【Caffe】caffe可视化训练过程实操

最新推荐文章于 2022-12-14 18:27:41 发布
最新推荐文章于 2022-12-14 18:27:41 发布
阅读量762 收藏 1
点赞数
分类专栏： 深度学习
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/Taily_Duan/article/details/85329476
版权

http://ethereon.github.io/netscope/#/editor 这个可以可视化caffemodel模型

使用lenet网络做训练，将两个loss曲线一个accuracy曲线画在一个图上

#!/usr/bin/python
#coding:utf-8
import os
import sys
import numpy as np
import matplotlib.pyplot as plt

caffe_root = '/usr/local/Cellar/caffe/'

sys.path.insert(0, caffe_root + 'python')
import caffe
# MODEL_FILE = caffe_root+'examples/mnist/lenet.prototxt'
# PRETRAINED = caffe_root+'examples/mnist/lenet_iter_10000.caffemodel'
# IMAGE_FILE = caffe_root+'examples/images/test4.bmp'

import matplotlib.pyplot as plt
from numpy import zeros, arange
from math import ceil
import caffe

#caffe.set_device(0)
caffe.set_mode_cpu()
# 使用SGDsolver,即随机梯度下降算法,这个要看你solver文件里,一般不写的话就是sgd
# solver=caffe.AdamSolver('/root/caffe/examples/mnist/lenet_solver_adam.prototxt')
solver = caffe.SGDSolver(caffe_root + 'examples/mnist/lenet_solver_test_show.prototxt')

# 等价于solver文件中的max_iter，即最大解算次数
niter = 10000
# 每隔100次收集一次数据
display = 100

# 每次测试进行100次解算，10000/100
test_iter = 100
# 每500次训练进行一次测试（100次解算），60000/64
test_interval = 500

# 初始化
train_loss = zeros(int(niter * 1.0 / display))
test_loss = zeros(int(niter * 1.0 / test_interval))
test_acc = zeros(int(niter * 1.0 / test_interval))

# iteration 0，不计入
solver.step(1)

# 辅助变量
_train_loss = 0;
_test_loss = 0;
_accuracy = 0
# 进行解算
for it in range(niter):
    # 进行一次解算
    solver.step(1)
    # 每迭代一次，训练batch_size张图片
    _train_loss += solver.net.blobs['loss'].data  # 注意，这里的loss表示你定义网络中loss层使用的名称，原博客中定义该网络使用的是SoftmaxWithLoss

    if it % display == 0:
        # 计算平均train loss
        train_loss[it // display] = _train_loss / display
        _train_loss = 0

    if it % test_interval == 0:
        for test_it in range(test_iter):
            # 进行一次测试
            solver.test_nets[0].forward()
            # 计算test loss
            _test_loss += solver.test_nets[0].blobs['loss'].data  # loss名称和上面的一样
            # 计算test accuracy
            _accuracy += solver.test_nets[0].blobs['accuracy'].data  # 这里和上面一样需要注意一下

        # 计算平均test loss
        test_loss[it / test_interval] = _test_loss / test_iter
        # 计算平均test accuracy
        test_acc[it / test_interval] = _accuracy / test_iter
        _test_loss = 0
        _accuracy = 0

    # 绘制train loss、test loss和accuracy曲线
print '\nplot the train loss and test accuracy\n'
_, ax1 = plt.subplots()
ax2 = ax1.twinx()

# train loss -> 绿色
ax1.plot(display * arange(len(train_loss)), train_loss, 'g')
# test loss -> 黄色
ax1.plot(test_interval * arange(len(test_loss)), test_loss, 'y')
# test accuracy -> 红色
ax2.plot(test_interval * arange(len(test_acc)), test_acc, 'r')

ax1.set_xlabel('iteration')
ax1.set_ylabel('loss')
ax2.set_ylabel('accuracy')
plt.show()

#coding:utf-8

添加后支持中文字符；

caffe_root = '/usr/local/Cellar/caffe/'

配置自己的caffe路径；

solver = caffe.SGDSolver(caffe_root + 'examples/mnist/lenet_solver_test_show.prototxt')

配置自己的solver文件路径；（注：其中的net和snapshot_prefix建议配置绝对路径）

net: "/usr/local/Cellar/caffe/examples/mnist/lenet_train_test.prototxt"

snapshot_prefix: "/Users/taily/pycharmproj/"

