] Iteration 4500 (612.965 iter/s, 0.163142s/100 iters), loss = 0.00435871
I1231 05:14:09.877979 112226 solver.cpp:237]     Train net output #0: loss = 0.00435888 (* 1 = 0.00435888 loss)
I1231 05:14:09.877987 112226 sgd_solver.cpp:105] Iteration 4500, lr = 0.00756788
I1231 05:14:09.997242 112226 solver.cpp:218] Iteration 4600 (838.528 iter/s, 0.119257s/100 iters), loss = 0.0074584
I1231 05:14:09.997275 112226 solver.cpp:237]     Train net output #0: loss = 0.00745859 (* 1 = 0.00745859 loss)
I1231 05:14:09.997282 112226 sgd_solver.cpp:105] Iteration 4600, lr = 0.00752897
I1231 05:14:10.093076 112271 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:10.112258 112226 solver.cpp:218] Iteration 4700 (869.734 iter/s, 0.114978s/100 iters), loss = 0.00427773
I1231 05:14:10.112284 112226 solver.cpp:237]     Train net output #0: loss = 0.00427795 (* 1 = 0.00427795 loss)
I1231 05:14:10.112290 112226 sgd_solver.cpp:105] Iteration 4700, lr = 0.00749052
I1231 05:14:10.226490 112226 solver.cpp:218] Iteration 4800 (875.668 iter/s, 0.114199s/100 iters), loss = 0.0174331
I1231 05:14:10.226513 112226 solver.cpp:237]     Train net output #0: loss = 0.0174333 (* 1 = 0.0174333 loss)
I1231 05:14:10.226521 112226 sgd_solver.cpp:105] Iteration 4800, lr = 0.00745253
I1231 05:14:10.341311 112226 solver.cpp:218] Iteration 4900 (871.145 iter/s, 0.114791s/100 iters), loss = 0.00480582
I1231 05:14:10.341336 112226 solver.cpp:237]     Train net output #0: loss = 0.00480605 (* 1 = 0.00480605 loss)
I1231 05:14:10.341343 112226 sgd_solver.cpp:105] Iteration 4900, lr = 0.00741498
I1231 05:14:10.453488 112226 solver.cpp:447] Snapshotting to binary proto file /home/zhaoys/myf/lenet/lenet_caffe/snapshot/train_iter_5000.caffemodel
I1231 05:14:10.461127 112226 sgd_solver.cpp:273] Snapshotting solver state to binary proto file /home/zhaoys/myf/lenet/lenet_caffe/snapshot/train_iter_5000.solverstate
I1231 05:14:10.464032 112226 solver.cpp:330] Iteration 5000, Testing net (#0)
I1231 05:14:10.510946 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:10.511458 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9897
I1231 05:14:10.511488 112226 solver.cpp:397]     Test net output #1: loss = 0.0317376 (* 1 = 0.0317376 loss)
I1231 05:14:10.512598 112226 solver.cpp:218] Iteration 5000 (583.925 iter/s, 0.171255s/100 iters), loss = 0.0539401
I1231 05:14:10.512640 112226 solver.cpp:237]     Train net output #0: loss = 0.0539403 (* 1 = 0.0539403 loss)
I1231 05:14:10.512650 112226 sgd_solver.cpp:105] Iteration 5000, lr = 0.00737788
I1231 05:14:10.626550 112226 solver.cpp:218] Iteration 5100 (877.929 iter/s, 0.113904s/100 iters), loss = 0.0210981
I1231 05:14:10.626577 112226 solver.cpp:237]     Train net output #0: loss = 0.0210984 (* 1 = 0.0210984 loss)
I1231 05:14:10.626585 112226 sgd_solver.cpp:105] Iteration 5100, lr = 0.0073412
I1231 05:14:10.740978 112226 solver.cpp:218] Iteration 5200 (874.155 iter/s, 0.114396s/100 iters), loss = 0.00715472
I1231 05:14:10.741003 112226 solver.cpp:237]     Train net output #0: loss = 0.00715494 (* 1 = 0.00715494 loss)
I1231 05:14:10.741010 112226 sgd_solver.cpp:105] Iteration 5200, lr = 0.00730495
I1231 05:14:10.855633 112226 solver.cpp:218] Iteration 5300 (872.461 iter/s, 0.114618s/100 iters), loss = 0.00303569
I1231 05:14:10.855677 112226 solver.cpp:237]     Train net output #0: loss = 0.00303591 (* 1 = 0.00303591 loss)
I1231 05:14:10.855684 112226 sgd_solver.cpp:105] Iteration 5300, lr = 0.00726911
I1231 05:14:10.969705 112226 solver.cpp:218] Iteration 5400 (877.013 iter/s, 0.114023s/100 iters), loss = 0.00846723
