1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
|
#!/usr/bin/env ruby
require 'sinatra'
require 'sinatra/cross_origin'
require 'nanomsg'
require 'zipf'
require 'digest'
require_relative "#{ARGV[0]}"
INPUT = ReadFile.readlines INPUT_FILE
`mkdir -p #{WORK_DIR}/g`
def start_daemon cmd, name, addr
STDERR.write "> starting #{name} daemon\n"
cmd.gsub! '__ADDR__', addr
pid = fork do
exec cmd
end
sock = NanoMsg::PairSocket.new
sock.connect addr
STDERR.write "< got #{sock.recv} from #{name}\n"
return sock, pid
end
def stop_all_daemons env
STDERR.write "shutting down all daemons\n"
env.each { |name,p|
p[:socket].send "shutdown"
STDERR.write "< #{name} is #{p[:socket].recv}\n"
}
end
daemons = {
:extractor => "python -m cdec.sa.extract -c #{DATA_DIR}/sa.ini --online -u -S '__ADDR__'",
:aligner_fwd => "#{CDEC_NET}/word-aligner/net_fa -f #{DATA_DIR}/a/forward.params -m #{FWD_MEAN_SRCLEN_MULT} -T #{FWD_TENSION} --sock_url '__ADDR__'",
:aligner_back => "#{CDEC_NET}/word-aligner/net_fa -f #{DATA_DIR}/a/backward.params -m #{BACK_MEAN_SRCLEN_MULT} -T #{BACK_TENSION} --sock_url '__ADDR__'",
:atools => "#{CDEC_NET}/utils/atools_net -c grow-diag-final-and -S '__ADDR__'",
:dtrain => "#{CDEC_NET}/training/dtrain/dtrain_net_interface -c #{DATA_DIR}/dtrain.ini -o #{WORK_DIR}/weights.final -a '__ADDR__'" ##{DTRAIN_EXTRA}"
}
env = {}
port = BEGIN_PORT_RANGE
daemons.each { |name,cmd|
sock, pid = start_daemon cmd, name, "tcp://127.0.0.1:#{port}"
env[name] = { :socket => sock, :pid => pid }
port += 1
}
set :bind, SERVER_IP
set :port, WEB_PORT
set :allow_origin, :any
set :allow_methods, [:get, :post, :options]
set :allow_credentials, true
set :max_age, "1728000"
set :expose_headers, ['Content-Type']
get '/' do
cross_origin
"Nothing to see here."
end
get '/next' do
cross_origin
if params[:example]
source, reference = params[:example].strip.split(" ||| ")
# update weights
grammar = "#{WORK_DIR}/g/#{Digest::SHA256.hexdigest(source)}.grammar"
annotated_source = "<seg grammar=\"#{grammar}\"> #{source} </seg>"
msg = "#{annotated_source} ||| #{reference}"
STDERR.write "[dtrain] > sending '#{msg}' for update\n"
env[:dtrain][:socket].send msg
STDERR.write "[dtrain] waiting for confirmation ...\n"
STDERR.write "[dtrain] < says it's #{env[:dtrain][:socket].recv}\n"
# update grammar extractor
# get forward alignment
msg = "#{source} ||| #{reference}"
STDERR.write "[aligner_fwd] > sending '#{msg}' for forced alignment\n"
env[:aligner_fwd][:socket].send msg
STDERR.write "[aligner_fwd] waiting for alignment ...\n"
a_fwd = env[:aligner_fwd][:socket].recv.strip
STDERR.write "[aligner_fwd] < got alignment: '#{a_fwd}'\n"
# get backward alignment
msg = "#{source} ||| #{reference}"
STDERR.write "[aligner_back] > sending '#{msg}' for forced alignment\n"
env[:aligner_back][:socket].send msg
STDERR.write "[aligner_back] waiting for alignment ...\n"
a_back = env[:aligner_back][:socket].recv.strip
STDERR.write "[aligner_back] < got alignment: '#{a_back}'\n"
# combine alignments
msg = "#{a_fwd} ||| #{a_back}"
STDERR.write "[atools] > sending '#{msg}' to combine alignments\n"
env[:atools][:socket].send msg
STDERR.write "[atools] waiting for alignment ...\n"
a = env[:atools][:socket].recv.strip
STDERR.write "[atools] < got alignment '#{a}'\n"
# actual extractor
msg = "TEST ||| #{source} ||| #{reference} ||| #{a}"
STDERR.write "[extractor] > sending '#{msg}' for learning\n"
env[:extractor][:socket].send "TEST ||| #{source} ||| #{reference} ||| #{a}"
STDERR.write "[extractor] waiting for confirmation ...\n"
STDERR.write "[extractor] < got '#{env[:extractor][:socket].recv}'\n"
end
source = INPUT.shift
if !source # input is done -> displays 'Thank you!'
STDERR.write ">>> end of input, sending 'fi'\n"
"fi"
else # translate next sentence
source.strip!
# generate grammar for current sentence
grammar = "#{WORK_DIR}/g/#{Digest::SHA256.hexdigest(source)}.grammar" # FIXME: keep grammars?
msg = "- ||| #{source} ||| #{grammar}" # FIXME: content identifier useful?
STDERR.write "[extractor] > asking to generate grammar: '#{msg}'\n"
env[:extractor][:socket].send msg
STDERR.write "[extractor] waiting for confirmation ...\n"
STDERR.write "[extractor] < says it generated #{env[:extractor][:socket].recv.strip}\n"
# translation
msg = "act:translate ||| <seg grammar=\"#{grammar}\"> #{source} </seg>"
STDERR.write "[dtrain] > asking to translate: '#{msg}'\n"
env[:dtrain][:socket].send msg
STDERR.write "[dtrain] waiting for translation ...\n"
transl = env[:dtrain][:socket].recv.force_encoding "UTF-8"
STDERR.write "[dtrain] < received translation: '#{transl}'\n"
"#{source}\t#{transl}"
end
end
# stop daemons and shut down server
get '/shutdown' do
stop_all_daemons env
exit
end
|
