首先安装好显卡----已经装好了？普大喜奔！没装好？那就用cpu，，也是一样的。
拷贝cudnn v5.0 头文件和库文件以及执行文件到cuda8中
-----------------------------准备工作--------------------------------------
git clone https://github.com/BVLC/caffe.git
git branch -a
git checkout windows
cmake-gui
configure + vs2013 Win64
修改设置atlas选项为 open
build-malab on
检查numpy是否安装，是否配置正确
configure
generate
----------------------------build---------------------------------------
choose release mode
build all
reload and build all
build-install
set to path
-----------------------------wget----------------------------------------------
wget --no-check-certificate ~kriz/cifar-10-binary.tar.gz
./get_cifar10.sh （只执行解压缩那部分的语句，其余的不要。）即：
tar -xvf cifar-10-binary.tar.gz && rm -f cifar-10-binary.tar.gz
mv cifar-10-batches-bin/* . && rm -rf cifar-10-batches-bin
（放在了example目录下了）
====================================华丽的分割线--摘要===========================================
convert_cifar_data.exe data/cifar10 examples/cifar10 lmdb
compute_image_mean.exe -backend=lmdb examples/cifar10/cifar10_train_lmdb examples/cifar10/mean.binaryproto
caffe train --solver=examples/cifar10/cifar10_quick_solver.prototxt
caffe train --solver=examples/cifar10/cifar10_quick_solver_lr1.prototxt --snapshot=examples/cifar10/cifar10_quick_iter_4000.solverstate
caffe test -model examples/cifar10/cifar10_quick_train_test.prototxt -weights examples/cifar10/cifar10_quick_iter_5000.caffemodel.h5 -iterations 100
---------------------------------------
synset_words.txt 的内容如下：
airplane
automobile
bird
cat
deer
dog
frog
horse
ship
truck
===================================超级华丽的分割线1============================================
===================================超级华丽的分割线2============================================
===================================超级华丽的分割线3============================================
===================================超级华丽的分割线4============================================
======================================= 详情 ===============================================
convert_cifar_data.exe data/cifar10 examples/cifar10 lmdb
compute_image_mean.exe -backend=lmdb examples/cifar10/cifar10_train_lmdb examples/cifar10/mean.binaryproto
caffe.exe train --solver=examples/cifar10/cifar10_quick_solver.prototxt
I0703 15:08:42.303948 73968 net.cpp:137] Memory required for data: 1230000
I0703 15:08:42.303948 73968 layer_factory.cpp:58] Creating layer conv1
I0703 15:08:42.303948 73968 net.cpp:84] Creating Layer conv1
I0703 15:08:42.303948 73968 net.cpp:406] conv1 <- data
I0703 15:08:42.303948 73968 net.cpp:380] conv1 -> conv1
I0703 15:08:42.303948 73968 net.cpp:122] Setting up conv1
I0703 15:08:42.303948 73968 net.cpp:129] Top shape: 100 32 32 32 (3276800)
I0703 15:08:42.319548 73968 net.cpp:137] Memory required for data: 14337200
I0703 15:08:42.319548 73968 layer_factory.cpp:58] Creating layer pool1
I0703 15:08:42.319548 73968 net.cpp:84] Creating Layer pool1
I0703 15:08:42.319548 73968 net.cpp:406] pool1 <- conv1
I0703 15:08:42.319548 73968 net.cpp:380] pool1 -> pool1
I0703 15:08:42.319548 73968 net.cpp:122] Setting up pool1
I0703 15:08:42.319548 73968 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0703 15:08:42.319548 73968 net.cpp:137] Memory required for data: 17614000
I0703 15:08:42.319548 73968 layer_factory.cpp:58] Creating layer relu1
I0703 15:08:42.319548 73968 net.cpp:84] Creating Layer relu1
I0703 15:08:42.319548 73968 net.cpp:406] relu1 <- pool1
I0703 15:08:42.319548 73968 net.cpp:367] relu1 -> pool1 (in-place)
I0703 15:08:42.319548 73968 net.cpp:122] Setting up relu1