I1231 05:14:10.969732 112226 solver.cpp:237]     Train net output #0: loss = 0.00846745 (* 1 = 0.00846745 loss)
I1231 05:14:10.969738 112226 sgd_solver.cpp:105] Iteration 5400, lr = 0.00723368
I1231 05:14:11.082397 112226 solver.cpp:330] Iteration 5500, Testing net (#0)
I1231 05:14:11.130837 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:11.131626 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9891
I1231 05:14:11.131644 112226 solver.cpp:397]     Test net output #1: loss = 0.0324316 (* 1 = 0.0324316 loss)
I1231 05:14:11.132758 112226 solver.cpp:218] Iteration 5500 (613.415 iter/s, 0.163022s/100 iters), loss = 0.0167479
I1231 05:14:11.132776 112226 solver.cpp:237]     Train net output #0: loss = 0.0167481 (* 1 = 0.0167481 loss)
I1231 05:14:11.132781 112226 sgd_solver.cpp:105] Iteration 5500, lr = 0.00719865
I1231 05:14:11.246398 112226 solver.cpp:218] Iteration 5600 (880.166 iter/s, 0.113615s/100 iters), loss = 0.000499987
I1231 05:14:11.246423 112226 solver.cpp:237]     Train net output #0: loss = 0.00050021 (* 1 = 0.00050021 loss)
I1231 05:14:11.246464 112226 sgd_solver.cpp:105] Iteration 5600, lr = 0.00716402
I1231 05:14:11.269551 112271 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:11.359910 112226 solver.cpp:218] Iteration 5700 (881.2 iter/s, 0.113482s/100 iters), loss = 0.00245953
I1231 05:14:11.359935 112226 solver.cpp:237]     Train net output #0: loss = 0.00245974 (* 1 = 0.00245974 loss)
I1231 05:14:11.359942 112226 sgd_solver.cpp:105] Iteration 5700, lr = 0.00712977
I1231 05:14:11.477833 112226 solver.cpp:218] Iteration 5800 (848.263 iter/s, 0.117888s/100 iters), loss = 0.0389832
I1231 05:14:11.477866 112226 solver.cpp:237]     Train net output #0: loss = 0.0389835 (* 1 = 0.0389835 loss)
I1231 05:14:11.477871 112226 sgd_solver.cpp:105] Iteration 5800, lr = 0.0070959
I1231 05:14:11.593122 112226 solver.cpp:218] Iteration 5900 (867.754 iter/s, 0.11524s/100 iters), loss = 0.00710046
I1231 05:14:11.593165 112226 solver.cpp:237]     Train net output #0: loss = 0.00710067 (* 1 = 0.00710067 loss)
I1231 05:14:11.593173 112226 sgd_solver.cpp:105] Iteration 5900, lr = 0.0070624
I1231 05:14:11.707418 112226 solver.cpp:330] Iteration 6000, Testing net (#0)
I1231 05:14:11.758401 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:11.759146 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9907
I1231 05:14:11.759176 112226 solver.cpp:397]     Test net output #1: loss = 0.0290163 (* 1 = 0.0290163 loss)
I1231 05:14:11.760239 112226 solver.cpp:218] Iteration 6000 (598.555 iter/s, 0.167069s/100 iters), loss = 0.00378046
I1231 05:14:11.760258 112226 solver.cpp:237]     Train net output #0: loss = 0.00378066 (* 1 = 0.00378066 loss)
I1231 05:14:11.760263 112226 sgd_solver.cpp:105] Iteration 6000, lr = 0.00702927
I1231 05:14:11.872548 112226 solver.cpp:218] Iteration 6100 (890.593 iter/s, 0.112285s/100 iters), loss = 0.00106458
I1231 05:14:11.872566 112226 solver.cpp:237]     Train net output #0: loss = 0.00106477 (* 1 = 0.00106477 loss)
I1231 05:14:11.872571 112226 sgd_solver.cpp:105] Iteration 6100, lr = 0.0069965
I1231 05:14:11.989754 112226 solver.cpp:218] Iteration 6200 (853.497 iter/s, 0.117165s/100 iters), loss = 0.014281
I1231 05:14:11.989837 112226 solver.cpp:237]     Train net output #0: loss = 0.0142812 (* 1 = 0.0142812 loss)
I1231 05:14:11.989850 112226 sgd_solver.cpp:105] Iteration 6200, lr = 0.00696408
I1231 05:14:12.109879 112226 solver.cpp:218] Iteration 6300 (833.088 iter/s, 0.120035s/100 iters), loss = 0.00767381
I1231 05:14:12.109920 112226 solver.cpp:237]     Train net output #0: loss = 0.00767398 (* 1 = 0.00767398 loss)
