docker下使用caffe的命令记录

• 查看所有的images

sudo docker images

• 利用某个image生成container

  sudo docker run -it --net=host -v /home/tingting/Dropbox/code/craterDetection/docker:/root/workspace tingting/caffe:version1 bash

• 打开jupyter notebook

jupyter notebook

在浏览器中输入：localhost:8890

这里一定要注意，有时因为打开多个containers，新的container的jupyter notebook的接口是不同的，之前因为以为所有的接口窦唯8888，所以直接输入了8888，但其实这样打开的是8888那个接口对应的container

• 将prototxt文件生成网络结构图

# Load the module
import caffe
import caffe.draw
from caffe.proto import caffe_pb2
from google.protobuf import text_format

# Set the parameters
netName = 'test_net'

input_net_proto_file = netName + '.prototxt'
output_image_file =  netName + '.jpg'
rankdir ='LR'

'''
# net configuration file
input_net_proto_file = 'test_net.prototxt' 
# output configuration image file name
output_image_file ='test_net.jpg' 
# net arrange method， which is in ['LR'、'TB'、'RL']
rankdir ='LR' 
'''

# Read the net
net = caffe_pb2.NetParameter()
text_format.Merge(open(input_net_proto_file).read(), net)

# Draw the net
print('Drawing net to %s' % output_image_file)
caffe.draw.draw_net_to_file(net, output_image_file, rankdir)
print('done...') 

 

•  利用一个新的terminal打开正在运行的container

# watch the containers which are running

sudo dockers ps

# go into one conainer

sudo docker exec -it sharp_ardinghelli bash