TorchNN/Training.lua at master · Legomaniak/TorchNN · GitHub

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
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
require 'FillData'
require 'ReadDataset'
require 'TensorSaveLoad'
require 'TestModel'


function TrainingFromFile(model, listIn, listOut, selector, trainer)
  local etime = sys.clock();
  -- TRAINING
  model:training()

  local nFile = table.getn(listIn)
  for noFile = 1,nFile do
    if Settings.DispFile then plog.info("Training from File: " .. noFile .. "/" .. nFile) end
    local dataIn = NactiData(Settings.ListFolder .. listIn[i]);
    local dataOut = NactiData(Settings.ListFolder .. listOut[i]);
    selector(model, dataIn, dataOut, trainer);
  end
  log.info("Training " .. Settings.ModelName .. " completed in " .. sys.clock() - etime)
end

function Training(model, dataIn, dataOut, selector, trainer)
  local etime = sys.clock();
  -- TRAINING
  model:training()

--  local nFile = dataIn:size(1)
  local nFile = table.getn(dataIn)
  for noFile = 1,nFile do
    if Settings.DispFile then plog.info("Training from File: " .. noFile .. "/" .. nFile) end
--    selector(model, dataIn:select(1,noFile), dataOut:select(1,noFile), trainer);
    selector(model, dataIn[noFile], dataOut[noFile], trainer);
  end
  log.info("Training " .. Settings.ModelName .. " completed in " .. sys.clock() - etime)
end

function TrainingRaw(model, DataIn, DataOut)
  local etime = sys.clock();
  -- TRAINING
  model:training()
--  local nFile = DataIn:size(1)
  local nFile = table.getn(DataIn)
  for noFile = 1,nFile do
    if Settings.DispFile then plog.info("Training from File: " .. noFile .. "/" .. nFile) end
     -- prepare inputs & outputs tensors
--      local dataIn=DataIn:select(1,noFile)
--      local dataOut=DataOut:select(1,noFile)
      local dataIn=DataIn[noFile]
      local dataOut=DataOut[noFile]
      for batch = 1,Settings.BatchN do
        if Settings.DispBatch then plog.info("Training Batch: " .. batch .. "/" .. Settings.BatchN) end
        local tIn ={}
        table.insert(tIn,Settings.BatchSize)
        local tOut ={}
        table.insert(tOut,Settings.BatchSize)
        for i = 2,dataIn:dim() do
          table.insert(tIn,dataIn:size(i))
        end
        for i = 2,dataOut:dim() do
          table.insert(tOut,dataOut:size(i))
        end

        local inputs = torch.Tensor(torch.LongStorage(tIn))
        local outputs = torch.Tensor(torch.LongStorage(tOut))

        -- process batches
        for i = 1, Settings.BatchSize do
          local dataM=torch.random(1,dataIn:size(1))
          inputs[i] = dataIn:select(1,dataM)
          outputs[i] = dataOut:select(1,dataM)
        end
        TrainOptim(model, inputs, outputs);
      end
    end
  log.info("Training " .. Settings.ModelName .. " completed in " .. sys.clock() - etime)
end

function TestModelFromFile(model, listIn, listOut, selector, tester)
  local etime = sys.clock();
  -- EVALUATION
  model:evaluate();
  local err = 0
  local all = 0

  local nFile = table.getn(listIn)
  for noFile = 1,nFile do
    if Settings.DispFile then plog.info("Testing from File: " .. noFile .. "/" .. nFile) end
    local dataIn = NactiData(Settings.ListFolder .. listIn[nFile]);
    local dataOut = NactiData(Settings.ListFolder .. listOut[nFile]);
    local f = function(model, dataIn, dataOut)
      local e,a = tester(model, dataIn, dataOut);
      err=err+e
      all=all+a
    end
    selector(model, dataIn, dataOut, f);
  end
  -- logs & export model
  err = 100*err/all
  log.info("Testing " .. Settings.ModelName .. " completed in " .. sys.clock() - etime)
  return err
end

function TestModel(model, dataIn, dataOut, selector, tester)
  local etime = sys.clock();
  -- EVALUATION
  model:evaluate();
  local err = 0
  local all = 0
  local nFile = dataIn:size(1)
  for noFile = 1,nFile do
    if Settings.DispFile then plog.info("Testing from File: " .. noFile .. "/" .. nFile) end
    local f = function(model, dataIn, dataOut)
      local a,e = tester(model, dataIn, dataOut);
      err=err+e
      all=all+a
    end
    selector(model, dataIn:select(1,noFile), dataOut:select(1,noFile), f);
  end
  -- logs & export model
  err = 100*err/all
  log.info("Testing " .. Settings.ModelName .. " completed in " .. sys.clock() - etime)
  return err
end

function TestModelX(model, dataIn, dataOut, selector, tester)
  local etime = sys.clock();
  -- EVALUATION
  model:evaluate();
  local err = torch.Tensor(Settings.OutputSize,Settings.OutputSize):fill(0)
  local all = torch.Tensor(Settings.OutputSize):fill(0)
  local errn = torch.Tensor(Settings.OutputSize):fill(0)
--  local nFile = dataIn:size(1)
  local nFile = table.getn(dataIn)
  for noFile = 1,nFile do
    if Settings.DispFile then plog.info("Testing from File: " .. noFile .. "/" .. nFile) end
    local f = function(model, dataIn, dataOut)
      local a,e = tester(model, dataIn, dataOut);
      err=err+e
      all=all+a
    end
    selector(model, dataIn[noFile], dataOut[noFile], f);
  end
  -- logs & export model
  for i=1,err:size(1) do
    if all[i]==0 then
      err[i]=0
    else
      --err[i]=100*err[i]:sum()/all[i]
      err[i]=100*err[i]/all[i]
    end
  end
  --err = 100*err/all
  log.info("Testing " .. Settings.ModelName .. " completed in " .. sys.clock() - etime)
  return err
end