I0703 15:08:42.319548 73968 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0703 15:08:42.319548 73968 net.cpp:137] Memory required for data: 20890800
I0703 15:08:42.319548 73968 layer_factory.cpp:58] Creating layer conv2
I0703 15:08:42.319548 73968 net.cpp:84] Creating Layer conv2
I0703 15:08:42.319548 73968 net.cpp:406] conv2 <- pool1
I0703 15:08:42.319548 73968 net.cpp:380] conv2 -> conv2
I0703 15:08:42.319548 73968 net.cpp:122] Setting up conv2
I0703 15:08:42.319548 73968 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0703 15:08:42.319548 73968 net.cpp:137] Memory required for data: 24167600
I0703 15:08:42.319548 73968 layer_factory.cpp:58] Creating layer relu2
I0703 15:08:42.319548 73968 net.cpp:84] Creating Layer relu2
I0703 15:08:42.319548 73968 net.cpp:406] relu2 <- conv2
I0703 15:08:42.319548 73968 net.cpp:367] relu2 -> conv2 (in-place)
I0703 15:08:42.335150 73968 net.cpp:122] Setting up relu2
I0703 15:08:42.335150 73968 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0703 15:08:42.335150 73968 net.cpp:137] Memory required for data: 27444400
I0703 15:08:42.335150 73968 layer_factory.cpp:58] Creating layer pool2
I0703 15:08:42.335150 73968 net.cpp:84] Creating Layer pool2
I0703 15:08:42.335150 73968 net.cpp:406] pool2 <- conv2
I0703 15:08:42.335150 73968 net.cpp:380] pool2 -> pool2
I0703 15:08:42.335150 73968 net.cpp:122] Setting up pool2
I0703 15:08:42.335150 73968 net.cpp:129] Top shape: 100 32 8 8 (204800)
I0703 15:08:42.335150 73968 net.cpp:137] Memory required for data: 28263600
I0703 15:08:42.335150 73968 layer_factory.cpp:58] Creating layer conv3
I0703 15:08:42.335150 73968 net.cpp:84] Creating Layer conv3
I0703 15:08:42.335150 73968 net.cpp:406] conv3 <- pool2
I0703 15:08:42.335150 73968 net.cpp:380] conv3 -> conv3
I0703 15:08:42.335150 73968 net.cpp:122] Setting up conv3
I0703 15:08:42.335150 73968 net.cpp:129] Top shape: 100 64 8 8 (409600)
I0703 15:08:42.335150 73968 net.cpp:137] Memory required for data: 29902000
I0703 15:08:42.335150 73968 layer_factory.cpp:58] Creating layer relu3
I0703 15:08:42.335150 73968 net.cpp:84] Creating Layer relu3
I0703 15:08:42.335150 73968 net.cpp:406] relu3 <- conv3
I0703 15:08:42.335150 73968 net.cpp:367] relu3 -> conv3 (in-place)
I0703 15:08:42.335150 73968 net.cpp:122] Setting up relu3
I0703 15:08:42.335150 73968 net.cpp:129] Top shape: 100 64 8 8 (409600)
I0703 15:08:42.335150 73968 net.cpp:137] Memory required for data: 31540400
I0703 15:08:42.335150 73968 layer_factory.cpp:58] Creating layer pool3
I0703 15:08:42.350749 73968 net.cpp:84] Creating Layer pool3
I0703 15:08:42.350749 73968 net.cpp:406] pool3 <- conv3
I0703 15:08:42.350749 73968 net.cpp:380] pool3 -> pool3
I0703 15:08:42.350749 73968 net.cpp:122] Setting up pool3
I0703 15:08:42.350749 73968 net.cpp:129] Top shape: 100 64 4 4 (102400)
I0703 15:08:42.350749 73968 net.cpp:137] Memory required for data: 31950000
I0703 15:08:42.350749 73968 layer_factory.cpp:58] Creating layer ip1
I0703 15:08:42.350749 73968 net.cpp:84] Creating Layer ip1
I0703 15:08:42.350749 73968 net.cpp:406] ip1 <- pool3
I0703 15:08:42.350749 73968 net.cpp:380] ip1 -> ip1
I0703 15:08:42.350749 73968 net.cpp:122] Setting up ip1
I0703 15:08:42.350749 73968 net.cpp:129] Top shape: 100 64 (6400)
I0703 15:08:42.350749 73968 net.cpp:137] Memory required for data: 31975600
I0703 15:08:42.350749 73968 layer_factory.cpp:58] Creating layer ip2
I0703 15:08:42.350749 73968 net.cpp:84] Creating Layer ip2
I0703 15:08:42.350749 73968 net.cpp:406] ip2 <- ip1
