本例利用python学写caffe模型的prototxt。先载入一些必要的模块,把路径改成自己的:
import os
os.chdir('/home/lml/caffe-master/')
import sys
sys.path.insert(0, './python')
import caffe
from pylab import
%matplotlib inline
接下来运行给出的LeNet例子(确定已经下载了数据,转换好了数据格式):
# Download and prepare data
!data/mnist/get_mnist.sh
!examples/mnist/create_mnist.sh
Downloading…
–2015-09-21 09:11:03– http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz 正在解析主机
yann.lecun.com (yann.lecun.com)… 128.122.47.89 正在连接 yann.lecun.com
(yann.lecun.com)|128.122.47.89|:80… 已连接。 已发出 HTTP 请求,正在等待回应… 200
OK 长度: 9912422 (9.5M) [application/x-gzip] 正在保存至:
“train-images-idx3-ubyte.gz”
100%[======================================>] 9,912,422 1.39MB/s
用时 12s 2015-09-21 09:11:16 (775 KB/s) - 已保存
“train-images-idx3-ubyte.gz” [9912422/9912422])
–2015-09-21 09:11:16– http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz 正在解析主机
yann.lecun.com (yann.lecun.com)… 128.122.47.89 正在连接 yann.lecun.com
(yann.lecun.com)|128.122.47.89|:80… 已连接。 已发出 HTTP 请求,正在等待回应… 200
OK 长度: 28881 (28K) [application/x-gzip] 正在保存至:
“train-labels-idx1-ubyte.gz”
100%[======================================>] 28,881 47.4KB/s
用时 0.6s 2015-09-21 09:11:18 (47.4 KB/s) - 已保存
“train-labels-idx1-ubyte.gz” [28881/28881])
–2015-09-21 09:11:18– http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz 正在解析主机
yann.lecun.com (yann.lecun.com)… 128.122.47.89 正在连接 yann.lecun.com
(yann.lecun.com)|128.122.47.89|:80… 已连接。 已发出 HTTP 请求,正在等待回应… 200
OK 长度: 1648877 (1.6M) [application/x-gzip] 正在保存至:
“t10k-images-idx3-ubyte.gz”
100%[======================================>] 1,648,877 452KB/s
用时 3.6s 2015-09-21 09:11:22 (452 KB/s) - 已保存
“t10k-images-idx3-ubyte.gz” [1648877/1648877])
–2015-09-21 09:11:22– http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz 正在解析主机
yann.lecun.com (yann.lecun.com)… 128.122.47.89 正在连接 yann.lecun.com
(yann.lecun.com)|128.122.47.89|:80… 已连接。 已发出 HTTP 请求,正在等待回应… 200
OK 长度: 4542 (4.4K) [application/x-gzip] 正在保存至:
“t10k-labels-idx1-ubyte.gz”
100%[======================================>] 4,542 13.2KB/s
用时 0.3s 2015-09-21 09:11:23 (13.2 KB/s) - 已保存
“t10k-labels-idx1-ubyte.gz” [4542/4542]) Unzipping… Done. Creating
lmdb… Done.
