npow / rnn-em

Recurrent Neural Networks with External Memory

Recurrent Neural Networks with External Memory

This is a Theano implementation of the RNN-EM model as described in this paper.

This repository uses code from mesnilgr/is13 to load the ATIS dataset.

Usage

git clone --recursive [email protected]:npow/RNN-EM.git
python main.py

rnn-em's Issues

Parameters set

hi,What parameters can get the best final result?

report a bug!

we must do the change as:

diff --git a/is13 b/is13
index 04bc81d..8284fa7 160000
--- a/is13
+++ b/is13
@@ -1 +1 @@
-Subproject commit 04bc81d3a783eea1cccc9c7e4bc0268c22d3690b
+Subproject commit 8284fa7c182720cf47b0b406b6b2ff764c84563a
diff --git a/rnn_em.py b/rnn_em.py
index c0b02b8..311ed1b 100644
--- a/rnn_em.py
+++ b/rnn_em.py
@@ -121,7 +121,8 @@ class model(object):

     self.train = theano.function( inputs  = [idxs, y, lr],
                                   outputs = nll,

•                                  updates = updates )
  

•                                  updates = updates,
  

•                                  on_unused_input='warn')
  
   self.normalize = theano.function( inputs = [],
                    updates = {self.emb:\
  

  envy@ub1404envy:/media/envy/data1t/github/RNN-EM$

envy@ub1404envy:/media/envy/data1t/github/RNN-EM$ sudo python main.py
[sudo] password for envy:

---

Namespace(bs=9, decay=0, emb_size=100, fold=4, hidden_size=100, lr=0.0627142536696559, memory_size=40, n_epochs=50, n_memory_slots=1, seed=345, verbose=1, win=7)
Traceback (most recent call last):
File "main.py", line 63, in
n_memory_slots = s.n_memory_slots )
File "/media/envy/data1t/github/RNN-EM/rnn_em.py", line 124, in init
updates = updates )
File "/usr/local/lib/python2.7/dist-packages/theano/compile/function.py", line 317, in function
output_keys=output_keys)
File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py", line 526, in pfunc
output_keys=output_keys)
File "/usr/local/lib/python2.7/dist-packages/theano/compile/function_module.py", line 1777, in orig_function
output_keys=output_keys).create(
File "/usr/local/lib/python2.7/dist-packages/theano/compile/function_module.py", line 1416, in init
self._check_unused_inputs(inputs, outputs, on_unused_input)
File "/usr/local/lib/python2.7/dist-packages/theano/compile/function_module.py", line 1554, in _check_unused_inputs
i.variable, err_msg))
theano.compile.function_module.UnusedInputError: theano.function was asked to create a function computing outputs given certain inputs, but the provided input variable at index 2 is not part of the computational graph needed to compute the outputs: lr.
To make this error into a warning, you can pass the parameter on_unused_input='warn' to theano.function. To disable it completely, use on_unused_input='ignore'.
envy@ub1404envy:/media/envy/data1t/github/RNN-EM$