I0703 15:08:42.350749 73968 net.cpp:380] ip2 -> ip2
I0703 15:08:42.350749 73968 net.cpp:122] Setting up ip2
I0703 15:08:42.350749 73968 net.cpp:129] Top shape: 100 10 (1000)
I0703 15:08:42.350749 73968 net.cpp:137] Memory required for data: 31979600
I0703 15:08:42.350749 73968 layer_factory.cpp:58] Creating layer ip2_ip2_0_split
I0703 15:08:42.350749 73968 net.cpp:84] Creating Layer ip2_ip2_0_split
I0703 15:08:42.350749 73968 net.cpp:406] ip2_ip2_0_split <- ip2
I0703 15:08:42.350749 73968 net.cpp:380] ip2_ip2_0_split -> ip2_ip2_0_split_0
I0703 15:08:42.350749 73968 net.cpp:380] ip2_ip2_0_split -> ip2_ip2_0_split_1
I0703 15:08:42.350749 73968 net.cpp:122] Setting up ip2_ip2_0_split
I0703 15:08:42.350749 73968 net.cpp:129] Top shape: 100 10 (1000)
I0703 15:08:42.350749 73968 net.cpp:129] Top shape: 100 10 (1000)
I0703 15:08:42.350749 73968 net.cpp:137] Memory required for data: 31987600
I0703 15:08:42.350749 73968 layer_factory.cpp:58] Creating layer accuracy
I0703 15:08:42.350749 73968 net.cpp:84] Creating Layer accuracy
I0703 15:08:42.350749 73968 net.cpp:406] accuracy <- ip2_ip2_0_split_0
I0703 15:08:42.350749 73968 net.cpp:406] accuracy <- label_cifar_1_split_0
I0703 15:08:42.366350 73968 net.cpp:380] accuracy -> accuracy
I0703 15:08:42.366350 73968 net.cpp:122] Setting up accuracy
I0703 15:08:42.366350 73968 net.cpp:129] Top shape: (1)
I0703 15:08:42.366350 73968 net.cpp:137] Memory required for data: 31987604
I0703 15:08:42.366350 73968 layer_factory.cpp:58] Creating layer loss
I0703 15:08:42.366350 73968 net.cpp:84] Creating Layer loss
I0703 15:08:42.366350 73968 net.cpp:406] loss <- ip2_ip2_0_split_1
I0703 15:08:42.366350 73968 net.cpp:406] loss <- label_cifar_1_split_1
I0703 15:08:42.366350 73968 net.cpp:380] loss -> loss
I0703 15:08:42.366350 73968 layer_factory.cpp:58] Creating layer loss
I0703 15:08:42.366350 73968 net.cpp:122] Setting up loss
I0703 15:08:42.366350 73968 net.cpp:129] Top shape: (1)
I0703 15:08:42.366350 73968 net.cpp:132]
with loss weight 1
I0703 15:08:42.366350 73968 net.cpp:137] Memory required for data: 31987608
I0703 15:08:42.366350 73968 net.cpp:198] loss needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:200] accuracy does not need backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] ip2_ip2_0_split needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] ip2 needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] ip1 needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] pool3 needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] relu3 needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] conv3 needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] pool2 needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] relu2 needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] conv2 needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] relu1 needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] pool1 needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:198] conv1 needs backward computation.
I0703 15:08:42.366350 73968 net.cpp:200] label_cifar_1_split does not need backward computation.
I0703 15:08:42.366350 73968 net.cpp:200] cifar does not need backward computation.
I0703 15:08:42.366350 73968 net.cpp:242] This network produces output accuracy
I0703 15:08:42.366350 73968 net.cpp:242] This network produces output loss
I0703 15:08:42.366350 73968 net.cpp:255] Network initialization done.
I0703 15:08:42.366350 73968 solver.cpp:56] Solver scaffolding done.