训练完成后显示训练过程的loss曲线和accuracy曲线：

绿色——训练数据上的loss

黄色——测试数据上的loss

红色——准确率accuracy（近1000次迭代后准确率就接近1了，后续迭代缓慢提高了accurary）

训练过程中打印的log：

/anaconda2/envs/py27/bin/python /Users/taily/pycharmproj/test-caffe-show/test_digit.py
WARNING: Logging before InitGoogleLogging() is written to STDERR
I1228 17:55:50.809736 2733998976 upgrade_proto.cpp:1113] snapshot_prefix was a directory and is replaced to /Users/taily/pycharmproj/lenet_solver_test_show
I1228 17:55:50.809761 2733998976 solver.cpp:45] 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: "/Users/taily/pycharmproj/lenet_solver_test_show"
solver_mode: CPU
net: "/usr/local/Cellar/caffe/examples/mnist/lenet_train_test.prototxt"
I1228 17:55:50.809949 2733998976 solver.cpp:102] Creating training net from net file: /usr/local/Cellar/caffe/examples/mnist/lenet_train_test.prototxt
I1228 17:55:50.810111 2733998976 net.cpp:296] The NetState phase (0) differed from the phase (1) specified by a rule in layer mnist
I1228 17:55:50.810127 2733998976 net.cpp:296] The NetState phase (0) differed from the phase (1) specified by a rule in layer accuracy
I1228 17:55:50.810133 2733998976 net.cpp:53] Initializing net from parameters: 
name: "LeNet"
state {
  phase: TRAIN
}
layer {
  name: "mnist"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TRAIN
  }
  transform_param {
    scale: 0.00390625
  }
  data_param {
    source: "/usr/local/Cellar/caffe/examples/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"
}
I1228 17:55:50.810202 2733998976 layer_factory.hpp:77] Creating layer mnist
I1228 17:55:50.810294 2733998976 db_lmdb.cpp:35] Opened lmdb /usr/local/Cellar/caffe/examples/mnist/mnist_train_lmdb
I1228 17:55:50.810323 2733998976 net.cpp:86] Creating Layer mnist
I1228 17:55:50.810329 2733998976 net.cpp:382] mnist -> data
I1228 17:55:50.810349 2733998976 net.cpp:382] mnist -> label
I1228 17:55:50.810376 2733998976 data_layer.cpp:45] output data size: 64,1,28,28
I1228 17:55:50.815076 2733998976 net.cpp:124] Setting up mnist
I1228 17:55:50.815102 2733998976 net.cpp:131] Top shape: 64 1 28 28 (50176)
I1228 17:55:50.815109 2733998976 net.cpp:131] Top shape: 64 (64)
I1228 17:55:50.815114 2733998976 net.cpp:139] Memory required for data: 200960
I1228 17:55:50.815119 2733998976 layer_factory.hpp:77] Creating layer conv1
I1228 17:55:50.815127 2733998976 net.cpp:86] Creating Layer conv1
I1228 17:55:50.815132 2733998976 net.cpp:408] conv1 <- data
I1228 17:55:50.815138 2733998976 net.cpp:382] conv1 -> conv1
I1228 17:55:50.815182 2733998976 net.cpp:124] Setting up conv1
I1228 17:55:50.815187 2733998976 net.cpp:131] Top shape: 64 20 24 24 (737280)
I1228 17:55:50.815193 2733998976 net.cpp:139] Memory required for data: 3150080
I1228 17:55:50.815203 2733998976 layer_factory.hpp:77] Creating layer pool1
I1228 17:55:50.815213 2733998976 net.cpp:86] Creating Layer pool1
I1228 17:55:50.815218 2733998976 net.cpp:408] pool1 <- conv1
I1228 17:55:50.815223 2733998976 net.cpp:382] pool1 -> pool1
I1228 17:55:50.815233 2733998976 net.cpp:124] Setting up pool1
I1228 17:55:50.815238 2733998976 net.cpp:131] Top shape: 64 20 12 12 (184320)
I1228 17:55:50.815246 2733998976 net.cpp:139] Memory required for data: 3887360
I1228 17:55:50.815250 2733998976 layer_factory.hpp:77] Creating layer conv2
I1228 17:55:50.815256 2733998976 net.cpp:86] Creating Layer conv2
I1228 17:55:50.815260 2733998976 net.cpp:408] conv2 <- pool1
I1228 17:55:50.815266 2733998976 net.cpp:382] conv2 -> conv2
I1228 17:55:50.815492 2733998976 net.cpp:124] Setting up conv2
I1228 17:55:50.815498 2733998976 net.cpp:131] Top shape: 64 50 8 8 (204800)
I1228 17:55:50.815503 2733998976 net.cpp:139] Memory required for data: 4706560
I1228 17:55:50.815510 2733998976 layer_factory.hpp:77] Creating layer pool2
I1228 17:55:50.815515 2733998976 net.cpp:86] Creating Layer pool2
I1228 17:55:50.815521 2733998976 net.cpp:408] pool2 <- conv2
I1228 17:55:50.815526 2733998976 net.cpp:382] pool2 -> pool2
I1228 17:55:50.815531 2733998976 net.cpp:124] Setting up pool2
I1228 17:55:50.815572 2733998976 net.cpp:131] Top shape: 64 50 4 4 (51200)
I1228 17:55:50.815589 2733998976 net.cpp:139] Memory required for data: 4911360
I1228 17:55:50.815593 2733998976 layer_factory.hpp:77] Creating layer ip1
I1228 17:55:50.815603 2733998976 net.cpp:86] Creating Layer ip1
I1228 17:55:50.815608 2733998976 net.cpp:408] ip1 <- pool2
I1228 17:55:50.815613 2733998976 net.cpp:382] ip1 -> ip1
I1228 17:55:50.818809 2733998976 net.cpp:124] Setting up ip1
I1228 17:55:50.818819 2733998976 net.cpp:131] Top shape: 64 500 (32000)
I1228 17:55:50.818823 2733998976 net.cpp:139] Memory required for data: 5039360
I1228 17:55:50.818830 2733998976 layer_factory.hpp:77] Creating layer relu1
I1228 17:55:50.818840 2733998976 net.cpp:86] Creating Layer relu1
I1228 17:55:50.818843 2733998976 net.cpp:408] relu1 <- ip1
I1228 17:55:50.818848 2733998976 net.cpp:369] relu1 -> ip1 (in-place)
I1228 17:55:50.818855 2733998976 net.cpp:124] Setting up relu1
I1228 17:55:50.818857 2733998976 net.cpp:131] Top shape: 64 500 (32000)
I1228 17:55:50.818862 2733998976 net.cpp:139] Memory required for data: 5167360
I1228 17:55:50.818866 2733998976 layer_factory.hpp:77] Creating layer ip2
I1228 17:55:50.818871 2733998976 net.cpp:86] Creating Layer ip2
I1228 17:55:50.818876 2733998976 net.cpp:408] ip2 <- ip1
I1228 17:55:50.818881 2733998976 net.cpp:382] ip2 -> ip2
I1228 17:55:50.818918 2733998976 net.cpp:124] Setting up ip2
I1228 17:55:50.818923 2733998976 net.cpp:131] Top shape: 64 10 (640)
I1228 17:55:50.818928 2733998976 net.cpp:139] Memory required for data: 5169920
I1228 17:55:50.818933 2733998976 layer_factory.hpp:77] Creating layer loss
I1228 17:55:50.818941 2733998976 net.cpp:86] Creating Layer loss
I1228 17:55:50.818945 2733998976 net.cpp:408] loss <- ip2
I1228 17:55:50.818949 2733998976 net.cpp:408] loss <- label
I1228 17:55:50.818954 2733998976 net.cpp:382] loss -> loss
I1228 17:55:50.818962 2733998976 layer_factory.hpp:77] Creating layer loss
I1228 17:55:50.818972 2733998976 net.cpp:124] Setting up loss
I1228 17:55:50.818977 2733998976 net.cpp:131] Top shape: (1)
I1228 17:55:50.818981 2733998976 net.cpp:134]     with loss weight 1
I1228 17:55:50.818989 2733998976 net.cpp:139] Memory required for data: 5169924
I1228 17:55:50.818992 2733998976 net.cpp:200] loss needs backward computation.
I1228 17:55:50.818996 2733998976 net.cpp:200] ip2 needs backward computation.
I1228 17:55:50.819001 2733998976 net.cpp:200] relu1 needs backward computation.
I1228 17:55:50.819005 2733998976 net.cpp:200] ip1 needs backward computation.
I1228 17:55:50.819010 2733998976 net.cpp:200] pool2 needs backward computation.
I1228 17:55:50.819012 2733998976 net.cpp:200] conv2 needs backward computation.
I1228 17:55:50.819017 2733998976 net.cpp:200] pool1 needs backward computation.
I1228 17:55:50.819021 2733998976 net.cpp:200] conv1 needs backward computation.
I1228 17:55:50.819026 2733998976 net.cpp:202] mnist does not need backward computation.
I1228 17:55:50.819031 2733998976 net.cpp:244] This network produces output loss
I1228 17:55:50.819036 2733998976 net.cpp:257] Network initialization done.
I1228 17:55:50.819180 2733998976 solver.cpp:190] Creating test net (#0) specified by net file: /usr/local/Cellar/caffe/examples/mnist/lenet_train_test.prototxt
I1228 17:55:50.819196 2733998976 net.cpp:296] The NetState phase (1) differed from the phase (0) specified by a rule in layer mnist
I1228 17:55:50.819205 2733998976 net.cpp:53] 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: "/usr/local/Cellar/caffe/examples/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"
}
I1228 17:55:50.819308 2733998976 layer_factory.hpp:77] Creating layer mnist
I1228 17:55:50.819391 2733998976 db_lmdb.cpp:35] Opened lmdb /usr/local/Cellar/caffe/examples/mnist/mnist_test_lmdb
I1228 17:55:50.819408 2733998976 net.cpp:86] Creating Layer mnist
I1228 17:55:50.819414 2733998976 net.cpp:382] mnist -> data
I1228 17:55:50.819420 2733998976 net.cpp:382] mnist -> label
I1228 17:55:50.819430 2733998976 data_layer.cpp:45] output data size: 100,1,28,28
I1228 17:55:50.820066 2733998976 net.cpp:124] Setting up mnist
I1228 17:55:50.820091 2733998976 net.cpp:131] Top shape: 100 1 28 28 (78400)
I1228 17:55:50.820097 2733998976 net.cpp:131] Top shape: 100 (100)
I1228 17:55:50.820101 2733998976 net.cpp:139] Memory required for data: 314000
I1228 17:55:50.820106 2733998976 layer_factory.hpp:77] Creating layer label_mnist_1_split
I1228 17:55:50.820112 2733998976 net.cpp:86] Creating Layer label_mnist_1_split
I1228 17:55:50.820116 2733998976 net.cpp:408] label_mnist_1_split <- label
I1228 17:55:50.820122 2733998976 net.cpp:382] label_mnist_1_split -> label_mnist_1_split_0
I1228 17:55:50.820128 2733998976 net.cpp:382] label_mnist_1_split -> label_mnist_1_split_1
I1228 17:55:50.820134 2733998976 net.cpp:124] Setting up label_mnist_1_split
I1228 17:55:50.820139 2733998976 net.cpp:131] Top shape: 100 (100)
I1228 17:55:50.820143 2733998976 net.cpp:131] Top shape: 100 (100)
I1228 17:55:50.820148 2733998976 net.cpp:139] Memory required for data: 314800
I1228 17:55:50.820152 2733998976 layer_factory.hpp:77] Creating layer conv1
I1228 17:55:50.820159 2733998976 net.cpp:86] Creating Layer conv1
I1228 17:55:50.820163 2733998976 net.cpp:408] conv1 <- data
I1228 17:55:50.820169 2733998976 net.cpp:382] conv1 -> conv1
I1228 17:55:50.820188 2733998976 net.cpp:124] Setting up conv1
