lstm
lstm copied to clipboard
wojzaremba
I have a time series history data but the the length is only 28 readings taken over 6 months, can I use RNN(LSTM) to predict next 3 months readings? if...
I want to save the trainned model after every epoch, I save the `local model` defined in this code and also extra data .i.e step....every time reload from the previous...
How to run on a specific treebank text corpus, like [Italian treebank corpus](http://universaldependencies.org/it/overview/introduction.html)? Thank you.
Hi, I can not understand why in the following code, the `reshaped_gates` is split along dimension 2 rather than 1, since output of `reshaped_gates` is a `4 x rnn_size` matrix....
Is there any specific reason for using torch.MemoryFile(''w'') in the g_cloneManyTimes() function rather than simply using net:clone()? Thanks
Hello, Thanks for your code. but when I tried your code, it always shows the problem that : /home/shi/torch/install/share/lua/5.1/nn/LogSoftMax.lua:4: attempt to call field 'LogSoftMax_updateOutput' (a nil value) stack traceback: /home/shi/torch/install/share/lua/5.1/nn/LogSoftMax.lua:4:...
Hi, I found `cutorch.synchronize()` was used in [main.lua#L184](https://github.com/wojzaremba/lstm/blob/master/main.lua#L184). Isn't `backward()` blocking? And in my experiment, `cutorch.synchronize()` will cost much time.
I'm having trouble understanding `replicate`. ``` LUA -- Stacks replicated, shifted versions of x_inp -- into a single matrix of size x_inp:size(1) x batch_size. local function replicate(x_inp, batch_size) local s...
https://github.com/wojzaremba/lstm/blob/master/main.lua#L212 Shouldn't this be 1 to len?