I0703 15:08:42.366350 73968 caffe.cpp:249] Starting Optimization
I0703 15:08:42.366350 73968 solver.cpp:272] Solving CIFAR10_quick
I0703 15:08:42.366350 73968 solver.cpp:273] Learning Rate Policy: fixed
I0703 15:08:42.413151 73968 solver.cpp:330] Iteration 0, Testing net (#0)
I0703 15:08:43.629990 70280 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:08:43.661191 73968 solver.cpp:397]
Test net output #0: accuracy = 0.0986
I0703 15:08:43.661191 73968 solver.cpp:397]
Test net output #1: loss = 2.30244 (* 1 = 2.30244 loss)
I0703 15:08:43.707993 73968 solver.cpp:218] Iteration 0 (-5.42863e-042 iter/s, 1.29273s/100 iters), loss = 2.30274
I0703 15:08:43.707993 73968 solver.cpp:237]
Train net output #0: loss = 2.30274 (* 1 = 2.30274 loss)
I0703 15:08:43.707993 73968 sgd_solver.cpp:105] Iteration 0, lr = 0.001
I0703 15:08:46.640887 73968 solver.cpp:218] Iteration 100 (34.1913 iter/s, 2.92472s/100 iters), loss = 1.65303
I0703 15:08:46.640887 73968 solver.cpp:237]
Train net output #0: loss = 1.65303 (* 1 = 1.65303 loss)
I0703 15:08:46.640887 73968 sgd_solver.cpp:105] Iteration 100, lr = 0.001
I0703 15:08:49.563181 73968 solver.cpp:218] Iteration 200 (34.1799 iter/s, 2.9257s/100 iters), loss = 1.60865
I0703 15:08:49.563181 73968 solver.cpp:237]
Train net output #0: loss = 1.60865 (* 1 = 1.60865 loss)
I0703 15:08:49.563181 73968 sgd_solver.cpp:105] Iteration 200, lr = 0.001
I0703 15:08:52.480474 73968 solver.cpp:218] Iteration 300 (34.2469 iter/s, 2.91998s/100 iters), loss = 1.185
I0703 15:08:52.480474 73968 solver.cpp:237]
Train net output #0: loss = 1.185 (* 1 = 1.185 loss)
I0703 15:08:52.480474 73968 sgd_solver.cpp:105] Iteration 300, lr = 0.001
I0703 15:08:55.428969 73968 solver.cpp:218] Iteration 400 (34.0078 iter/s, 2.9405s/100 iters), loss = 1.22841
I0703 15:08:55.428969 73968 solver.cpp:237]
Train net output #0: loss = 1.22841 (* 1 = 1.22841 loss)
I0703 15:08:55.428969 73968 sgd_solver.cpp:105] Iteration 400, lr = 0.001
I0703 15:08:58.253659 71148 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:08:58.347262 73968 solver.cpp:330] Iteration 500, Testing net (#0)
I0703 15:08:59.486099 70280 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:08:59.517300 73968 solver.cpp:397]
Test net output #0: accuracy = 0.5535
I0703 15:08:59.517300 73968 solver.cpp:397]
Test net output #1: loss = 1.2742 (* 1 = 1.2742 loss)
I0703 15:08:59.548501 73968 solver.cpp:218] Iteration 500 (24.261 iter/s, 4.12185s/100 iters), loss = 1.23252
I0703 15:08:59.548501 73968 solver.cpp:237]
Train net output #0: loss = 1.23252 (* 1 = 1.23252 loss)
I0703 15:08:59.548501 73968 sgd_solver.cpp:105] Iteration 500, lr = 0.001
I0703 15:09:02.512596 73968 solver.cpp:218] Iteration 600 (33.7982 iter/s, 2.95873s/100 iters), loss = 1.25093
I0703 15:09:02.512596 73968 solver.cpp:237]
Train net output #0: loss = 1.25093 (* 1 = 1.25093 loss)
I0703 15:09:02.512596 73968 sgd_solver.cpp:105] Iteration 600, lr = 0.001
I0703 15:09:05.493698 73968 solver.cpp:218] Iteration 700 (33.3689 iter/s, 2.99681s/100 iters), loss = 1.15823
I0703 15:09:05.494699 73968 solver.cpp:237]
Train net output #0: loss = 1.15823 (* 1 = 1.15823 loss)
I0703 15:09:05.494699 73968 sgd_solver.cpp:105] Iteration 700, lr = 0.001