I1228 17:55:50.820192 2733998976 net.cpp:131] Top shape: 100 20 24 24 (1152000)
I1228 17:55:50.820197 2733998976 net.cpp:139] Memory required for data: 4922800
I1228 17:55:50.820204 2733998976 layer_factory.hpp:77] Creating layer pool1
I1228 17:55:50.820210 2733998976 net.cpp:86] Creating Layer pool1
I1228 17:55:50.820214 2733998976 net.cpp:408] pool1 <- conv1
I1228 17:55:50.820222 2733998976 net.cpp:382] pool1 -> pool1
I1228 17:55:50.820230 2733998976 net.cpp:124] Setting up pool1
I1228 17:55:50.820235 2733998976 net.cpp:131] Top shape: 100 20 12 12 (288000)
I1228 17:55:50.820240 2733998976 net.cpp:139] Memory required for data: 6074800
I1228 17:55:50.820243 2733998976 layer_factory.hpp:77] Creating layer conv2
I1228 17:55:50.820250 2733998976 net.cpp:86] Creating Layer conv2
I1228 17:55:50.820255 2733998976 net.cpp:408] conv2 <- pool1
I1228 17:55:50.820261 2733998976 net.cpp:382] conv2 -> conv2
I1228 17:55:50.820474 2733998976 net.cpp:124] Setting up conv2
I1228 17:55:50.820480 2733998976 net.cpp:131] Top shape: 100 50 8 8 (320000)
I1228 17:55:50.820485 2733998976 net.cpp:139] Memory required for data: 7354800
I1228 17:55:50.820492 2733998976 layer_factory.hpp:77] Creating layer pool2
I1228 17:55:50.820497 2733998976 net.cpp:86] Creating Layer pool2
I1228 17:55:50.820647 2733998976 net.cpp:408] pool2 <- conv2
I1228 17:55:50.820652 2733998976 net.cpp:382] pool2 -> pool2
I1228 17:55:50.820659 2733998976 net.cpp:124] Setting up pool2
I1228 17:55:50.820663 2733998976 net.cpp:131] Top shape: 100 50 4 4 (80000)
I1228 17:55:50.820701 2733998976 net.cpp:139] Memory required for data: 7674800
I1228 17:55:50.820734 2733998976 layer_factory.hpp:77] Creating layer ip1
I1228 17:55:50.820765 2733998976 net.cpp:86] Creating Layer ip1
I1228 17:55:50.820768 2733998976 net.cpp:408] ip1 <- pool2
I1228 17:55:50.820775 2733998976 net.cpp:382] ip1 -> ip1
I1228 17:55:50.824555 2733998976 net.cpp:124] Setting up ip1
I1228 17:55:50.824573 2733998976 net.cpp:131] Top shape: 100 500 (50000)
I1228 17:55:50.824594 2733998976 net.cpp:139] Memory required for data: 7874800
I1228 17:55:50.824602 2733998976 layer_factory.hpp:77] Creating layer relu1
I1228 17:55:50.824621 2733998976 net.cpp:86] Creating Layer relu1
I1228 17:55:50.824627 2733998976 net.cpp:408] relu1 <- ip1
I1228 17:55:50.824632 2733998976 net.cpp:369] relu1 -> ip1 (in-place)
I1228 17:55:50.824640 2733998976 net.cpp:124] Setting up relu1
I1228 17:55:50.824643 2733998976 net.cpp:131] Top shape: 100 500 (50000)
I1228 17:55:50.824648 2733998976 net.cpp:139] Memory required for data: 8074800
I1228 17:55:50.824652 2733998976 layer_factory.hpp:77] Creating layer ip2
I1228 17:55:50.824679 2733998976 net.cpp:86] Creating Layer ip2
I1228 17:55:50.824683 2733998976 net.cpp:408] ip2 <- ip1
I1228 17:55:50.824689 2733998976 net.cpp:382] ip2 -> ip2
I1228 17:55:50.824791 2733998976 net.cpp:124] Setting up ip2
I1228 17:55:50.824796 2733998976 net.cpp:131] Top shape: 100 10 (1000)
I1228 17:55:50.824800 2733998976 net.cpp:139] Memory required for data: 8078800
I1228 17:55:50.824806 2733998976 layer_factory.hpp:77] Creating layer ip2_ip2_0_split
I1228 17:55:50.824812 2733998976 net.cpp:86] Creating Layer ip2_ip2_0_split
I1228 17:55:50.824816 2733998976 net.cpp:408] ip2_ip2_0_split <- ip2
I1228 17:55:50.824821 2733998976 net.cpp:382] ip2_ip2_0_split -> ip2_ip2_0_split_0
I1228 17:55:50.824827 2733998976 net.cpp:382] ip2_ip2_0_split -> ip2_ip2_0_split_1
I1228 17:55:50.824833 2733998976 net.cpp:124] Setting up ip2_ip2_0_split
I1228 17:55:50.824837 2733998976 net.cpp:131] Top shape: 100 10 (1000)
I1228 17:55:50.824842 2733998976 net.cpp:131] Top shape: 100 10 (1000)
I1228 17:55:50.824865 2733998976 net.cpp:139] Memory required for data: 8086800
I1228 17:55:50.824869 2733998976 layer_factory.hpp:77] Creating layer accuracy
I1228 17:55:50.824875 2733998976 net.cpp:86] Creating Layer accuracy
I1228 17:55:50.824880 2733998976 net.cpp:408] accuracy <- ip2_ip2_0_split_0
I1228 17:55:50.824885 2733998976 net.cpp:408] accuracy <- label_mnist_1_split_0
I1228 17:55:50.824903 2733998976 net.cpp:382] accuracy -> accuracy
I1228 17:55:50.824909 2733998976 net.cpp:124] Setting up accuracy
I1228 17:55:50.824913 2733998976 net.cpp:131] Top shape: (1)
I1228 17:55:50.824918 2733998976 net.cpp:139] Memory required for data: 8086804
I1228 17:55:50.824923 2733998976 layer_factory.hpp:77] Creating layer loss
I1228 17:55:50.824928 2733998976 net.cpp:86] Creating Layer loss
I1228 17:55:50.824931 2733998976 net.cpp:408] loss <- ip2_ip2_0_split_1
I1228 17:55:50.824936 2733998976 net.cpp:408] loss <- label_mnist_1_split_1
I1228 17:55:50.824941 2733998976 net.cpp:382] loss -> loss
I1228 17:55:50.824947 2733998976 layer_factory.hpp:77] Creating layer loss
I1228 17:55:50.824957 2733998976 net.cpp:124] Setting up loss
I1228 17:55:50.824962 2733998976 net.cpp:131] Top shape: (1)
I1228 17:55:50.824966 2733998976 net.cpp:134]     with loss weight 1
I1228 17:55:50.824973 2733998976 net.cpp:139] Memory required for data: 8086808
I1228 17:55:50.824977 2733998976 net.cpp:200] loss needs backward computation.
I1228 17:55:50.824982 2733998976 net.cpp:202] accuracy does not need backward computation.
I1228 17:55:50.824986 2733998976 net.cpp:200] ip2_ip2_0_split needs backward computation.
I1228 17:55:50.824990 2733998976 net.cpp:200] ip2 needs backward computation.
I1228 17:55:50.824995 2733998976 net.cpp:200] relu1 needs backward computation.
I1228 17:55:50.825000 2733998976 net.cpp:200] ip1 needs backward computation.
I1228 17:55:50.825003 2733998976 net.cpp:200] pool2 needs backward computation.
I1228 17:55:50.825008 2733998976 net.cpp:200] conv2 needs backward computation.
I1228 17:55:50.825012 2733998976 net.cpp:200] pool1 needs backward computation.
I1228 17:55:50.825016 2733998976 net.cpp:200] conv1 needs backward computation.
I1228 17:55:50.825021 2733998976 net.cpp:202] label_mnist_1_split does not need backward computation.
I1228 17:55:50.825026 2733998976 net.cpp:202] mnist does not need backward computation.
I1228 17:55:50.825029 2733998976 net.cpp:244] This network produces output accuracy
I1228 17:55:50.825034 2733998976 net.cpp:244] This network produces output loss
I1228 17:55:50.825042 2733998976 net.cpp:257] Network initialization done.
I1228 17:55:50.825074 2733998976 solver.cpp:57] Solver scaffolding done.
I1228 17:55:50.826243 2733998976 solver.cpp:347] Iteration 0, Testing net (#0)
I1228 17:55:53.022482 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:55:53.110131 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.0886
I1228 17:55:53.110157 2733998976 solver.cpp:414]     Test net output #1: loss = 2.39438 (* 1 = 2.39438 loss)
I1228 17:55:53.144232 2733998976 solver.cpp:239] Iteration 0 (0 iter/s, 2.319s/100 iters), loss = 2.39822
I1228 17:55:53.144258 2733998976 solver.cpp:258]     Train net output #0: loss = 2.39822 (* 1 = 2.39822 loss)
I1228 17:55:53.144269 2733998976 sgd_solver.cpp:112] Iteration 0, lr = 0.01
I1228 17:55:55.400316 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:55:58.461671 2733998976 solver.cpp:239] Iteration 100 (18.8076 iter/s, 5.317s/100 iters), loss = 0.199418
I1228 17:55:58.461697 2733998976 solver.cpp:258]     Train net output #0: loss = 0.199418 (* 1 = 0.199418 loss)
I1228 17:55:58.461704 2733998976 sgd_solver.cpp:112] Iteration 100, lr = 0.00992565
I1228 17:56:01.460424 2733998976 solver.cpp:239] Iteration 200 (33.3556 iter/s, 2.998s/100 iters), loss = 0.130344
I1228 17:56:01.460450 2733998976 solver.cpp:258]     Train net output #0: loss = 0.130344 (* 1 = 0.130344 loss)
I1228 17:56:01.460458 2733998976 sgd_solver.cpp:112] Iteration 200, lr = 0.00985258
I1228 17:56:04.451843 2733998976 solver.cpp:239] Iteration 300 (33.4336 iter/s, 2.991s/100 iters), loss = 0.196874
I1228 17:56:04.451869 2733998976 solver.cpp:258]     Train net output #0: loss = 0.196874 (* 1 = 0.196874 loss)
I1228 17:56:04.451877 2733998976 sgd_solver.cpp:112] Iteration 300, lr = 0.00978075
I1228 17:56:07.440696 2733998976 solver.cpp:239] Iteration 400 (33.4672 iter/s, 2.988s/100 iters), loss = 0.0869522
I1228 17:56:07.440726 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0869522 (* 1 = 0.0869522 loss)
I1228 17:56:07.440733 2733998976 sgd_solver.cpp:112] Iteration 400, lr = 0.00971013
I1228 17:56:10.421128 2733998976 solver.cpp:347] Iteration 500, Testing net (#0)
I1228 17:56:12.586988 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:56:12.675629 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.972
I1228 17:56:12.675654 2733998976 solver.cpp:414]     Test net output #1: loss = 0.086382 (* 1 = 0.086382 loss)
I1228 17:56:12.704449 2733998976 solver.cpp:239] Iteration 500 (19.0006 iter/s, 5.263s/100 iters), loss = 0.0954096
I1228 17:56:12.704476 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0954096 (* 1 = 0.0954096 loss)
I1228 17:56:12.704483 2733998976 sgd_solver.cpp:112] Iteration 500, lr = 0.00964069
I1228 17:56:14.965848 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:56:18.031318 2733998976 solver.cpp:239] Iteration 600 (18.7758 iter/s, 5.326s/100 iters), loss = 0.0956796
I1228 17:56:18.031345 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0956796 (* 1 = 0.0956796 loss)
I1228 17:56:18.031353 2733998976 sgd_solver.cpp:112] Iteration 600, lr = 0.0095724
I1228 17:56:21.054196 2733998976 solver.cpp:239] Iteration 700 (33.0907 iter/s, 3.022s/100 iters), loss = 0.119464
I1228 17:56:21.054222 2733998976 solver.cpp:258]     Train net output #0: loss = 0.119464 (* 1 = 0.119464 loss)
