diff options
| author | txh18 <cloudygooseg@gmail.com> | 2015-11-17 23:31:14 +0800 |
|---|---|---|
| committer | txh18 <cloudygooseg@gmail.com> | 2015-11-17 23:31:14 +0800 |
| commit | 54226ba490b65c223a230dc5de5cea56b1543594 (patch) | |
| tree | 76709dfd982d1a89067d8464d8ec638d48b6bdf9 | |
| parent | fc4c5a71053b837ed6143659a6e7b45792ed9e51 (diff) | |
bug fix for select_linear layer-by-layer update
| -rw-r--r-- | nerv/examples/lmptb/lmptb/layer/select_linear.lua | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/nerv/examples/lmptb/lmptb/layer/select_linear.lua b/nerv/examples/lmptb/lmptb/layer/select_linear.lua index a0ae207..b215d48 100644 --- a/nerv/examples/lmptb/lmptb/layer/select_linear.lua +++ b/nerv/examples/lmptb/lmptb/layer/select_linear.lua @@ -30,8 +30,8 @@ function SL:init(batch_size) end function SL:update(bp_err, input, output) - for i = 1, input[1]:ncol(), 1 do - local word_vec = self.ltp.trans[input[1][0][i - 1]] + for i = 1, input[1]:nrow(), 1 do + local word_vec = self.ltp.trans[input[1][i - 1][0]] word_vec:add(word_vec, bp_err[1][i - 1], 1, - self.gconf.lrate / self.gconf.batch_size) end --I tried the update_select_rows kernel which uses atomicAdd, but it generates unreproducable result |