I0703 15:09:08.483223 73968 solver.cpp:218] Iteration 800 (33.6528 iter/s, 2.97152s/100 iters), loss = 1.02646
I0703 15:09:08.483223 73968 solver.cpp:237]
Train net output #0: loss = 1.02646 (* 1 = 1.02646 loss)
I0703 15:09:08.483223 73968 sgd_solver.cpp:105] Iteration 800, lr = 0.001
I0703 15:09:11.447319 73968 solver.cpp:218] Iteration 900 (33.8917 iter/s, 2.95058s/100 iters), loss = 1.09516
I0703 15:09:11.447319 73968 solver.cpp:237]
Train net output #0: loss = 1.09516 (* 1 = 1.09516 loss)
I0703 15:09:11.447319 73968 sgd_solver.cpp:105] Iteration 900, lr = 0.001
I0703 15:09:14.285416 71148 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:09:14.394620 73968 solver.cpp:330] Iteration 1000, Testing net (#0)
I0703 15:09:15.551057 70280 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:09:15.597858 73968 solver.cpp:397]
Test net output #0: accuracy = 0.6325
I0703 15:09:15.597858 73968 solver.cpp:397]
Test net output #1: loss = 1.05842 (* 1 = 1.05842 loss)
I0703 15:09:15.629060 73968 solver.cpp:218] Iteration 1000 (23.9025 iter/s, 4.18367s/100 iters), loss = 0.983432
I0703 15:09:15.629060 73968 solver.cpp:237]
Train net output #0: loss = 0.983432 (* 1 = 0.983432 loss)
I0703 15:09:15.629060 73968 sgd_solver.cpp:105] Iteration 1000, lr = 0.001
I0703 15:09:18.599957 73968 solver.cpp:218] Iteration 1100 (33.8006 iter/s, 2.95853s/100 iters), loss = 1.06141
I0703 15:09:18.599957 73968 solver.cpp:237]
Train net output #0: loss = 1.06141 (* 1 = 1.06141 loss)
I0703 15:09:18.599957 73968 sgd_solver.cpp:105] Iteration 1100, lr = 0.001
I0703 15:09:19.317580 73968 blocking_queue.cpp:49] Waiting for data
I0703 15:09:21.548452 73968 solver.cpp:218] Iteration 1200 (33.9381 iter/s, 2.94654s/100 iters), loss = 0.950789
I0703 15:09:21.548452 73968 solver.cpp:237]
Train net output #0: loss = 0.950789 (* 1 = 0.950789 loss)
I0703 15:09:21.548452 73968 sgd_solver.cpp:105] Iteration 1200, lr = 0.001
I0703 15:09:24.512547 73968 solver.cpp:218] Iteration 1300 (33.7456 iter/s, 2.96335s/100 iters), loss = 0.845029
I0703 15:09:24.512547 73968 solver.cpp:237]
Train net output #0: loss = 0.845029 (* 1 = 0.845029 loss)
I0703 15:09:24.512547 73968 sgd_solver.cpp:105] Iteration 1300, lr = 0.001
I0703 15:09:27.538540 73968 solver.cpp:218] Iteration 1400 (33.0713 iter/s, 3.02377s/100 iters), loss = 0.854708
I0703 15:09:27.538540 73968 solver.cpp:237]
Train net output #0: loss = 0.854708 (* 1 = 0.854708 loss)
I0703 15:09:27.538540 73968 sgd_solver.cpp:105] Iteration 1400, lr = 0.001
I0703 15:09:30.405486 71148 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:09:30.499089 73968 solver.cpp:330] Iteration 1500, Testing net (#0)
I0703 15:09:31.642729 70280 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:09:31.689529 73968 solver.cpp:397]
Test net output #0: accuracy = 0.666
I0703 15:09:31.689529 73968 solver.cpp:397]
Test net output #1: loss = 0.965108 (* 1 = 0.965108 loss)
I0703 15:09:31.706130 73968 solver.cpp:218] Iteration 1500 (23.9048 iter/s, 4.18326s/100 iters), loss = 0.803068
I0703 15:09:31.706130 73968 solver.cpp:237]
Train net output #0: loss = 0.803068 (* 1 = 0.803068 loss)
I0703 15:09:31.706130 73968 sgd_solver.cpp:105] Iteration 1500, lr = 0.001
I0703 15:09:34.671032 73968 solver.cpp:218] Iteration 1600 (33.9581 iter/s, 2.94481s/100 iters), loss = 0.89545