I1228 17:56:21.054229 2733998976 sgd_solver.cpp:112] Iteration 700, lr = 0.00950522
I1228 17:56:24.041611 2733998976 solver.cpp:239] Iteration 800 (33.4784 iter/s, 2.987s/100 iters), loss = 0.212201
I1228 17:56:24.041638 2733998976 solver.cpp:258]     Train net output #0: loss = 0.212201 (* 1 = 0.212201 loss)
I1228 17:56:24.041646 2733998976 sgd_solver.cpp:112] Iteration 800, lr = 0.00943913
I1228 17:56:27.032083 2733998976 solver.cpp:239] Iteration 900 (33.4448 iter/s, 2.99s/100 iters), loss = 0.154787
I1228 17:56:27.032121 2733998976 solver.cpp:258]     Train net output #0: loss = 0.154787 (* 1 = 0.154787 loss)
I1228 17:56:27.032128 2733998976 sgd_solver.cpp:112] Iteration 900, lr = 0.00937411
I1228 17:56:28.022493 210243584 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:56:30.003214 2733998976 solver.cpp:347] Iteration 1000, Testing net (#0)
I1228 17:56:32.152966 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:56:32.241582 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9823
I1228 17:56:32.241607 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0546377 (* 1 = 0.0546377 loss)
I1228 17:56:32.270887 2733998976 solver.cpp:239] Iteration 1000 (19.0913 iter/s, 5.238s/100 iters), loss = 0.053198
I1228 17:56:32.270915 2733998976 solver.cpp:258]     Train net output #0: loss = 0.053198 (* 1 = 0.053198 loss)
I1228 17:56:32.270922 2733998976 sgd_solver.cpp:112] Iteration 1000, lr = 0.00931012
I1228 17:56:34.504310 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:56:37.551901 2733998976 solver.cpp:239] Iteration 1100 (18.9394 iter/s, 5.28s/100 iters), loss = 0.00487837
I1228 17:56:37.551930 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00487837 (* 1 = 0.00487837 loss)
I1228 17:56:37.551937 2733998976 sgd_solver.cpp:112] Iteration 1100, lr = 0.00924715
I1228 17:56:40.550760 2733998976 solver.cpp:239] Iteration 1200 (33.3556 iter/s, 2.998s/100 iters), loss = 0.00953127
I1228 17:56:40.550787 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00953127 (* 1 = 0.00953127 loss)
I1228 17:56:40.550796 2733998976 sgd_solver.cpp:112] Iteration 1200, lr = 0.00918515
I1228 17:56:43.530740 2733998976 solver.cpp:239] Iteration 1300 (33.5683 iter/s, 2.979s/100 iters), loss = 0.0194495
I1228 17:56:43.530769 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0194495 (* 1 = 0.0194495 loss)
I1228 17:56:43.530776 2733998976 sgd_solver.cpp:112] Iteration 1300, lr = 0.00912412
I1228 17:56:46.519441 2733998976 solver.cpp:239] Iteration 1400 (33.4672 iter/s, 2.988s/100 iters), loss = 0.00659618
I1228 17:56:46.519477 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00659618 (* 1 = 0.00659618 loss)
I1228 17:56:46.519493 2733998976 sgd_solver.cpp:112] Iteration 1400, lr = 0.00906403
I1228 17:56:49.531745 2733998976 solver.cpp:347] Iteration 1500, Testing net (#0)
I1228 17:56:51.735550 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:56:51.827378 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9852
I1228 17:56:51.827402 2733998976 solver.cpp:414]     Test net output #1: loss = 0.046295 (* 1 = 0.046295 loss)
I1228 17:56:51.857616 2733998976 solver.cpp:239] Iteration 1500 (18.7336 iter/s, 5.338s/100 iters), loss = 0.0957371
I1228 17:56:51.857666 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0957371 (* 1 = 0.0957371 loss)
I1228 17:56:51.857674 2733998976 sgd_solver.cpp:112] Iteration 1500, lr = 0.00900485
I1228 17:56:54.138475 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:56:57.181025 2733998976 solver.cpp:239] Iteration 1600 (18.7864 iter/s, 5.323s/100 iters), loss = 0.129418
I1228 17:56:57.181051 2733998976 solver.cpp:258]     Train net output #0: loss = 0.129418 (* 1 = 0.129418 loss)
I1228 17:56:57.181057 2733998976 sgd_solver.cpp:112] Iteration 1600, lr = 0.00894657
I1228 17:57:00.391135 2733998976 solver.cpp:239] Iteration 1700 (31.1526 iter/s, 3.21s/100 iters), loss = 0.0193124
I1228 17:57:00.391162 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0193124 (* 1 = 0.0193124 loss)
I1228 17:57:00.391170 2733998976 sgd_solver.cpp:112] Iteration 1700, lr = 0.00888916
I1228 17:57:03.499020 2733998976 solver.cpp:239] Iteration 1800 (32.1854 iter/s, 3.107s/100 iters), loss = 0.0188734
I1228 17:57:03.499047 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0188734 (* 1 = 0.0188734 loss)
I1228 17:57:03.499055 2733998976 sgd_solver.cpp:112] Iteration 1800, lr = 0.0088326
I1228 17:57:05.743053 210243584 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:57:06.658465 2733998976 solver.cpp:239] Iteration 1900 (31.6556 iter/s, 3.159s/100 iters), loss = 0.110986
I1228 17:57:06.658493 2733998976 solver.cpp:258]     Train net output #0: loss = 0.110986 (* 1 = 0.110986 loss)
I1228 17:57:06.658500 2733998976 sgd_solver.cpp:112] Iteration 1900, lr = 0.00877687
I1228 17:57:09.826277 2733998976 solver.cpp:347] Iteration 2000, Testing net (#0)
I1228 17:57:12.128520 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:57:12.220306 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9862
I1228 17:57:12.220331 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0407267 (* 1 = 0.0407267 loss)
I1228 17:57:12.252471 2733998976 solver.cpp:239] Iteration 2000 (17.8795 iter/s, 5.593s/100 iters), loss = 0.0132918
I1228 17:57:12.252497 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0132918 (* 1 = 0.0132918 loss)
I1228 17:57:12.252504 2733998976 sgd_solver.cpp:112] Iteration 2000, lr = 0.00872196
I1228 17:57:14.704600 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:57:17.740623 2733998976 solver.cpp:239] Iteration 2100 (18.2216 iter/s, 5.488s/100 iters), loss = 0.0271667
I1228 17:57:17.740653 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0271667 (* 1 = 0.0271667 loss)
I1228 17:57:17.740664 2733998976 sgd_solver.cpp:112] Iteration 2100, lr = 0.00866784
I1228 17:57:20.737411 2733998976 solver.cpp:239] Iteration 2200 (33.3778 iter/s, 2.996s/100 iters), loss = 0.0226718
I1228 17:57:20.737437 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0226718 (* 1 = 0.0226718 loss)
I1228 17:57:20.737444 2733998976 sgd_solver.cpp:112] Iteration 2200, lr = 0.0086145
I1228 17:57:23.731586 2733998976 solver.cpp:239] Iteration 2300 (33.4001 iter/s, 2.994s/100 iters), loss = 0.105805
I1228 17:57:23.731614 2733998976 solver.cpp:258]     Train net output #0: loss = 0.105805 (* 1 = 0.105805 loss)
I1228 17:57:23.731621 2733998976 sgd_solver.cpp:112] Iteration 2300, lr = 0.00856192
I1228 17:57:26.741955 2733998976 solver.cpp:239] Iteration 2400 (33.2226 iter/s, 3.01s/100 iters), loss = 0.00785064
I1228 17:57:26.741982 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00785064 (* 1 = 0.00785064 loss)
I1228 17:57:26.741991 2733998976 sgd_solver.cpp:112] Iteration 2400, lr = 0.00851008
I1228 17:57:29.723547 2733998976 solver.cpp:347] Iteration 2500, Testing net (#0)
I1228 17:57:31.921429 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:57:32.011425 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9842
I1228 17:57:32.011451 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0485859 (* 1 = 0.0485859 loss)
I1228 17:57:32.044616 2733998976 solver.cpp:239] Iteration 2500 (18.8608 iter/s, 5.302s/100 iters), loss = 0.0214106
I1228 17:57:32.044641 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0214106 (* 1 = 0.0214106 loss)
I1228 17:57:32.044648 2733998976 sgd_solver.cpp:112] Iteration 2500, lr = 0.00845897
I1228 17:57:34.297142 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:57:37.362659 2733998976 solver.cpp:239] Iteration 2600 (18.8041 iter/s, 5.318s/100 iters), loss = 0.0625504
I1228 17:57:37.362687 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0625504 (* 1 = 0.0625504 loss)
I1228 17:57:37.362694 2733998976 sgd_solver.cpp:112] Iteration 2600, lr = 0.00840857
I1228 17:57:40.430872 2733998976 solver.cpp:239] Iteration 2700 (32.5945 iter/s, 3.068s/100 iters), loss = 0.0545578
I1228 17:57:40.430898 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0545578 (* 1 = 0.0545578 loss)
I1228 17:57:40.430907 2733998976 sgd_solver.cpp:112] Iteration 2700, lr = 0.00835886
I1228 17:57:43.639781 2733998976 solver.cpp:239] Iteration 2800 (31.1721 iter/s, 3.208s/100 iters), loss = 0.00138936
I1228 17:57:43.639809 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00138936 (* 1 = 0.00138936 loss)
I1228 17:57:43.639816 2733998976 sgd_solver.cpp:112] Iteration 2800, lr = 0.00830984
I1228 17:57:43.869724 210243584 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:57:46.593061 2733998976 solver.cpp:239] Iteration 2900 (33.8639 iter/s, 2.953s/100 iters), loss = 0.0231533
I1228 17:57:46.593088 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0231533 (* 1 = 0.0231533 loss)
I1228 17:57:46.593096 2733998976 sgd_solver.cpp:112] Iteration 2900, lr = 0.00826148
I1228 17:57:49.484295 2733998976 solver.cpp:347] Iteration 3000, Testing net (#0)
I1228 17:57:51.664630 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:57:51.754335 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9871
I1228 17:57:51.754359 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0406435 (* 1 = 0.0406435 loss)
I1228 17:57:51.781070 2733998976 solver.cpp:239] Iteration 3000 (19.279 iter/s, 5.187s/100 iters), loss = 0.0197396
I1228 17:57:51.781095 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0197396 (* 1 = 0.0197396 loss)
I1228 17:57:51.781102 2733998976 sgd_solver.cpp:112] Iteration 3000, lr = 0.00821377