I1231 05:14:12.109926 112226 sgd_solver.cpp:105] Iteration 6300, lr = 0.00693201
I1231 05:14:12.223362 112226 solver.cpp:218] Iteration 6400 (881.605 iter/s, 0.11343s/100 iters), loss = 0.00609869
I1231 05:14:12.223404 112226 solver.cpp:237]     Train net output #0: loss = 0.00609886 (* 1 = 0.00609886 loss)
I1231 05:14:12.223414 112226 sgd_solver.cpp:105] Iteration 6400, lr = 0.00690029
I1231 05:14:12.335515 112226 solver.cpp:330] Iteration 6500, Testing net (#0)
I1231 05:14:12.386811 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:12.387568 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9906
I1231 05:14:12.387588 112226 solver.cpp:397]     Test net output #1: loss = 0.0315652 (* 1 = 0.0315652 loss)
I1231 05:14:12.388682 112226 solver.cpp:218] Iteration 6500 (605.06 iter/s, 0.165273s/100 iters), loss = 0.0120906
I1231 05:14:12.388700 112226 solver.cpp:237]     Train net output #0: loss = 0.0120908 (* 1 = 0.0120908 loss)
I1231 05:14:12.388705 112226 sgd_solver.cpp:105] Iteration 6500, lr = 0.0068689
I1231 05:14:12.455711 112271 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:12.503032 112226 solver.cpp:218] Iteration 6600 (874.71 iter/s, 0.114324s/100 iters), loss = 0.0408356
I1231 05:14:12.503060 112226 solver.cpp:237]     Train net output #0: loss = 0.0408357 (* 1 = 0.0408357 loss)
I1231 05:14:12.503067 112226 sgd_solver.cpp:105] Iteration 6600, lr = 0.00683784
I1231 05:14:12.616698 112226 solver.cpp:218] Iteration 6700 (880.024 iter/s, 0.113633s/100 iters), loss = 0.0120867
I1231 05:14:12.616721 112226 solver.cpp:237]     Train net output #0: loss = 0.0120869 (* 1 = 0.0120869 loss)
I1231 05:14:12.616727 112226 sgd_solver.cpp:105] Iteration 6700, lr = 0.00680711
I1231 05:14:12.730803 112226 solver.cpp:218] Iteration 6800 (876.611 iter/s, 0.114076s/100 iters), loss = 0.00273887
I1231 05:14:12.730825 112226 solver.cpp:237]     Train net output #0: loss = 0.00273904 (* 1 = 0.00273904 loss)
I1231 05:14:12.730832 112226 sgd_solver.cpp:105] Iteration 6800, lr = 0.0067767
I1231 05:14:12.848994 112226 solver.cpp:218] Iteration 6900 (846.324 iter/s, 0.118158s/100 iters), loss = 0.0063858
I1231 05:14:12.849032 112226 solver.cpp:237]     Train net output #0: loss = 0.00638596 (* 1 = 0.00638596 loss)
I1231 05:14:12.849041 112226 sgd_solver.cpp:105] Iteration 6900, lr = 0.0067466
I1231 05:14:12.968724 112226 solver.cpp:330] Iteration 7000, Testing net (#0)
I1231 05:14:13.027690 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:13.028416 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9914
I1231 05:14:13.028482 112226 solver.cpp:397]     Test net output #1: loss = 0.0280982 (* 1 = 0.0280982 loss)
I1231 05:14:13.029805 112226 solver.cpp:218] Iteration 7000 (553.213 iter/s, 0.180762s/100 iters), loss = 0.00817811
I1231 05:14:13.029870 112226 solver.cpp:237]     Train net output #0: loss = 0.00817828 (* 1 = 0.00817828 loss)
I1231 05:14:13.029883 112226 sgd_solver.cpp:105] Iteration 7000, lr = 0.00671681
I1231 05:14:13.155633 112226 solver.cpp:218] Iteration 7100 (795.18 iter/s, 0.125758s/100 iters), loss = 0.00938367
I1231 05:14:13.155678 112226 solver.cpp:237]     Train net output #0: loss = 0.00938383 (* 1 = 0.00938383 loss)
I1231 05:14:13.155695 112226 sgd_solver.cpp:105] Iteration 7100, lr = 0.00668733
I1231 05:14:13.271452 112226 solver.cpp:218] Iteration 7200 (863.825 iter/s, 0.115764s/100 iters), loss = 0.00807085
I1231 05:14:13.271486 112226 solver.cpp:237]     Train net output #0: loss = 0.00807101 (* 1 = 0.00807101 loss)
I1231 05:14:13.271493 112226 sgd_solver.cpp:105] Iteration 7200, lr = 0.00665815