I0703 15:09:34.671032 73968 solver.cpp:237]
Train net output #0: loss = 0.89545 (* 1 = 0.89545 loss)
I0703 15:09:34.671032 73968 sgd_solver.cpp:105] Iteration 1600, lr = 0.001
I0703 15:09:37.666788 73968 solver.cpp:218] Iteration 1700 (33.1725 iter/s, 3.01455s/100 iters), loss = 0.858076
I0703 15:09:37.667788 73968 solver.cpp:237]
Train net output #0: loss = 0.858076 (* 1 = 0.858076 loss)
I0703 15:09:37.667788 73968 sgd_solver.cpp:105] Iteration 1700, lr = 0.001
I0703 15:09:40.681776 73968 solver.cpp:218] Iteration 1800 (33.4777 iter/s, 2.98707s/100 iters), loss = 0.739417
I0703 15:09:40.681776 73968 solver.cpp:237]
Train net output #0: loss = 0.739417 (* 1 = 0.739417 loss)
I0703 15:09:40.681776 73968 sgd_solver.cpp:105] Iteration 1800, lr = 0.001
I0703 15:09:43.649132 73968 solver.cpp:218] Iteration 1900 (33.6842 iter/s, 2.96875s/100 iters), loss = 0.755557
I0703 15:09:43.649132 73968 solver.cpp:237]
Train net output #0: loss = 0.755557 (* 1 = 0.755557 loss)
I0703 15:09:43.649132 73968 sgd_solver.cpp:105] Iteration 1900, lr = 0.001
I0703 15:09:46.534909 71148 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:09:46.628511 73968 solver.cpp:330] Iteration 2000, Testing net (#0)
I0703 15:09:47.772374 70280 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:09:47.819176 73968 solver.cpp:397]
Test net output #0: accuracy = 0.7016
I0703 15:09:47.819176 73968 solver.cpp:397]
Test net output #1: loss = 0.878213 (* 1 = 0.878213 loss)
I0703 15:09:47.850378 73968 solver.cpp:218] Iteration 2000 (23.8369 iter/s, 4.19518s/100 iters), loss = 0.678401
I0703 15:09:47.850378 73968 solver.cpp:237]
Train net output #0: loss = 0.678401 (* 1 = 0.678401 loss)
I0703 15:09:47.850378 73968 sgd_solver.cpp:105] Iteration 2000, lr = 0.001
I0703 15:09:50.791185 73968 solver.cpp:218] Iteration 2100 (33.7445 iter/s, 2.96345s/100 iters), loss = 0.790343
I0703 15:09:50.792186 73968 solver.cpp:237]
Train net output #0: loss = 0.790343 (* 1 = 0.790343 loss)
I0703 15:09:50.792186 73968 sgd_solver.cpp:105] Iteration 2100, lr = 0.001
I0703 15:09:53.723479 73968 solver.cpp:218] Iteration 2200 (34.139 iter/s, 2.9292s/100 iters), loss = 0.792828
I0703 15:09:53.723479 73968 solver.cpp:237]
Train net output #0: loss = 0.792828 (* 1 = 0.792828 loss)
I0703 15:09:53.723479 73968 sgd_solver.cpp:105] Iteration 2200, lr = 0.001
I0703 15:09:56.723778 73968 solver.cpp:218] Iteration 2300 (33.3467 iter/s, 2.9988s/100 iters), loss = 0.634268
I0703 15:09:56.724778 73968 solver.cpp:237]
Train net output #0: loss = 0.634268 (* 1 = 0.634268 loss)
I0703 15:09:56.724778 73968 sgd_solver.cpp:105] Iteration 2300, lr = 0.001
I0703 15:09:59.751081 73968 solver.cpp:218] Iteration 2400 (33.0722 iter/s, 3.02369s/100 iters), loss = 0.730874
I0703 15:09:59.752081 73968 solver.cpp:237]
Train net output #0: loss = 0.730874 (* 1 = 0.730874 loss)
I0703 15:09:59.752081 73968 sgd_solver.cpp:105] Iteration 2400, lr = 0.001
I0703 15:10:02.642370 71148 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:10:02.746381 73968 solver.cpp:330] Iteration 2500, Testing net (#0)
I0703 15:10:03.921499 70280 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:10:03.966502 73968 solver.cpp:397]
Test net output #0: accuracy = 0.7126
I0703 15:10:03.966502 73968 solver.cpp:397]