I1228 17:57:54.068579 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:57:57.061866 2733998976 solver.cpp:239] Iteration 3100 (18.9394 iter/s, 5.28s/100 iters), loss = 0.0151984
I1228 17:57:57.061902 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0151984 (* 1 = 0.0151984 loss)
I1228 17:57:57.061918 2733998976 sgd_solver.cpp:112] Iteration 3100, lr = 0.0081667
I1228 17:58:00.250874 2733998976 solver.cpp:239] Iteration 3200 (31.3676 iter/s, 3.188s/100 iters), loss = 0.00594261
I1228 17:58:00.250900 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00594261 (* 1 = 0.00594261 loss)
I1228 17:58:00.250908 2733998976 sgd_solver.cpp:112] Iteration 3200, lr = 0.00812025
I1228 17:58:03.203016 2733998976 solver.cpp:239] Iteration 3300 (33.8753 iter/s, 2.952s/100 iters), loss = 0.0419102
I1228 17:58:03.203045 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0419102 (* 1 = 0.0419102 loss)
I1228 17:58:03.203053 2733998976 sgd_solver.cpp:112] Iteration 3300, lr = 0.00807442
I1228 17:58:06.174073 2733998976 solver.cpp:239] Iteration 3400 (33.6587 iter/s, 2.971s/100 iters), loss = 0.00628851
I1228 17:58:06.174099 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00628851 (* 1 = 0.00628851 loss)
I1228 17:58:06.174108 2733998976 sgd_solver.cpp:112] Iteration 3400, lr = 0.00802918
I1228 17:58:09.104461 2733998976 solver.cpp:347] Iteration 3500, Testing net (#0)
I1228 17:58:11.270108 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:58:11.359313 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9867
I1228 17:58:11.359341 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0408062 (* 1 = 0.0408062 loss)
I1228 17:58:11.386420 2733998976 solver.cpp:239] Iteration 3500 (19.1865 iter/s, 5.212s/100 iters), loss = 0.00765101
I1228 17:58:11.386447 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00765101 (* 1 = 0.00765101 loss)
I1228 17:58:11.386454 2733998976 sgd_solver.cpp:112] Iteration 3500, lr = 0.00798454
I1228 17:58:13.678453 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:58:16.771239 2733998976 solver.cpp:239] Iteration 3600 (18.5736 iter/s, 5.384s/100 iters), loss = 0.0324877
I1228 17:58:16.771266 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0324877 (* 1 = 0.0324877 loss)
I1228 17:58:16.771273 2733998976 sgd_solver.cpp:112] Iteration 3600, lr = 0.00794046
I1228 17:58:19.771371 2733998976 solver.cpp:239] Iteration 3700 (33.3333 iter/s, 3s/100 iters), loss = 0.0184573
I1228 17:58:19.771399 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0184573 (* 1 = 0.0184573 loss)
I1228 17:58:19.771406 2733998976 sgd_solver.cpp:112] Iteration 3700, lr = 0.00789695
I1228 17:58:21.106961 210243584 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:58:22.749763 2733998976 solver.cpp:239] Iteration 3800 (33.5796 iter/s, 2.978s/100 iters), loss = 0.0156665
I1228 17:58:22.749791 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0156665 (* 1 = 0.0156665 loss)
I1228 17:58:22.749799 2733998976 sgd_solver.cpp:112] Iteration 3800, lr = 0.007854
I1228 17:58:25.847599 2733998976 solver.cpp:239] Iteration 3900 (32.2893 iter/s, 3.097s/100 iters), loss = 0.0366668
I1228 17:58:25.847625 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0366668 (* 1 = 0.0366668 loss)
I1228 17:58:25.847632 2733998976 sgd_solver.cpp:112] Iteration 3900, lr = 0.00781158
I1228 17:58:28.841053 2733998976 solver.cpp:347] Iteration 4000, Testing net (#0)
I1228 17:58:31.186780 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:58:31.278307 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9899
I1228 17:58:31.278332 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0315919 (* 1 = 0.0315919 loss)
I1228 17:58:31.307127 2733998976 solver.cpp:239] Iteration 4000 (18.3184 iter/s, 5.459s/100 iters), loss = 0.0191784
I1228 17:58:31.307155 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0191784 (* 1 = 0.0191784 loss)
I1228 17:58:31.307163 2733998976 sgd_solver.cpp:112] Iteration 4000, lr = 0.00776969
I1228 17:58:33.665038 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:58:37.257160 2733998976 solver.cpp:239] Iteration 4100 (16.8067 iter/s, 5.95s/100 iters), loss = 0.0143265
I1228 17:58:37.257194 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0143265 (* 1 = 0.0143265 loss)
I1228 17:58:37.257203 2733998976 sgd_solver.cpp:112] Iteration 4100, lr = 0.00772833
I1228 17:58:40.578187 2733998976 solver.cpp:239] Iteration 4200 (30.1205 iter/s, 3.32s/100 iters), loss = 0.0118002
I1228 17:58:40.578214 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0118002 (* 1 = 0.0118002 loss)
I1228 17:58:40.578222 2733998976 sgd_solver.cpp:112] Iteration 4200, lr = 0.00768748
I1228 17:58:43.781576 2733998976 solver.cpp:239] Iteration 4300 (31.2207 iter/s, 3.203s/100 iters), loss = 0.0523296
I1228 17:58:43.781610 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0523296 (* 1 = 0.0523296 loss)
I1228 17:58:43.781620 2733998976 sgd_solver.cpp:112] Iteration 4300, lr = 0.00764712
I1228 17:58:47.871171 2733998976 solver.cpp:239] Iteration 4400 (24.4559 iter/s, 4.089s/100 iters), loss = 0.0258025
I1228 17:58:47.871197 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0258025 (* 1 = 0.0258025 loss)
I1228 17:58:47.871206 2733998976 sgd_solver.cpp:112] Iteration 4400, lr = 0.00760726
I1228 17:58:51.312611 2733998976 solver.cpp:347] Iteration 4500, Testing net (#0)
I1228 17:58:53.480944 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:58:53.568606 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9896
I1228 17:58:53.568632 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0343229 (* 1 = 0.0343229 loss)
I1228 17:58:53.596662 2733998976 solver.cpp:239] Iteration 4500 (17.4672 iter/s, 5.725s/100 iters), loss = 0.00411332
I1228 17:58:53.596688 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00411332 (* 1 = 0.00411332 loss)
I1228 17:58:53.596695 2733998976 sgd_solver.cpp:112] Iteration 4500, lr = 0.00756788
I1228 17:58:55.893582 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:58:59.003871 2733998976 solver.cpp:239] Iteration 4600 (18.4945 iter/s, 5.407s/100 iters), loss = 0.0100597
I1228 17:58:59.003899 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0100597 (* 1 = 0.0100597 loss)
I1228 17:58:59.003907 2733998976 sgd_solver.cpp:112] Iteration 4600, lr = 0.00752897
I1228 17:59:01.549305 210243584 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:59:02.282960 2733998976 solver.cpp:239] Iteration 4700 (30.4971 iter/s, 3.279s/100 iters), loss = 0.00317848
I1228 17:59:02.283004 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00317848 (* 1 = 0.00317848 loss)
I1228 17:59:02.283015 2733998976 sgd_solver.cpp:112] Iteration 4700, lr = 0.00749052
I1228 17:59:05.461637 2733998976 solver.cpp:239] Iteration 4800 (31.4663 iter/s, 3.178s/100 iters), loss = 0.015481
I1228 17:59:05.461663 2733998976 solver.cpp:258]     Train net output #0: loss = 0.015481 (* 1 = 0.015481 loss)
I1228 17:59:05.461670 2733998976 sgd_solver.cpp:112] Iteration 4800, lr = 0.00745253
I1228 17:59:08.421372 2733998976 solver.cpp:239] Iteration 4900 (33.7952 iter/s, 2.959s/100 iters), loss = 0.0076528
I1228 17:59:08.421401 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0076528 (* 1 = 0.0076528 loss)
I1228 17:59:08.421408 2733998976 sgd_solver.cpp:112] Iteration 4900, lr = 0.00741498
I1228 17:59:11.388818 2733998976 solver.cpp:464] Snapshotting to binary proto file /Users/taily/pycharmproj/lenet_solver_test_show_iter_5000.caffemodel
I1228 17:59:11.393275 2733998976 sgd_solver.cpp:284] Snapshotting solver state to binary proto file /Users/taily/pycharmproj/lenet_solver_test_show_iter_5000.solverstate
I1228 17:59:11.394623 2733998976 solver.cpp:347] Iteration 5000, Testing net (#0)
I1228 17:59:13.615495 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:59:13.703135 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9897
I1228 17:59:13.703161 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0317835 (* 1 = 0.0317835 loss)
I1228 17:59:13.731845 2733998976 solver.cpp:239] Iteration 5000 (18.8324 iter/s, 5.31s/100 iters), loss = 0.0358031
I1228 17:59:13.731875 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0358031 (* 1 = 0.0358031 loss)
I1228 17:59:13.731883 2733998976 sgd_solver.cpp:112] Iteration 5000, lr = 0.00737788
I1228 17:59:16.064182 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:59:19.139829 2733998976 solver.cpp:239] Iteration 5100 (18.4945 iter/s, 5.407s/100 iters), loss = 0.017382
I1228 17:59:19.139858 2733998976 solver.cpp:258]     Train net output #0: loss = 0.017382 (* 1 = 0.017382 loss)
I1228 17:59:19.139865 2733998976 sgd_solver.cpp:112] Iteration 5100, lr = 0.0073412
I1228 17:59:23.241438 2733998976 solver.cpp:239] Iteration 5200 (24.3843 iter/s, 4.101s/100 iters), loss = 0.0074814
I1228 17:59:23.241466 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0074814 (* 1 = 0.0074814 loss)
I1228 17:59:23.241473 2733998976 sgd_solver.cpp:112] Iteration 5200, lr = 0.00730495
I1228 17:59:26.214295 2733998976 solver.cpp:239] Iteration 5300 (33.6474 iter/s, 2.972s/100 iters), loss = 0.00124816
I1228 17:59:26.214323 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00124816 (* 1 = 0.00124816 loss)
I1228 17:59:26.214330 2733998976 sgd_solver.cpp:112] Iteration 5300, lr = 0.00726911
I1228 17:59:29.230228 2733998976 solver.cpp:239] Iteration 5400 (33.1675 iter/s, 3.015s/100 iters), loss = 0.00897871
