Fix: Issue in rnn/Sequencer.lua #7
Description
I got an error like this:
imi@imi-All-Series:~/graulef/deep-soli$ th net/main.lua --file ../datapre --list config/file_half.json --load ../uni_image_np_50.t7 --inputsize 32 --inputch 4 --label 11 --datasize 32 --datach 4 --batch 4 --maxseq 40 --cuda --cudnn
[eval] data with 1364 seq
[net] loading model ../uni_image_np_50.t7
nn.Sequencer @ nn.Recursor @ nn.MaskZero @ nn.Sequential {
[input -> (1) -> (2) -> (3) -> (4) -> (5) -> (6) -> (7) -> (8) -> (9) -> (10) -> (11) -> (12) -> (13) -> (14) -> (15) -> (16) -> (17) -> (18) -> (19) -> (20) -> (21) -> output]
(1): cudnn.SpatialConvolution(4 -> 32, 3x3, 2,2)
(2): nn.SpatialBatchNormalization (4D) (32)
(3): cudnn.ReLU
(4): cudnn.SpatialConvolution(32 -> 64, 3x3, 2,2)
(5): nn.SpatialBatchNormalization (4D) (64)
(6): cudnn.ReLU
(7): nn.SpatialDropout(0.400000)
(8): cudnn.SpatialConvolution(64 -> 128, 3x3, 2,2)
(9): nn.SpatialBatchNormalization (4D) (128)
(10): cudnn.ReLU
(11): nn.SpatialDropout(0.400000)
(12): nn.Reshape(1152)
(13): nn.Linear(1152 -> 512)
(14): nn.BatchNormalization (2D) (512)
(15): cudnn.ReLU
(16): nn.Dropout(0.5, busy)
(17): nn.Linear(512 -> 512)
(18): nn.LSTM(512 -> 512)
(19): nn.Dropout(0.5, busy)
(20): nn.Linear(512 -> 13)
(21): cudnn.LogSoftMax
}
/home/imi/graulef/datapre/0_12_20/label.json
/home/imi/graulef/datapre/0_10_18/label.json
/home/imi/graulef/datapre/0_6_8/label.json
/home/imi/graulef/datapre/0_3_0/label.json
Evaluation passed
/home/imi/torch/install/bin/luajit: /home/imi/torch/install/share/lua/5.1/rnn/Sequencer.lua:93: attempt to index field 'tableoutput' (a nil value)
stack traceback:
/home/imi/torch/install/share/lua/5.1/rnn/Sequencer.lua:93: in function 'forward'
./net/rnntrain.lua:31: in function 'batchEval'
./net/train.lua:25: in function 'epochEval'
./net/train.lua:47: in function 'train'
net/main.lua:45: in main chunk
[C]: in function 'dofile'
.../imi/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk
[C]: at 0x004065d0
Fixed it by adding self.tableoutput = {} to line 73 in rnn/Sequencer.lua, so basically just reinitializing the table as that was somehow forgotten in the original version.