hi~I ran the demo train-a-digit-classifier, and found when I set the batchsize more bigger,like 300, the procedure occured an error of out of range.
Actually, I also noticed the torch/optim project is updated for some simlar errors, but after I updated the optim page using "luarocks install optim", it still didn't work.
So, how can I fix this problem for big batchsize.By the way, I am kind of confusion about the ConfusionMatrix;what's the effect of the matrix?
karandwivedi42
commented
8 years ago
hi~I ran the demo train-a-digit-classifier, and found when I set the batchsize more bigger,like 300, the procedure occured an error of out of range. Actually, I also noticed the torch/optim project is updated for some simlar errors, but after I updated the optim page using "luarocks install optim", it still didn't work. So, how can I fix this problem for big batchsize.By the way, I am kind of confusion about the ConfusionMatrix;what's the effect of the matrix?