Test net output #1: loss = 0.845339 (* 1 = 0.845339 loss)
I0703 15:10:03.996505 73968 solver.cpp:218] Iteration 2500 (23.57 iter/s, 4.24269s/100 iters), loss = 0.613872
I0703 15:10:03.997506 73968 solver.cpp:237]
Train net output #0: loss = 0.613872 (* 1 = 0.613872 loss)
I0703 15:10:03.997506 73968 sgd_solver.cpp:105] Iteration 2500, lr = 0.001
I0703 15:10:06.947801 73968 solver.cpp:218] Iteration 2600 (33.9124 iter/s, 2.94878s/100 iters), loss = 0.713529
I0703 15:10:06.948801 73968 solver.cpp:237]
Train net output #0: loss = 0.713529 (* 1 = 0.713529 loss)
I0703 15:10:06.949801 73968 sgd_solver.cpp:105] Iteration 2600, lr = 0.001
I0703 15:10:09.897095 73968 solver.cpp:218] Iteration 2700 (33.94 iter/s, 2.94638s/100 iters), loss = 0.749147
I0703 15:10:09.898097 73968 solver.cpp:237]
Train net output #0: loss = 0.749147 (* 1 = 0.749147 loss)
I0703 15:10:09.898097 73968 sgd_solver.cpp:105] Iteration 2700, lr = 0.001
I0703 15:10:12.863761 73968 solver.cpp:218] Iteration 2800 (33.7369 iter/s, 2.96411s/100 iters), loss = 0.570446
I0703 15:10:12.864761 73968 solver.cpp:237]
Train net output #0: loss = 0.570446 (* 1 = 0.570446 loss)
I0703 15:10:12.865762 73968 sgd_solver.cpp:105] Iteration 2800, lr = 0.001
I0703 15:10:15.809679 73968 solver.cpp:218] Iteration 2900 (33.9821 iter/s, 2.94273s/100 iters), loss = 0.713307
I0703 15:10:15.810678 73968 solver.cpp:237]
Train net output #0: loss = 0.713307 (* 1 = 0.713307 loss)
I0703 15:10:15.811678 73968 sgd_solver.cpp:105] Iteration 2900, lr = 0.001
I0703 15:10:18.606696 71148 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:10:18.704706 73968 solver.cpp:330] Iteration 3000, Testing net (#0)
I0703 15:10:19.830818 70280 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:10:19.873823 73968 solver.cpp:397]
Test net output #0: accuracy = 0.7197
I0703 15:10:19.874824 73968 solver.cpp:397]
Test net output #1: loss = 0.835526 (* 1 = 0.835526 loss)
I0703 15:10:19.903826 73968 solver.cpp:218] Iteration 3000 (24.4437 iter/s, 4.09103s/100 iters), loss = 0.57974
I0703 15:10:19.904826 73968 solver.cpp:237]
Train net output #0: loss = 0.57974 (* 1 = 0.57974 loss)
I0703 15:10:19.905827 73968 sgd_solver.cpp:105] Iteration 3000, lr = 0.001
I0703 15:10:22.878123 73968 solver.cpp:218] Iteration 3100 (33.657 iter/s, 2.97115s/100 iters), loss = 0.664127
I0703 15:10:22.879123 73968 solver.cpp:237]
Train net output #0: loss = 0.664127 (* 1 = 0.664127 loss)
I0703 15:10:22.879123 73968 sgd_solver.cpp:105] Iteration 3100, lr = 0.001
I0703 15:10:25.912427 73968 solver.cpp:218] Iteration 3200 (32.9824 iter/s, 3.03192s/100 iters), loss = 0.726144
I0703 15:10:25.913427 73968 solver.cpp:237]
Train net output #0: loss = 0.726144 (* 1 = 0.726144 loss)
I0703 15:10:25.914427 73968 sgd_solver.cpp:105] Iteration 3200, lr = 0.001
I0703 15:10:28.902726 73968 solver.cpp:218] Iteration 3300 (33.4837 iter/s, 2.98653s/100 iters), loss = 0.597564
I0703 15:10:28.902726 73968 solver.cpp:237]
Train net output #0: loss = 0.597564 (* 1 = 0.597564 loss)
I0703 15:10:28.903726 73968 sgd_solver.cpp:105] Iteration 3300, lr = 0.001
I0703 15:10:31.947031 73968 solver.cpp:218] Iteration 3400 (32.872 iter/s, 3.0421s/100 iters), loss = 0.663627
I0703 15:10:31.948030 73968 solver.cpp:237]
Train net output #0: loss = 0.663627 (* 1 = 0.663627 loss)
