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
|
local PushTable, parent = torch.class("nn.PushTable", "nn.Module")
function PushTable:__init(index)
self._index = index
self._pulls = {}
self.output = {}
self._gradInput = torch.Tensor()
self.gradInput = {}
self._forward = false
end
function PushTable:pull(index)
local pull = nn.PullTable(self, index)
table.insert(self._pulls, pull)
return pull
end
function PushTable:updateOutput(inputTable)
for i, input in ipairs(inputTable) do
if i < self._index then
self.output[i] = input
elseif i > self._index then
self.output[i-1] = input
end
end
local input = inputTable[self._index]
for i,pull in ipairs(self._pulls) do
pull:_updateOutput(input)
end
self._forward = true
return self.output
end
function PushTable:_updateGradInput(gradOutput)
if self._forward then
if torch.type(self.gradInput) ~= torch.type(gradOutput) then
self._gradInput = gradOutput.new()
end
self._gradInput:resizeAs(gradOutput)
self._gradInput:copy(gradOutput)
else
self._gradInput:add(gradOutput)
end
self._forward = false
end
function PushTable:updateGradInput(inputTable, gradOutputTable)
for i, gradOutput in ipairs(gradOutputTable) do
if i < self._index then
self.gradInput[i] = gradOutput
elseif i > self._index then
self.gradInput[i+1] = gradOutput
end
end
self.gradInput[self._index] = self._gradInput
assert(#inputTable == #self.gradInput, "tables size mismatch")
return self.gradInput
end
function PushTable:type(type, tensorCache)
assert(type, 'PullTable: must provide a type to convert to')
tensorCache = tensorCache or {}
-- find all tensors and convert them
for key,param in pairs(self) do
if(key ~= "_pulls") then
self[key] = nn.utils.recursiveType(param, type, tensorCache)
end
end
return self
end
|
