-
Notifications
You must be signed in to change notification settings - Fork 313
/
AbstractSequencer.lua
40 lines (34 loc) · 1.56 KB
/
AbstractSequencer.lua
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
local AbstractSequencer, parent = torch.class("nn.AbstractSequencer", "nn.Container")
function AbstractSequencer:getStepModule(step)
error"DEPRECATED 27 Oct 2015. Wrap your internal modules into a Recursor instead"
end
function AbstractSequencer:sharedClone(shareParams, shareGradParams, clones, pointers, stepClone)
-- stepClone is ignored (always false, i.e. uses sharedClone)
return parent.sharedClone(self, shareParams, shareGradParams, clones, pointers)
end
-- AbstractSequence handles its own rho internally (dynamically)
function AbstractSequencer:maxBPTTstep(rho)
end
-- Toggle to feed long sequences using multiple forwards.
-- 'eval' only affects evaluation (recommended for RNNs)
-- 'train' only affects training
-- 'neither' affects neither training nor evaluation
-- 'both' affects both training and evaluation (recommended for LSTMs)
-- Essentially, forget() isn't called on rnn module when remember is on
function AbstractSequencer:remember(remember)
self._remember = (remember == nil) and 'both' or remember
local _ = require 'moses'
assert(_.contains({'both','eval','train','neither'}, self._remember),
"AbstractSequencer : unrecognized value for remember : "..self._remember)
return self
end
function AbstractSequencer:hasMemory()
local _ = require 'moses'
if (self.train ~= false) and _.contains({'both','train'}, self._remember) then -- train (defaults to nil...)
return true
elseif (self.train == false) and _.contains({'both','eval'}, self._remember) then -- evaluate
return true
else
return false
end
end