I0703 15:10:31.948030 73968 sgd_solver.cpp:105] Iteration 3400, lr = 0.001
I0703 15:10:34.875890 71148 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:10:34.976495 73968 solver.cpp:330] Iteration 3500, Testing net (#0)
I0703 15:10:36.121947 70280 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:10:36.185348 73968 solver.cpp:397]
Test net output #0: accuracy = 0.7187
I0703 15:10:36.185348 73968 solver.cpp:397]
Test net output #1: loss = 0.862067 (* 1 = 0.862067 loss)
I0703 15:10:36.216549 73968 solver.cpp:218] Iteration 3500 (23.5687 iter/s, 4.24291s/100 iters), loss = 0.560303
I0703 15:10:36.216549 73968 solver.cpp:237]
Train net output #0: loss = 0.560303 (* 1 = 0.560303 loss)
I0703 15:10:36.216549 73968 sgd_solver.cpp:105] Iteration 3500, lr = 0.001
I0703 15:10:39.186511 73968 solver.cpp:218] Iteration 3600 (33.6513 iter/s, 2.97165s/100 iters), loss = 0.619981
I0703 15:10:39.186511 73968 solver.cpp:237]
Train net output #0: loss = 0.619981 (* 1 = 0.619981 loss)
I0703 15:10:39.186511 73968 sgd_solver.cpp:105] Iteration 3600, lr = 0.001
I0703 15:10:42.190245 73968 solver.cpp:218] Iteration 3700 (33.4028 iter/s, 2.99376s/100 iters), loss = 0.798935
I0703 15:10:42.190245 73968 solver.cpp:237]
Train net output #0: loss = 0.798935 (* 1 = 0.798935 loss)
I0703 15:10:42.190245 73968 sgd_solver.cpp:105] Iteration 3700, lr = 0.001
I0703 15:10:45.185573 73968 solver.cpp:218] Iteration 3800 (33.3392 iter/s, 2.99947s/100 iters), loss = 0.563457
I0703 15:10:45.185573 73968 solver.cpp:237]
Train net output #0: loss = 0.563457 (* 1 = 0.563457 loss)
I0703 15:10:45.185573 73968 sgd_solver.cpp:105] Iteration 3800, lr = 0.001
I0703 15:10:48.180902 73968 solver.cpp:218] Iteration 3900 (33.4396 iter/s, 2.99047s/100 iters), loss = 0.629793
I0703 15:10:48.180902 73968 solver.cpp:237]
Train net output #0: loss = 0.629793 (* 1 = 0.629793 loss)
I0703 15:10:48.180902 73968 sgd_solver.cpp:105] Iteration 3900, lr = 0.001
I0703 15:10:51.034423 71148 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:10:51.128026 73968 solver.cpp:447] Snapshotting to binary proto file examples/cifar10/cifar10_quick_iter_4000.caffemodel
I0703 15:10:51.159227 73968 sgd_solver.cpp:273] Snapshotting solver state to binary proto file examples/cifar10/cifar10_quick_iter_4000.solverstate
I0703 15:10:51.174827 73968 solver.cpp:310] Iteration 4000, loss = 0.518084
I0703 15:10:51.174827 73968 solver.cpp:330] Iteration 4000, Testing net (#0)
I0703 15:10:52.338073 70280 data_layer.cpp:73] Restarting data prefetching from start.
I0703 15:10:52.378077 73968 solver.cpp:397]
Test net output #0: accuracy = 0.7254
I0703 15:10:52.379077 73968 solver.cpp:397]
Test net output #1: loss = 0.841199 (* 1 = 0.841199 loss)
I0703 15:10:52.380077 73968 solver.cpp:315] Optimization Done.
I0703 15:10:52.380077 73968 caffe.cpp:260] Optimization Done.
D:\ws_caffe\caffe>
-----------------------------------------------------------------------------------------------------------
caffe train --solver=examples/cifar10/cifar10_quick_solver_lr1.prototxt --snapshot=examples/cifar10/cifar10_quick_iter_4000.solverstate
caffe test -model examples/cifar10/cifar10_quick_train_test.prototxt -weights examples/cifar10/cifar10_quick_iter_5000.caffemodel.h5 -iterations 100