还需要两个文件:
- net prototxt,定义网络结构,并指定对应的训练/测试数据
- solver prototxt,定义学习的参数
先从网络开始,用python代码以caffe的protobuf模型格式写网络。这个网络读取LMDB格式的预生成数据,也可以用MemoryDataLayer从ndarray中直接读取。
from caffe import layers as L
from caffe import params as P
def lenet(lmdb, batch_size):
# our version of LeNet: a series of linear and simple nonlinear transformations
n = caffe.NetSpec()
n.data, n.label = L.Data(batch_size=batch_size, backend=P.Data.LMDB, source=lmdb,
transform_param=dict(scale=1./255), ntop=2)
n.conv1 = L.Convolution(n.data, kernel_size=5, num_output=20, weight_filler=dict(type='xavier'))
n.pool1 = L.Pooling(n.conv1, kernel_size=2, stride=2, pool=P.Pooling.MAX)
n.conv2 = L.Convolution(n.pool1, kernel_size=5, num_output=50, weight_filler=dict(type='xavier'))
n.pool2 = L.Pooling(n.conv2, kernel_size=2, stride=2, pool=P.Pooling.MAX)
n.ip1 = L.InnerProduct(n.pool2, num_output=500, weight_filler=dict(type='xavier'))
n.relu1 = L.ReLU(n.ip1, in_place=True)
n.ip2 = L.InnerProduct(n.relu1, num_output=10, weight_filler=dict(type='xavier'))
n.loss = L.SoftmaxWithLoss(n.ip2, n.label)
return n.to_proto()
with open('examples/mnist/lenet_auto_train.prototxt', 'w') as f:
f.write(str(lenet('examples/mnist/mnist_train_lmdb', 64)))
with open('examples/mnist/lenet_auto_test.prototxt', 'w') as f:
f.write(str(lenet('examples/mnist/mnist_test_lmdb', 100)))
这个网络利用Google的protobuf库写入到磁盘,虽然这样写出来的网络模型有些冗长,但是可读性好,可以直接读、写、修改。看一下这样自动生成的训练网络:
!cat examples/mnist/lenet_auto_train.prototxt
layer { name: "data" type: "Data" top: "data" top: "label" transform_param { scale: 0.00392156862745 }
data_param { source: "examples/mnist/mnist_train_lmdb" batch_size: 64 backend: LMDB }
}
layer { name: "conv1" type: "Convolution" bottom: "data" top: "conv1" convolution_param { num_output: 20 kernel_size: 5 weight_filler { type: "xavier" }
}
}
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" convolution_param { num_output: 50 kernel_size: 5 weight_filler { type: "xavier" }
}
}
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" inner_product_param { num_output: 500 weight_filler { type: "xavier" }
}
}
layer { name: "relu1" type: "ReLU" bottom: "ip1" top: "ip1" }
layer { name: "ip2" type: "InnerProduct" bottom: "ip1" top: "ip2" inner_product_param { num_output: 10 weight_filler { type: "xavier" }
}
}
layer { name: "loss" type: "SoftmaxWithLoss" bottom: "ip2" bottom: "label" top: "loss" }
看看学习参数,同样是prototxt文件,用带动量的SGD(随机梯度下降),权重递减,以及特定的学习率:
!cat examples/mnist/lenet_auto_solver.prototxt
# The train/test net protocol buffer definition
train_net: "examples/mnist/lenet_auto_train.prototxt"
test_net: "examples/mnist/lenet_auto_test.prototxt"
# test_iter specifies how many forward passes the test should carry out.
# In the case of MNIST, we have test batch size 100 and 100 test iterations,
# covering the full 10,000 testing images.
test_iter: 100
# Carry out testing every 500 training iterations.
test_interval: 500
# The base learning rate, momentum and the weight decay of the network.
base_lr: 0.01
momentum: 0.9
weight_decay: 0.0005
# The learning rate policy
lr_policy: "inv"
gamma: 0.0001
power: 0.75
# Display every 100 iterations
display: 100
# The maximum number of iterations
max_iter: 10000
# snapshot intermediate results
snapshot: 5000
snapshot_prefix: "examples/mnist/lenet"
选择GPU,载入solver,这里用的是SGD,Adagrad和Nesterov加速梯度也是可行的:
caffe.set_device(0)
caffe.set_mode_gpu()
solver = caffe.SGDSolver('examples/mnist/lenet_auto_solver.prototxt')
查看中间特征(blobs)和参数(params)的维数,对网络结构加深理解:
# each output is (batch size, feature dim, spatial dim)
[(k, v.data.shape) for k, v in solver.net.blobs.items()]
[(‘data’, (64, 1, 28, 28)),
(‘label’, (64,)),
(‘conv1’, (64, 20, 24, 24)),
(‘pool1’, (64, 20, 12, 12)),
(‘conv2’, (64, 50, 8, 8)),
(‘pool2’, (64, 50, 4, 4)),
(‘ip1’, (64, 500)),
(‘ip2’, (64, 10)),
(‘loss’, ())]
# just print the weight sizes (not biases)
[(k, v[0].data.shape) for k, v in solver.net.params.items()]
[(‘conv1’, (20, 1, 5, 5)),
(‘conv2’, (50, 20, 5, 5)),
(‘ip1’, (500, 800)),
(‘ip2’, (10, 500))]
在开始训练前,检查一下所有东西是不是都载入了,接下来在测试集和训练集上执行一个前向的过程,确保网络中包含了数据:
solver.net.forward() # train net
solver.test_nets[0].forward() # test net (there can be more than one)
{‘loss’: array(2.394181489944458, dtype=float32)}
显示训练集的前8个图像和它们的标签来看看:
# we use a little trick to tile the first eight images
imshow(solver.net.blobs['data'].data[:8, 0].transpose(1, 0, 2).reshape(28, 828), cmap='gray')
print solver.net.blobs['label'].data[:8]
[ 5. 0. 4. 1. 9. 2. 1. 3.]