I1231 05:14:13.387769 112226 solver.cpp:218] Iteration 7300 (860.097 iter/s, 0.116266s/100 iters), loss = 0.0224124
I1231 05:14:13.387817 112226 solver.cpp:237]     Train net output #0: loss = 0.0224126 (* 1 = 0.0224126 loss)
I1231 05:14:13.387827 112226 sgd_solver.cpp:105] Iteration 7300, lr = 0.00662927
I1231 05:14:13.505359 112226 solver.cpp:218] Iteration 7400 (850.806 iter/s, 0.117536s/100 iters), loss = 0.00614564
I1231 05:14:13.505391 112226 solver.cpp:237]     Train net output #0: loss = 0.00614582 (* 1 = 0.00614582 loss)
I1231 05:14:13.505398 112226 sgd_solver.cpp:105] Iteration 7400, lr = 0.00660067
I1231 05:14:13.616973 112271 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:13.620801 112226 solver.cpp:330] Iteration 7500, Testing net (#0)
I1231 05:14:13.672514 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:13.673836 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9897
I1231 05:14:13.673854 112226 solver.cpp:397]     Test net output #1: loss = 0.0328055 (* 1 = 0.0328055 loss)
I1231 05:14:13.675462 112226 solver.cpp:218] Iteration 7500 (589.968 iter/s, 0.169501s/100 iters), loss = 0.00213244
I1231 05:14:13.675484 112226 solver.cpp:237]     Train net output #0: loss = 0.00213262 (* 1 = 0.00213262 loss)
I1231 05:14:13.675490 112226 sgd_solver.cpp:105] Iteration 7500, lr = 0.00657236
I1231 05:14:13.792430 112226 solver.cpp:218] Iteration 7600 (855.208 iter/s, 0.116931s/100 iters), loss = 0.00654828
I1231 05:14:13.792476 112226 solver.cpp:237]     Train net output #0: loss = 0.00654845 (* 1 = 0.00654845 loss)
I1231 05:14:13.792485 112226 sgd_solver.cpp:105] Iteration 7600, lr = 0.00654433
I1231 05:14:13.909584 112226 solver.cpp:218] Iteration 7700 (853.959 iter/s, 0.117102s/100 iters), loss = 0.0209833
I1231 05:14:13.909611 112226 solver.cpp:237]     Train net output #0: loss = 0.0209835 (* 1 = 0.0209835 loss)
I1231 05:14:13.909656 112226 sgd_solver.cpp:105] Iteration 7700, lr = 0.00651658
I1231 05:14:14.032776 112226 solver.cpp:218] Iteration 7800 (812.002 iter/s, 0.123152s/100 iters), loss = 0.00290527
I1231 05:14:14.032826 112226 solver.cpp:237]     Train net output #0: loss = 0.00290546 (* 1 = 0.00290546 loss)
I1231 05:14:14.032837 112226 sgd_solver.cpp:105] Iteration 7800, lr = 0.00648911
I1231 05:14:14.150871 112226 solver.cpp:218] Iteration 7900 (847.181 iter/s, 0.118039s/100 iters), loss = 0.0050287
I1231 05:14:14.150908 112226 solver.cpp:237]     Train net output #0: loss = 0.00502888 (* 1 = 0.00502888 loss)
I1231 05:14:14.150919 112226 sgd_solver.cpp:105] Iteration 7900, lr = 0.0064619
I1231 05:14:14.266319 112226 solver.cpp:330] Iteration 8000, Testing net (#0)
I1231 05:14:14.313983 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:14.315620 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9915
I1231 05:14:14.315639 112226 solver.cpp:397]     Test net output #1: loss = 0.0284429 (* 1 = 0.0284429 loss)
I1231 05:14:14.316642 112226 solver.cpp:218] Iteration 8000 (603.39 iter/s, 0.16573s/100 iters), loss = 0.00472042
I1231 05:14:14.316660 112226 solver.cpp:237]     Train net output #0: loss = 0.0047206 (* 1 = 0.0047206 loss)
I1231 05:14:14.316665 112226 sgd_solver.cpp:105] Iteration 8000, lr = 0.00643496
I1231 05:14:14.432391 112226 solver.cpp:218] Iteration 8100 (864.125 iter/s, 0.115724s/100 iters), loss = 0.0119718
I1231 05:14:14.432417 112226 solver.cpp:237]     Train net output #0: loss = 0.0119719 (* 1 = 0.0119719 loss)
I1231 05:14:14.432425 112226 sgd_solver.cpp:105] Iteration 8100, lr = 0.00640827
I1231 05:14:14.548878 112226 solver.cpp:218] Iteration 8200 (858.704 iter/s, 0.116455s/100 iters), loss = 0.0115943