I1228 17:59:29.230257 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00897871 (* 1 = 0.00897871 loss)
I1228 17:59:29.230263 2733998976 sgd_solver.cpp:112] Iteration 5400, lr = 0.00723368
I1228 17:59:32.210072 2733998976 solver.cpp:347] Iteration 5500, Testing net (#0)
I1228 17:59:34.675516 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:59:34.809540 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.988
I1228 17:59:34.809629 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0347851 (* 1 = 0.0347851 loss)
I1228 17:59:34.855731 2733998976 solver.cpp:239] Iteration 5500 (17.7778 iter/s, 5.625s/100 iters), loss = 0.00980196
I1228 17:59:34.855765 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00980196 (* 1 = 0.00980196 loss)
I1228 17:59:34.855774 2733998976 sgd_solver.cpp:112] Iteration 5500, lr = 0.00719865
I1228 17:59:37.256271 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:59:40.275497 2733998976 solver.cpp:239] Iteration 5600 (18.4536 iter/s, 5.419s/100 iters), loss = 0.000928452
I1228 17:59:40.275527 2733998976 solver.cpp:258]     Train net output #0: loss = 0.000928452 (* 1 = 0.000928452 loss)
I1228 17:59:40.275533 2733998976 sgd_solver.cpp:112] Iteration 5600, lr = 0.00716402
I1228 17:59:40.876044 210243584 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:59:43.253707 2733998976 solver.cpp:239] Iteration 5700 (33.5796 iter/s, 2.978s/100 iters), loss = 0.0043696
I1228 17:59:43.253737 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0043696 (* 1 = 0.0043696 loss)
I1228 17:59:43.253746 2733998976 sgd_solver.cpp:112] Iteration 5700, lr = 0.00712977
I1228 17:59:46.250402 2733998976 solver.cpp:239] Iteration 5800 (33.3778 iter/s, 2.996s/100 iters), loss = 0.0301731
I1228 17:59:46.250432 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0301731 (* 1 = 0.0301731 loss)
I1228 17:59:46.250439 2733998976 sgd_solver.cpp:112] Iteration 5800, lr = 0.0070959
I1228 17:59:49.302268 2733998976 solver.cpp:239] Iteration 5900 (32.7761 iter/s, 3.051s/100 iters), loss = 0.00737168
I1228 17:59:49.302371 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00737168 (* 1 = 0.00737168 loss)
I1228 17:59:49.302392 2733998976 sgd_solver.cpp:112] Iteration 5900, lr = 0.0070624
I1228 17:59:52.800864 2733998976 solver.cpp:347] Iteration 6000, Testing net (#0)
I1228 17:59:55.019049 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 17:59:55.107859 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9903
I1228 17:59:55.107934 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0295323 (* 1 = 0.0295323 loss)
I1228 17:59:55.136704 2733998976 solver.cpp:239] Iteration 6000 (17.1409 iter/s, 5.834s/100 iters), loss = 0.00375669
I1228 17:59:55.136732 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00375669 (* 1 = 0.00375669 loss)
I1228 17:59:55.136740 2733998976 sgd_solver.cpp:112] Iteration 6000, lr = 0.00702927
I1228 17:59:57.629866 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:00:00.632270 2733998976 solver.cpp:239] Iteration 6100 (18.1984 iter/s, 5.495s/100 iters), loss = 0.00206975
I1228 18:00:00.632297 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00206975 (* 1 = 0.00206975 loss)
I1228 18:00:00.632303 2733998976 sgd_solver.cpp:112] Iteration 6100, lr = 0.0069965
I1228 18:00:03.699493 2733998976 solver.cpp:239] Iteration 6200 (32.6052 iter/s, 3.067s/100 iters), loss = 0.0100554
I1228 18:00:03.699522 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0100554 (* 1 = 0.0100554 loss)
I1228 18:00:03.699530 2733998976 sgd_solver.cpp:112] Iteration 6200, lr = 0.00696408
I1228 18:00:06.682682 2733998976 solver.cpp:239] Iteration 6300 (33.5233 iter/s, 2.983s/100 iters), loss = 0.00990718
I1228 18:00:06.682710 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00990718 (* 1 = 0.00990718 loss)
I1228 18:00:06.682718 2733998976 sgd_solver.cpp:112] Iteration 6300, lr = 0.00693201
I1228 18:00:10.203083 2733998976 solver.cpp:239] Iteration 6400 (28.4091 iter/s, 3.52s/100 iters), loss = 0.00909858
I1228 18:00:10.203112 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00909858 (* 1 = 0.00909858 loss)
I1228 18:00:10.203119 2733998976 sgd_solver.cpp:112] Iteration 6400, lr = 0.00690029
I1228 18:00:13.111776 2733998976 solver.cpp:347] Iteration 6500, Testing net (#0)
I1228 18:00:15.288513 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:00:15.375962 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9894
I1228 18:00:15.375988 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0316598 (* 1 = 0.0316598 loss)
I1228 18:00:15.404659 2733998976 solver.cpp:239] Iteration 6500 (19.2271 iter/s, 5.201s/100 iters), loss = 0.0130249
I1228 18:00:15.404686 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0130249 (* 1 = 0.0130249 loss)
I1228 18:00:15.404693 2733998976 sgd_solver.cpp:112] Iteration 6500, lr = 0.0068689
I1228 18:00:17.712810 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:00:19.522375 210243584 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:00:20.790387 2733998976 solver.cpp:239] Iteration 6600 (18.5701 iter/s, 5.385s/100 iters), loss = 0.0314408
I1228 18:00:20.790416 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0314408 (* 1 = 0.0314408 loss)
I1228 18:00:20.790424 2733998976 sgd_solver.cpp:112] Iteration 6600, lr = 0.00683784
I1228 18:00:23.775645 2733998976 solver.cpp:239] Iteration 6700 (33.5008 iter/s, 2.985s/100 iters), loss = 0.00651566
I1228 18:00:23.775673 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00651566 (* 1 = 0.00651566 loss)
I1228 18:00:23.775681 2733998976 sgd_solver.cpp:112] Iteration 6700, lr = 0.00680711
I1228 18:00:26.771030 2733998976 solver.cpp:239] Iteration 6800 (33.389 iter/s, 2.995s/100 iters), loss = 0.00297121
I1228 18:00:26.771059 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00297121 (* 1 = 0.00297121 loss)
I1228 18:00:26.771066 2733998976 sgd_solver.cpp:112] Iteration 6800, lr = 0.0067767
I1228 18:00:29.782070 2733998976 solver.cpp:239] Iteration 6900 (33.2116 iter/s, 3.011s/100 iters), loss = 0.00802709
I1228 18:00:29.782099 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00802709 (* 1 = 0.00802709 loss)
I1228 18:00:29.782107 2733998976 sgd_solver.cpp:112] Iteration 6900, lr = 0.0067466
I1228 18:00:32.752030 2733998976 solver.cpp:347] Iteration 7000, Testing net (#0)
I1228 18:00:34.922893 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:00:35.012979 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9898
I1228 18:00:35.013005 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0291224 (* 1 = 0.0291224 loss)
I1228 18:00:35.041568 2733998976 solver.cpp:239] Iteration 7000 (19.015 iter/s, 5.259s/100 iters), loss = 0.00646155
I1228 18:00:35.041594 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00646155 (* 1 = 0.00646155 loss)
I1228 18:00:35.041602 2733998976 sgd_solver.cpp:112] Iteration 7000, lr = 0.00671681
I1228 18:00:37.286623 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:00:40.305764 2733998976 solver.cpp:239] Iteration 7100 (18.997 iter/s, 5.264s/100 iters), loss = 0.00939924
I1228 18:00:40.305794 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00939924 (* 1 = 0.00939924 loss)
I1228 18:00:40.305802 2733998976 sgd_solver.cpp:112] Iteration 7100, lr = 0.00668733
I1228 18:00:43.588579 2733998976 solver.cpp:239] Iteration 7200 (30.4692 iter/s, 3.282s/100 iters), loss = 0.00617863
I1228 18:00:43.588609 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00617863 (* 1 = 0.00617863 loss)
I1228 18:00:43.588618 2733998976 sgd_solver.cpp:112] Iteration 7200, lr = 0.00665815
I1228 18:00:46.674247 2733998976 solver.cpp:239] Iteration 7300 (32.4149 iter/s, 3.085s/100 iters), loss = 0.0201554
I1228 18:00:46.674273 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0201554 (* 1 = 0.0201554 loss)
I1228 18:00:46.674281 2733998976 sgd_solver.cpp:112] Iteration 7300, lr = 0.00662927
I1228 18:00:49.679594 2733998976 solver.cpp:239] Iteration 7400 (33.2779 iter/s, 3.005s/100 iters), loss = 0.00633155
I1228 18:00:49.679622 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00633155 (* 1 = 0.00633155 loss)
I1228 18:00:49.679630 2733998976 sgd_solver.cpp:112] Iteration 7400, lr = 0.00660067
I1228 18:00:52.749370 210243584 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:00:52.866267 2733998976 solver.cpp:347] Iteration 7500, Testing net (#0)
I1228 18:00:55.053150 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:00:55.141665 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9894
I1228 18:00:55.141692 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0324159 (* 1 = 0.0324159 loss)
I1228 18:00:55.171073 2733998976 solver.cpp:239] Iteration 7500 (18.2116 iter/s, 5.491s/100 iters), loss = 0.00143481
I1228 18:00:55.171102 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00143481 (* 1 = 0.00143481 loss)
I1228 18:00:55.171108 2733998976 sgd_solver.cpp:112] Iteration 7500, lr = 0.00657236
I1228 18:00:57.504627 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:01:00.629341 2733998976 solver.cpp:239] Iteration 7600 (18.3217 iter/s, 5.458s/100 iters), loss = 0.00504313
I1228 18:01:00.629370 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00504313 (* 1 = 0.00504313 loss)
I1228 18:01:00.629377 2733998976 sgd_solver.cpp:112] Iteration 7600, lr = 0.00654433
I1228 18:01:03.611757 2733998976 solver.cpp:239] Iteration 7700 (33.5345 iter/s, 2.982s/100 iters), loss = 0.0370426
I1228 18:01:03.611785 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0370426 (* 1 = 0.0370426 loss)
