Merge pull request lisa-lab#17 from boulanni/master

nouiz · nouiz · commit b23ef621db10 · 2013-02-21T06:58:06.000-08:00
Small fixes to RNN-RBM tutorial
diff --git a/code/rnnrbm.py b/code/rnnrbm.py
@@ -271,9 +271,10 @@ def test_rnnrbm(batch_size=100, num_epochs=200):
     model = RnnRbm()
     model.train(glob.glob('../data/Nottingham/train/*.mid'),
                 batch_size=batch_size, num_epochs=num_epochs)
+    return model
 
 if __name__ == '__main__':
-    test_rnnrbm()
+    model = test_rnnrbm()
     model.generate('sample1.mid')
     model.generate('sample2.mid')
     pylab.show()
diff --git a/doc/rnnrbm.txt b/doc/rnnrbm.txt
@@ -11,7 +11,7 @@ Modeling and generating sequences of polyphonic music with the RNN-RBM
   and `restricted Boltzmann machines (RBM) <rbm.html>`_.
 
 .. note::
-  The code for this section is available for download here: `rnnrbm.py <../code/rnnrbm.py>`_.
+  The code for this section is available for download here: `rnnrbm.py <code/rnnrbm.py>`_.
 
   You will need the modified `Python MIDI package (GPL license) <http://www.iro.umontreal.ca/~lisa/deep/midi.zip>`_ in your ``$PYTHONPATH`` or in the working directory in order to convert MIDI files to and from piano-rolls.
   The script also assumes that the content of the `Nottingham Database of folk tunes <http://www.iro.umontreal.ca/~lisa/deep/data/Nottingham.zip>`_ has been extracted in the ``../data`` directory.
@@ -266,7 +266,7 @@ We now have all the necessary ingredients to start training our network on real
                                              n_hidden_recurrent)
 
           gradient = T.grad(cost, params, consider_constant=[v_sample])
-          updates_train.update(dict((p, p - lr * g) for p, g in zip(params,
+          updates_train.update(((p, p - lr * g) for p, g in zip(params,
                                                                   gradient)))
           self.train_function = theano.function([v], monitor,
                                                  updates=updates_train)
@@ -288,7 +288,10 @@ We now have all the necessary ingredients to start training our network on real
 
           assert len(files) > 0, 'Training set is empty!' \
                                  ' (did you download the data files?)'
-          dataset = [midiread(f, self.r, self.dt).piano_roll for f in files]
+          dataset = [midiread(f, self.r,
+                              self.dt).piano_roll.astype(theano.config.floatX)
+                     for f in files]
+          
           try:
               for epoch in xrange(num_epochs):
                   numpy.random.shuffle(dataset)