再显示测试集的前8个图像和它们的标签来看看:
imshow(solver.test_nets[0].blobs['data'].data[:8, 0].transpose(1, 0, 2).reshape(28, 828), cmap='gray')
print solver.test_nets[0].blobs['label'].data[:8]
[ 7. 2. 1. 0. 4. 1. 4. 9.]
看起来训练集和测试集的数据都已经顺利载入了,并且标签都是正确的。下面执行一个minibatch的SGD,看看会发生什么:
solver.step(1)
我们看一下第一层的滤波器经过传播后的更新情况,20个5×5的滤波器:
imshow(solver.net.params['conv1'][0].diff[:, 0].reshape(4, 5, 5, 5).transpose(0, 2, 1, 3).reshape(45, 55), cmap='gray')
让网络运行一会,注意到这个过程和通过caffe的binary训练是一样的。由于可以控制python中的循环,因此可以做一些其他的事情,例如自定义停止的标准,通过循环更新网络来改变求解过程:
%%time
niter = 200
test_interval = 25
# losses will also be stored in the log
train_loss = zeros(niter)
test_acc = zeros(int(np.ceil(niter / test_interval)))
output = zeros((niter, 8, 10))
# the main solver loop
for it in range(niter):
solver.step(1) # SGD by Caffe
# store the train loss
train_loss[it] = solver.net.blobs['loss'].data
# store the output on the first test batch
# (start the forward pass at conv1 to avoid loading new data)
solver.test_nets[0].forward(start='conv1')
output[it] = solver.test_nets[0].blobs['ip2'].data[:8]
# run a full test every so often
# (Caffe can also do this for us and write to a log, but we show here
# how to do it directly in Python, where more complicated things are easier.)
if it % test_interval == 0:
print 'Iteration', it, 'testing...'
correct = 0
for test_it in range(100):
solver.test_nets[0].forward()
correct += sum(solver.test_nets[0].blobs['ip2'].data.argmax(1)
== solver.test_nets[0].blobs['label'].data)
test_acc[it // test_interval] = correct / 1e4
Iteration 0 testing…
Iteration 25 testing…
Iteration 50 testing…
Iteration 75 testing…
Iteration 100 testing…
Iteration 125 testing…
Iteration 150 testing…
Iteration 175 testing…
CPU times: user 2.29 s, sys: 588 ms, total: 2.88 s
Wall time: 2.05 s
接下来画出训练损失和测试准确率:
_, ax1 = subplots()
ax2 = ax1.twinx()
ax1.plot(arange(niter), train_loss)
ax2.plot(test_interval arange(len(test_acc)), test_acc, 'r')
ax1.set_xlabel('iteration')
ax1.set_ylabel('train loss')
ax2.set_ylabel('test accuracy')
损失下降很快,并且会收敛(有些随机性),准确率相应上升。由于保存了第一个测试batch的结果,可以看一下预测得分是怎样变化的,x轴为时间,y轴为每个可能的标签,亮度表示置信度:
for i in range(8):
figure(figsize=(2, 2))
imshow(solver.test_nets[0].blobs['data'].data[i, 0], cmap='gray')
figure(figsize=(10, 2))
imshow(output[:50, i].T, interpolation='nearest', cmap='gray')
xlabel('iteration')
ylabel('label')
发现最后一个数字9最容易出错,会和数字4混淆。注意到这是原始输出得分,而不是softmax计算后的概率,下面的内容可以看到网络的置信度:
for i in range(8):
figure(figsize=(2, 2))
imshow(solver.test_nets[0].blobs['data'].data[i, 0], cmap='gray')
figure(figsize=(10, 2))
imshow(exp(output[:50, i].T) / exp(output[:50, i].T).sum(0), interpolation='nearest', cmap='gray')
xlabel('iteration')
ylabel('label')