I1231 05:14:14.548904 112226 solver.cpp:237]     Train net output #0: loss = 0.0115945 (* 1 = 0.0115945 loss)
I1231 05:14:14.548912 112226 sgd_solver.cpp:105] Iteration 8200, lr = 0.00638185
I1231 05:14:14.663906 112226 solver.cpp:218] Iteration 8300 (869.602 iter/s, 0.114995s/100 iters), loss = 0.0245502
I1231 05:14:14.663933 112226 solver.cpp:237]     Train net output #0: loss = 0.0245504 (* 1 = 0.0245504 loss)
I1231 05:14:14.663938 112226 sgd_solver.cpp:105] Iteration 8300, lr = 0.00635567
I1231 05:14:14.779588 112226 solver.cpp:218] Iteration 8400 (864.683 iter/s, 0.115649s/100 iters), loss = 0.00964116
I1231 05:14:14.779618 112226 solver.cpp:237]     Train net output #0: loss = 0.00964134 (* 1 = 0.00964134 loss)
I1231 05:14:14.779628 112226 sgd_solver.cpp:105] Iteration 8400, lr = 0.00632975
I1231 05:14:14.818081 112271 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:14.893966 112226 solver.cpp:330] Iteration 8500, Testing net (#0)
I1231 05:14:14.946905 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:14.947444 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9897
I1231 05:14:14.947468 112226 solver.cpp:397]     Test net output #1: loss = 0.0298499 (* 1 = 0.0298499 loss)
I1231 05:14:14.948550 112226 solver.cpp:218] Iteration 8500 (591.983 iter/s, 0.168924s/100 iters), loss = 0.00518011
I1231 05:14:14.948603 112226 solver.cpp:237]     Train net output #0: loss = 0.00518029 (* 1 = 0.00518029 loss)
I1231 05:14:14.948618 112226 sgd_solver.cpp:105] Iteration 8500, lr = 0.00630407
I1231 05:14:15.067173 112226 solver.cpp:218] Iteration 8600 (843.458 iter/s, 0.11856s/100 iters), loss = 0.00088433
I1231 05:14:15.067211 112226 solver.cpp:237]     Train net output #0: loss = 0.000884513 (* 1 = 0.000884513 loss)
I1231 05:14:15.067220 112226 sgd_solver.cpp:105] Iteration 8600, lr = 0.00627864
I1231 05:14:15.178897 112226 solver.cpp:218] Iteration 8700 (895.435 iter/s, 0.111678s/100 iters), loss = 0.0020867
I1231 05:14:15.178933 112226 solver.cpp:237]     Train net output #0: loss = 0.00208688 (* 1 = 0.00208688 loss)
I1231 05:14:15.178941 112226 sgd_solver.cpp:105] Iteration 8700, lr = 0.00625344
I1231 05:14:15.291426 112226 solver.cpp:218] Iteration 8800 (889.028 iter/s, 0.112482s/100 iters), loss = 0.00102185
I1231 05:14:15.291489 112226 solver.cpp:237]     Train net output #0: loss = 0.00102203 (* 1 = 0.00102203 loss)
I1231 05:14:15.291497 112226 sgd_solver.cpp:105] Iteration 8800, lr = 0.00622847
I1231 05:14:15.403802 112226 solver.cpp:218] Iteration 8900 (890.432 iter/s, 0.112305s/100 iters), loss = 0.0014528
I1231 05:14:15.403837 112226 solver.cpp:237]     Train net output #0: loss = 0.00145299 (* 1 = 0.00145299 loss)
I1231 05:14:15.403843 112226 sgd_solver.cpp:105] Iteration 8900, lr = 0.00620374
I1231 05:14:15.515713 112226 solver.cpp:330] Iteration 9000, Testing net (#0)
I1231 05:14:15.563016 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:15.563760 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9904
I1231 05:14:15.563781 112226 solver.cpp:397]     Test net output #1: loss = 0.0289717 (* 1 = 0.0289717 loss)
I1231 05:14:15.564855 112226 solver.cpp:218] Iteration 9000 (621.069 iter/s, 0.161013s/100 iters), loss = 0.0110505
I1231 05:14:15.564873 112226 solver.cpp:237]     Train net output #0: loss = 0.0110507 (* 1 = 0.0110507 loss)
I1231 05:14:15.564880 112226 sgd_solver.cpp:105] Iteration 9000, lr = 0.00617924
I1231 05:14:15.676730 112226 solver.cpp:218] Iteration 9100 (894.083 iter/s, 0.111846s/100 iters), loss = 0.00823897