I1228 18:01:03.611793 2733998976 sgd_solver.cpp:112] Iteration 7700, lr = 0.00651658
I1228 18:01:06.663183 2733998976 solver.cpp:239] Iteration 7800 (32.7761 iter/s, 3.051s/100 iters), loss = 0.00482317
I1228 18:01:06.663211 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00482317 (* 1 = 0.00482317 loss)
I1228 18:01:06.663219 2733998976 sgd_solver.cpp:112] Iteration 7800, lr = 0.00648911
I1228 18:01:09.659461 2733998976 solver.cpp:239] Iteration 7900 (33.3778 iter/s, 2.996s/100 iters), loss = 0.00459809
I1228 18:01:09.659488 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00459809 (* 1 = 0.00459809 loss)
I1228 18:01:09.659497 2733998976 sgd_solver.cpp:112] Iteration 7900, lr = 0.0064619
I1228 18:01:12.696130 2733998976 solver.cpp:347] Iteration 8000, Testing net (#0)
I1228 18:01:14.910470 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:01:14.999799 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9903
I1228 18:01:14.999821 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0296206 (* 1 = 0.0296206 loss)
I1228 18:01:15.030339 2733998976 solver.cpp:239] Iteration 8000 (18.622 iter/s, 5.37s/100 iters), loss = 0.00708911
I1228 18:01:15.030365 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00708911 (* 1 = 0.00708911 loss)
I1228 18:01:15.030373 2733998976 sgd_solver.cpp:112] Iteration 8000, lr = 0.00643496
I1228 18:01:17.349162 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:01:20.411386 2733998976 solver.cpp:239] Iteration 8100 (18.5839 iter/s, 5.381s/100 iters), loss = 0.0157724
I1228 18:01:20.411415 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0157724 (* 1 = 0.0157724 loss)
I1228 18:01:20.411423 2733998976 sgd_solver.cpp:112] Iteration 8100, lr = 0.00640827
I1228 18:01:23.374837 2733998976 solver.cpp:239] Iteration 8200 (33.7496 iter/s, 2.963s/100 iters), loss = 0.00885533
I1228 18:01:23.374866 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00885533 (* 1 = 0.00885533 loss)
I1228 18:01:23.374873 2733998976 sgd_solver.cpp:112] Iteration 8200, lr = 0.00638185
I1228 18:01:26.310763 2733998976 solver.cpp:239] Iteration 8300 (34.0716 iter/s, 2.935s/100 iters), loss = 0.0312884
I1228 18:01:26.310792 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0312884 (* 1 = 0.0312884 loss)
I1228 18:01:26.310801 2733998976 sgd_solver.cpp:112] Iteration 8300, lr = 0.00635568
I1228 18:01:29.262581 2733998976 solver.cpp:239] Iteration 8400 (33.8868 iter/s, 2.951s/100 iters), loss = 0.00718827
I1228 18:01:29.262609 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00718827 (* 1 = 0.00718827 loss)
I1228 18:01:29.262615 2733998976 sgd_solver.cpp:112] Iteration 8400, lr = 0.00632975
I1228 18:01:30.248950 210243584 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:01:32.192843 2733998976 solver.cpp:347] Iteration 8500, Testing net (#0)
I1228 18:01:34.367149 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:01:34.458750 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9908
I1228 18:01:34.458778 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0294309 (* 1 = 0.0294309 loss)
I1228 18:01:34.487138 2733998976 solver.cpp:239] Iteration 8500 (19.1424 iter/s, 5.224s/100 iters), loss = 0.0072237
I1228 18:01:34.487166 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0072237 (* 1 = 0.0072237 loss)
I1228 18:01:34.487175 2733998976 sgd_solver.cpp:112] Iteration 8500, lr = 0.00630407
I1228 18:01:36.752040 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:01:39.947311 2733998976 solver.cpp:239] Iteration 8600 (18.315 iter/s, 5.46s/100 iters), loss = 0.000905046
I1228 18:01:39.947340 2733998976 solver.cpp:258]     Train net output #0: loss = 0.000905046 (* 1 = 0.000905046 loss)
I1228 18:01:39.947347 2733998976 sgd_solver.cpp:112] Iteration 8600, lr = 0.00627864
I1228 18:01:42.910595 2733998976 solver.cpp:239] Iteration 8700 (33.7496 iter/s, 2.963s/100 iters), loss = 0.00216248
I1228 18:01:42.910624 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00216248 (* 1 = 0.00216248 loss)
I1228 18:01:42.910630 2733998976 sgd_solver.cpp:112] Iteration 8700, lr = 0.00625344
I1228 18:01:46.290315 2733998976 solver.cpp:239] Iteration 8800 (29.5946 iter/s, 3.379s/100 iters), loss = 0.00139346
I1228 18:01:46.290344 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00139346 (* 1 = 0.00139346 loss)
I1228 18:01:46.290354 2733998976 sgd_solver.cpp:112] Iteration 8800, lr = 0.00622847
I1228 18:01:49.550997 2733998976 solver.cpp:239] Iteration 8900 (30.6748 iter/s, 3.26s/100 iters), loss = 0.00137645
I1228 18:01:49.551023 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00137645 (* 1 = 0.00137645 loss)
I1228 18:01:49.551031 2733998976 sgd_solver.cpp:112] Iteration 8900, lr = 0.00620374
I1228 18:01:52.662079 2733998976 solver.cpp:347] Iteration 9000, Testing net (#0)
I1228 18:01:54.892765 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:01:54.987535 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9894
I1228 18:01:54.987560 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0305774 (* 1 = 0.0305774 loss)
I1228 18:01:55.016005 2733998976 solver.cpp:239] Iteration 9000 (18.3016 iter/s, 5.464s/100 iters), loss = 0.0261417
I1228 18:01:55.016034 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0261417 (* 1 = 0.0261417 loss)
I1228 18:01:55.016042 2733998976 sgd_solver.cpp:112] Iteration 9000, lr = 0.00617924
I1228 18:01:57.243510 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:02:00.529062 2733998976 solver.cpp:239] Iteration 9100 (18.1389 iter/s, 5.513s/100 iters), loss = 0.00703448
I1228 18:02:00.529090 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00703448 (* 1 = 0.00703448 loss)
I1228 18:02:00.529098 2733998976 sgd_solver.cpp:112] Iteration 9100, lr = 0.00615496
I1228 18:02:03.612413 2733998976 solver.cpp:239] Iteration 9200 (32.4359 iter/s, 3.083s/100 iters), loss = 0.00244489
I1228 18:02:03.612445 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00244489 (* 1 = 0.00244489 loss)
I1228 18:02:03.612454 2733998976 sgd_solver.cpp:112] Iteration 9200, lr = 0.0061309
I1228 18:02:06.645978 2733998976 solver.cpp:239] Iteration 9300 (32.9707 iter/s, 3.033s/100 iters), loss = 0.0102975
I1228 18:02:06.646006 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0102975 (* 1 = 0.0102975 loss)
I1228 18:02:06.646013 2733998976 sgd_solver.cpp:112] Iteration 9300, lr = 0.00610706
I1228 18:02:08.742673 210243584 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:02:09.636730 2733998976 solver.cpp:239] Iteration 9400 (33.4448 iter/s, 2.99s/100 iters), loss = 0.0121169
I1228 18:02:09.636757 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0121169 (* 1 = 0.0121169 loss)
I1228 18:02:09.636765 2733998976 sgd_solver.cpp:112] Iteration 9400, lr = 0.00608343
I1228 18:02:12.576611 2733998976 solver.cpp:347] Iteration 9500, Testing net (#0)
I1228 18:02:14.810715 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:02:14.900460 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9885
I1228 18:02:14.900486 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0348806 (* 1 = 0.0348806 loss)
I1228 18:02:14.929896 2733998976 solver.cpp:239] Iteration 9500 (18.8929 iter/s, 5.293s/100 iters), loss = 0.0023389
I1228 18:02:14.929924 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0023389 (* 1 = 0.0023389 loss)
I1228 18:02:14.929931 2733998976 sgd_solver.cpp:112] Iteration 9500, lr = 0.00606002
I1228 18:02:17.234755 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:02:20.608641 2733998976 solver.cpp:239] Iteration 9600 (17.6118 iter/s, 5.678s/100 iters), loss = 0.0030517
I1228 18:02:20.608669 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0030517 (* 1 = 0.0030517 loss)
I1228 18:02:20.608676 2733998976 sgd_solver.cpp:112] Iteration 9600, lr = 0.00603682
I1228 18:02:23.608698 2733998976 solver.cpp:239] Iteration 9700 (33.3333 iter/s, 3s/100 iters), loss = 0.0030205
I1228 18:02:23.608727 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0030205 (* 1 = 0.0030205 loss)
I1228 18:02:23.608734 2733998976 sgd_solver.cpp:112] Iteration 9700, lr = 0.00601382
I1228 18:02:26.671052 2733998976 solver.cpp:239] Iteration 9800 (32.6584 iter/s, 3.062s/100 iters), loss = 0.0140991
I1228 18:02:26.671078 2733998976 solver.cpp:258]     Train net output #0: loss = 0.0140991 (* 1 = 0.0140991 loss)
I1228 18:02:26.671085 2733998976 sgd_solver.cpp:112] Iteration 9800, lr = 0.00599102
I1228 18:02:29.744645 2733998976 solver.cpp:239] Iteration 9900 (32.5415 iter/s, 3.073s/100 iters), loss = 0.00612791
I1228 18:02:29.744678 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00612791 (* 1 = 0.00612791 loss)
I1228 18:02:29.744686 2733998976 sgd_solver.cpp:112] Iteration 9900, lr = 0.00596843
I1228 18:02:32.713112 2733998976 solver.cpp:464] Snapshotting to binary proto file /Users/taily/pycharmproj/lenet_solver_test_show_iter_10000.caffemodel
I1228 18:02:32.715968 2733998976 sgd_solver.cpp:284] Snapshotting solver state to binary proto file /Users/taily/pycharmproj/lenet_solver_test_show_iter_10000.solverstate
I1228 18:02:32.717589 2733998976 solver.cpp:347] Iteration 10000, Testing net (#0)
I1228 18:02:34.903210 210780160 data_layer.cpp:73] Restarting data prefetching from start.
I1228 18:02:34.991242 2733998976 solver.cpp:414]     Test net output #0: accuracy = 0.9909
I1228 18:02:34.991267 2733998976 solver.cpp:414]     Test net output #1: loss = 0.0293722 (* 1 = 0.0293722 loss)