I1231 05:14:15.676772 112226 solver.cpp:237]     Train net output #0: loss = 0.00823915 (* 1 = 0.00823915 loss)
I1231 05:14:15.676781 112226 sgd_solver.cpp:105] Iteration 9100, lr = 0.00615496
I1231 05:14:15.789703 112226 solver.cpp:218] Iteration 9200 (885.554 iter/s, 0.112924s/100 iters), loss = 0.00273166
I1231 05:14:15.789738 112226 solver.cpp:237]     Train net output #0: loss = 0.00273185 (* 1 = 0.00273185 loss)
I1231 05:14:15.789746 112226 sgd_solver.cpp:105] Iteration 9200, lr = 0.0061309
I1231 05:14:15.902246 112226 solver.cpp:218] Iteration 9300 (888.907 iter/s, 0.112498s/100 iters), loss = 0.00619383
I1231 05:14:15.902281 112226 solver.cpp:237]     Train net output #0: loss = 0.00619402 (* 1 = 0.00619402 loss)
I1231 05:14:15.902290 112226 sgd_solver.cpp:105] Iteration 9300, lr = 0.00610706
I1231 05:14:15.983361 112271 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:16.016964 112226 solver.cpp:218] Iteration 9400 (872.007 iter/s, 0.114678s/100 iters), loss = 0.0341899
I1231 05:14:16.016996 112226 solver.cpp:237]     Train net output #0: loss = 0.0341901 (* 1 = 0.0341901 loss)
I1231 05:14:16.017004 112226 sgd_solver.cpp:105] Iteration 9400, lr = 0.00608343
I1231 05:14:16.127959 112226 solver.cpp:330] Iteration 9500, Testing net (#0)
I1231 05:14:16.173899 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:16.174634 112226 solver.cpp:397]     Test net output #0: accuracy = 0.989
I1231 05:14:16.174654 112226 solver.cpp:397]     Test net output #1: loss = 0.0339525 (* 1 = 0.0339525 loss)
I1231 05:14:16.175726 112226 solver.cpp:218] Iteration 9500 (630.021 iter/s, 0.158725s/100 iters), loss = 0.00653742
I1231 05:14:16.175745 112226 solver.cpp:237]     Train net output #0: loss = 0.00653761 (* 1 = 0.00653761 loss)
I1231 05:14:16.175751 112226 sgd_solver.cpp:105] Iteration 9500, lr = 0.00606002
I1231 05:14:16.291050 112226 solver.cpp:218] Iteration 9600 (867.354 iter/s, 0.115293s/100 iters), loss = 0.00260782
I1231 05:14:16.291090 112226 solver.cpp:237]     Train net output #0: loss = 0.00260802 (* 1 = 0.00260802 loss)
I1231 05:14:16.291101 112226 sgd_solver.cpp:105] Iteration 9600, lr = 0.00603682
I1231 05:14:16.405395 112226 solver.cpp:218] Iteration 9700 (874.93 iter/s, 0.114295s/100 iters), loss = 0.00365551
I1231 05:14:16.405433 112226 solver.cpp:237]     Train net output #0: loss = 0.0036557 (* 1 = 0.0036557 loss)
I1231 05:14:16.405441 112226 sgd_solver.cpp:105] Iteration 9700, lr = 0.00601382
I1231 05:14:16.520783 112226 solver.cpp:218] Iteration 9800 (866.969 iter/s, 0.115344s/100 iters), loss = 0.013783
I1231 05:14:16.520808 112226 solver.cpp:237]     Train net output #0: loss = 0.0137832 (* 1 = 0.0137832 loss)
I1231 05:14:16.520841 112226 sgd_solver.cpp:105] Iteration 9800, lr = 0.00599102
I1231 05:14:16.636044 112226 solver.cpp:218] Iteration 9900 (867.875 iter/s, 0.115224s/100 iters), loss = 0.00643541
I1231 05:14:16.636071 112226 solver.cpp:237]     Train net output #0: loss = 0.00643561 (* 1 = 0.00643561 loss)
I1231 05:14:16.636078 112226 sgd_solver.cpp:105] Iteration 9900, lr = 0.00596843
I1231 05:14:16.748425 112226 solver.cpp:447] Snapshotting to binary proto file /home/zhaoys/myf/lenet/lenet_caffe/snapshot/train_iter_10000.caffemodel
I1231 05:14:16.754078 112226 sgd_solver.cpp:273] Snapshotting solver state to binary proto file /home/zhaoys/myf/lenet/lenet_caffe/snapshot/train_iter_10000.solverstate
I1231 05:14:16.757302 112226 solver.cpp:310] Iteration 10000, loss = 0.00403373
I1231 05:14:16.757325 112226 solver.cpp:330] Iteration 10000, Testing net (#0)
I1231 05:14:16.807796 112273 data_layer.cpp:73] Restarting data prefetching from start.
I1231 05:14:16.808288 112226 solver.cpp:397]     Test net output #0: accuracy = 0.9913
I1231 05:14:16.808310 112226 solver.cpp:397]     Test net output #1: loss = 0.0287575 (* 1 = 0.0287575 loss)
I1231 05:14:16.808320 112226 solver.cpp:315] Optimization Done.
I1231 05:14:16.808333 112226 caffe.cpp:259] Optimization Done.

 

三少Algorithm
关注
专栏目录
caffe下lenet实现（python脚本）
02-12
import os import sys import caffe from pylab import * ###pylab 是用来...from caffe import layers as L,params as P ###layers里面包含了Caffe所有内置的层（比如卷积，ReLU等） ；params则包含了各种枚举值###
LeNet简介以及Caffe实现
yuexingyu的博客
05-16 2732
LeNet是一个用来识别手写数字的最经典的卷积神经网络，是Yann LeCun在1998年设计并提出的，是早期卷积神经网络中最有代表性的实验系统之一。上图是LeNet网络模型框架，总共有8层，C层代表卷积层，S层代表下采样层。模型理解：第一层：输入层图像大小是32*32（Caffe中Mnist数据库为28*28）。第二层：C1层是卷积层，卷积核大小为5*5，总共有6个卷积核，通过卷积操作得到6个特...
LeNet网络详解及pytorch实现
Serins的博客
11-22 3014
博主是在校大学生，学习的也是深度学习方面的专业，入坑也有快五个月了，想通过这个方式记录自己学的东西，也方便随时回来巩固，如果有什么不对的地方请大家提出来一起学习，谢谢大家???????? 1.LeNet介绍 LeNet神经网络由深度学习三巨头之一的Yan LeCun提出，他同时也是卷积神经网络 (CNN，Convolutional Neural Networks)之父。LeNet主要用来进行手写字符的识别与分类，并在美国的银行中投入了使用。LeNet的实现确立了CNN的结构，现在神经网络中的许多内容在LeN
【深度学习】LeNet原理及代码实现
最新发布
言兼小师傅的博客
07-09 469
【代码】LeNet原理及代码实现。
Caffe入门：LeNet模型介绍与网络训练测试例子
m0_37357063的博客
11-20 2434
学习卷积神经网络，首先需要理解卷积的概念，二维离散卷积的概念，以及卷积核在图像中进行卷积操作得出结果图的实际直观含义。 卷积的实际计算方式：二维卷积更正式的名字是块循环矩阵（double block circulant matrix)，矩阵论，Toeplitz矩阵 caffe中，默认的矩阵计算实现都是基于矩阵乘法的。 就是把卷积核在图像中的滑动通过矩阵每一行元素位置的移动来实现的。 深度神...
caffe模型（1）：LeNet模型讲解
qq_39501363的博客
07-31 536
引言： LeNet作为入门级caffe网络模型，是用来识别手写数字的最经典的卷积神经网络，Yann LeCun于1998年设计提出，手写数字识别系统LeNet-5在20世纪90年代广泛应用于银行手写数字识别。 1.模型简介 如图所示，LeNet-5卷积神经网络整体框架较小，包含了卷积层、池化层、全连接层。包含输入输出层在内一共8层，每一层包含多个参数（权重），C层为卷积层...