plot the train loss and test accuracy

I1228 18:02:35.020229 2733998976 solver.cpp:239] Iteration 10000 (18.9573 iter/s, 5.275s/100 iters), loss = 0.00305677
I1228 18:02:35.020256 2733998976 solver.cpp:258]     Train net output #0: loss = 0.00305677 (* 1 = 0.00305677 loss)
I1228 18:02:35.020264 2733998976 sgd_solver.cpp:112] Iteration 10000, lr = 0.00594604

Process finished with exit code 0

 

 

 

Taily老段
关注
专栏目录
初级入门：基于Caffe的神经网络理论与实战
u014191607的博客
05-16 761
目录： 2基于caffe的神经网络基础知识与实战 2.1 问题定义 2.2 初识神经网络 2.2.1 神经元 2.2.2 前馈过程 2.2.3 反向过程 2.3 典型CNN网络——LeNet 2.3.1 深度卷积神经网络结构 2.3.1.1 卷积层 2.3.1.2 池化层 2.3.1.3 全连接层 2.3.1.4 激活层 2.3.2 前向传播过...
yolov3训练自己的数据集——第一次实操完整记录
热门推荐
夏日山间过
07-09 7万+
参考： darknet yolo源码解析 bacth参数对性能影响 小知识点 make -j4 　表示用４条线程编译（对应自己的电脑，看自己电脑cpu几核几线程） make -j32　用32线程编译 darknet源码解析 darknet概述 darknet是一个轻型的、完全基于C与CUDA的开源深度学习框架，支持CPU和GPU两种计算方式。 darknet优点 完全由c语言实现，没有任何依...
Caffe日志文件可视化分析
01-28
可代码是用QT，基于QCHART开发的，CAFFE日志的可视化分析工具，可以同时画出train loss、test loss、test accuracy
caffe训练过程中的可视化
haoji007的博客
07-21 534
阅读数：1203   import matplotlib.pyplot as plt import caffe caffe.set_device(0) caffe.set_mode_gpu() # 使用SGDSolver，即随机梯度下降算法 solver = caffe.SGDSolver('/home/xxx/mnist/solve...
caffe绘制训练过程的loss和accuracy曲线
weixin_33716557的博客
08-30 278
转自：http://blog.csdn.net/u013078356/article/details/51154847 在caffe的训练过程中，大家难免想图形化自己的训练数据，以便更好的展示结果。如 果自己写代码记录训练过程的数据，那就太麻烦了，caffe中其实已经自带了这样的小工具 caffe-master/tools/extra/parse_log.sh  caffe-master/too...
内涵：caffe之学习曲线可视化
算法工程师进阶之路
06-24 636
一、log文件的获取 默认，我们在终端上输入模型训练指令，训练的log信息会输出在标准输出也即屏幕上，我们要想查看某一步的信息需要滚动滑动条来查看，在终端中查看。可通过下面的方式来将log信息以文件的形式保存到指定位置。 自己写一个脚本文件，把训练的各种参数都加在里面。 #!/usr/bin/env sh set -e TOOLS=./build/tools $TOOLS/caffe
Openpose实操（一）
qq_43258953的博客
02-25 2081
一，OpenPose项目Github链接：https://github.com/CMU-Perceptual-Computing-Lab/openpose 二，windows源码下载地址： 源码下载 下载第一个推荐版。 三，下载训练好的caffe模型文件 百度云盘上供大家下载使用： 链接：https://pan.baidu.com/s/1qNB-AZFRL0JeTAmAL_xPSA 密码：t9u...
深度神经网络训练
Supermen333的博客
09-16 521
按我们的学习经验，从一个数据源开始——即使是用最传统、已经应用多年的机器学习算法，先完整地走完机器学习的整个工作流程，不断尝试各种算法深挖这些数据的价值，在运用过程中把数据、特征和算法搞透，真正积累出项目经验，才能更快、更靠谱的掌握深度学习技术。众所周知，深度学习是一个将理论算法与计算机工程技术紧密结合的领域。我们已经在 indico 的产品或者开发中使用了以下列表中的大部分的技术，但是对于剩下一些我们没有使用的，我将会借鉴他人的经验来帮助给出 Python 深度学习生态系统的清晰的、详尽的理解。
使用MindStudio实现AutoML模型自动生成
最新发布
weixin_43871400的博客
12-14 505
使用MindStudio实现AutoML模型自动生成
Caffe实践 - 基于VGG16 多标签分类的训练与部署
长风破浪会有时，直挂云帆济沧海
08-25 4995
基于Caffe VGG16 的多标签分类 这里采用从图像直接读取图片数据和标签的方式进行多标签分类. 1. 问题描述 假定每张图片具有 N 个标签(本文N=3)，分别为 label1,label2,label3,...,labelNlabel1,label2,label3,...,labelN{label1, label2, label3,...,labelN}. 数据集 tr...
caffe训练网络的实践
leo_is_ant的专栏
10-16 2761
Caffe现如今已经成为了深度学习一个非常火的框架。但是作为刚接触CNN或者caffe的小白可能对caffe的接口或者细节不是很清楚了解。本文作为尝试着利用caffe做一个简单的图片分类model，希望能够与大家快速入门caffe训练属于自己的数据。   首先，要准备自己的数据集。这次任务是训练手写的文本。请注意，并不是手写集Mnist，区别在于是手写完整的一页作为一类。一共有6类。肉眼看区别很
【caffe】可视化网络的训练过程的loss和accuracy（效果优于官方绘制脚本）
chen1234520nnn的博客
10-28 712
前言：我们在训练网络的时候，网络训练完毕后，需要可视化训练过程中loss和accuracy的变化情况，方便了解整个训练过程模型的优化情况。（之前使用caffe自带的脚本绘制曲线图，出现各种报错，遂采用python脚本来绘制曲线。） 1、保存训练过程的log。 我们在执行训练指令时，可以通过以下方式保存训练日志。 #这是sh指令，最后一句为设置保存训练日志 /data_1/caffe-mas...
基于Caffe的深度学习实践
San_Junipero的博客
01-31 5560
原网页引自：A Practical Introduction to Deep Learning with Caffe and Python 0.Introduction 这是一篇使用CNN进行分类的文章，数据集取自Kaggle-dog-cat。博客内容分为两部分：一、深度学习下的部分核心概念；二、实践教程。 实践教程也分为两部分：一、Section4部分，对数据从头学习CNN网络进行猫和
Caffe训练自己的数据之图片集的处理（根据前辈们的分享整理并自己实践的过程）
CrystalCYT的博客
04-30 1281
Caffe训练自己的数据之图片集的处理（根据前辈们的分享整理并自己实践的过程，windows）一、将收集好的数据先分类存放二、resize所有图片为256*256（matlab代码，记得更改为自己的路径）clear;clc;close all; tic; % 用于计算程序运行时间，和toc搭配使用 imagePath = 'C:\Users\sony\Documents\M
Caffe 官方demo训练学习-CIFAR-10
六月的羽_七月未央
08-02 569
按照git中最简单的CIFAR-10 操作过程 进入caffe根目录 下载数据集并建立模型 cd $CAFFE_ROOT ./data/cifar10/get_cifar10.sh ./examples/cifar10/create_cifar10.sh 训练 cd $CAFFE_ROOT ./examples/cifar10/train_quick.sh 训练完成后如...
caffe训练的实时可视化思路
XiaoY
01-15 6427
个人认为，可视化是能够增进对网络理解的强有力的方式，但是问过一个相关工作者，貌似只对loss曲线感兴趣。 caffe的对外接口其实很丰富，包括了matlab和python，虽然这个框架还存在一些bug（比如对错误的deploy文件的容错还有待提高，不过这锅貌似也不能caffe背），但已经够用了，在此感谢作者！最近比较忙，所以没有时间去实现了，只能暂时在这边构思一下~
Caffe研究实践 二 ------准备数据 训练 测试
计算机视觉算法、工程、产品
05-11 4992
准备数据 训练 测试
从零开始训练自己的数据模型超详细教程（以caffenet为例）带你熟悉整个过程（数据准备，均值文件，训练，测试）
风中落泪的花
04-22 4781
背景：我要解决的是一个二分类问题，分类出正脸和非正脸，从而过滤人脸质量较差的图片，如低头，侧脸或者是戴口罩的人脸，而分类出正脸中对正脸进行一个评分。好的，下面正式进入实验过程。 （1）模型准备** 如果你没有还没有装好caffe,那请参考： 1.caffe安装系列——史上最详细的安装步骤 2，Ubuntu 16.04+CUDA8.0+Caffe安装教程 3.windows...
"Caffe深度学习训练全过程及应用实例
Caffe的深度学习训练全过程可以分为以下几个子步骤： 1. 数据预处理：首先需要准备数据，Caffe使用的是自己的数据库格式LMDB（Lightning Memory-Mapped Database），因此需要将数据转换成这种格式。通常可以先将...
评论 4
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