CNN经典网络模型（一）：LeNet简介及代码实现（PyTorch超详细注释版）
热门推荐
qq_43307074的博客
10-01 1万+
CNN经典网络模型之一：LeNet，本文包含其简介及代码，在PyTorch中实现，进行超详细注释，适合新生小白阅读学习~
基于caffe lenet 中文车牌字符识别实现
03-25
在本文中，我们将深入探讨如何使用Caffe框架和LeNet模型来实现中文车牌字符的识别。Caffe是一种广泛使用的深度学习框架，以其高效、简洁和灵活性而受到开发者的青睐。LeNet则是Yann LeCun在1998年提出的经典卷积神经...
python_caffe_lenet5
10-28
本项目“python_caffe_lenet5”是针对深度学习初学者的一个实践教程，旨在帮助用户从零开始理解经典的LeNet-5神经网络模型，并通过Caffe框架和Python进行实现。Caffe是一款高效、灵活的深度学习框架，广泛应用于学术...
lenet_iter_10000.zip_caffe_caffe mnist_caffe 识别_lenet mnist_mnis
09-24
"lenet_mnist"特指了这个模型是LeNet架构，它是卷积神经网络（CNN）的经典实现之一，特别适合处理像MNIST这样的小规模图像分类问题。 压缩包内的"lenet_iter_10000.caffemodel"文件是训练完成后保存的模型权重和...
快速安装caffe步骤
12-21
我在win7虚拟机里面装的linux，用xshell用具链接linux，xftp实现win7和linux文件的传输。 3 可以使用 docker import 从容器快照文件中再导入为镜像，例如 cat ubuntu_caffe.tar | sudo docker import - test/ubuntu_...
lenet.caffemodel
08-12
lenet.caffemodel,非常好用。已测试成功！lenet.caffemodel,非常好用。已测试成功！
lenet mnist 数字10000次cnn结果 (.caffemodel)
12-18
lenet_iter_10000.caffemodel lenet_iter_5000.caffemodel
Caffe MNIST训练好的lenet模型
12-03
Caffe MNIST例子中训练好的LeNet模型，手写体数字识别准确率99%。
caffe总结（五）经典神经网络模型（LeNet , AlexNet , VGG , GoogLeNet）简介
weixin_42535423的博客
04-29 933
caffe总结（五） 一、LeNet 二、AlexNet ReLU 数据扩充 重叠池化 局部归一化（Local Response Normalization，简称LRN） Dropout 三、VGG 四、GoogLeNet 1、Inception V1 2、Inception V2 3、Inception V3 4、Inception V4 本文主要学习总结了关于经典网络模型（LeNet , AlexNet , VGG , GoogLeNet）的简单介绍。 一、LeNet C1层（卷积层）：6@28×28
深度学习——LeNet的实现（笔记）
jbkjhji的博客
11-30 904
先用卷积层来学习图片的空间信息，用池化层降低图片的敏感度。最后用全连接层转换为类别的空间，转换为10类。1.LeNet是用来分辨手写体的数字的，分辨出来是哪个数字的可能性。1.构建神经网络，加入激活函数是为了非线性。4.使用GPU计算模型在数据集上的精度。③然后使用全连接层来转换到类别空间。①LeNet是早期成功的神经网络。②先使用卷积层来学习图片空间信息。用到的数据集是MNIST。
【00】LeNet的发展及应用
xfsong2012的博客
10-09 2187
LeNet是Yann LeCun在1998年的发表的论文[1]提出来的第一个卷积神经网络，最早是用于识别手写数字和印刷字符。由于模型比较简单，所以方便新手入门学习。共有3个模型组成：LeNet-1：5 层模型，一个简单的 CNN。LeNet-4：6 层模型，是 LeNet-1 的改进版本。LeNet-5：7 层模型，最著名的版本。（最常用）其中LeNet是以LeCun的名字命名，后面的数字代表研究成果的代号。
深度学习经典网络模型——LeNet
m0_52819281的博客
06-21 2952
深度学习经典网络模型——LeNet
Lenet5实现及代码详解——以MINST数据集为例
qq_43569111的博客
10-18 3770
看了卷积神经网络（CNN）的原理及介绍，想着自己动手解决一个案例，在网上也看了很多博客，这里整理一下，顺便记录一下自己解决一个完成的CNN实例的过程，以便以后方便看。 如果有不足之处，欢迎大家指正。 数据获取与可视化操作 1、下载minst数据集 在网上可以很容易找到minst数据集，这里为了方便大家，直接给出链接，大家直接点击下载即可.。注意下载后不要解压缩。 2、数据的读取 利用python的...
Caffe深度学习教程：从LeNet到AlexNet的历史与实践
其中，历史部分回顾了深度学习的发展历程，从1989年的卷积神经网络（如LeNet，一个由卷积层、下采样操作及用于手写数字识别的分类器构成的基础模型）到2012年AlexNet的诞生，后者是ImageNet Large Scale Visual ...
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值