I have made scan always clone constants to remove and reference of an old

theano/scan_module/scan_utils.py

@@ -696,6 +696,11 @@ def reconstruct_graph(inputs, outputs, tag = None):
     givens = {}
     for nw_x, x in zip(nw_inputs, inputs):
         givens[x] = nw_x
+    allinputs = theano.gof.graph.inputs(outputs)
+    for inp in allinputs:
+        if isinstance(inp, theano.Constant):
+            givens[inp] = inp.clone()
+
     nw_outputs = clone( outputs, replace=givens)
     return (nw_inputs, nw_outputs)
 

theano/scan_module/tests/test_scan.py

 import time
 import unittest
 
+import cPickle
 import numpy
 
 import theano
@@ -187,6 +188,33 @@ class T_Scan(unittest.TestCase):
     def setUp(self):
         utt.seed_rng()
 
+    # generator network, only one output , type scalar ; no sequence or
+    # non sequence arguments
+    def test_pickling(self):
+        def f_pow2(x_tm1):
+            return 2*x_tm1
+
+        state = theano.tensor.scalar('state')
+        n_steps = theano.tensor.iscalar('nsteps')
+        output, updates = theano.scan(f_pow2, [],state, [],n_steps = n_steps, truncate_gradient
+                = -1, go_backwards = False)
+        _my_f = theano.function([state,n_steps], output, updates = updates,
+                               allow_input_downcast = True)
+
+        ### TESTING PICKLE-ing this function
+        cPickle.dump(_my_f, open('tmp_scan_test_pickle.pkl','wb'),-1)
+        my_f = cPickle.load(open('tmp_scan_test_pickle.pkl'))
+
+        rng = numpy.random.RandomState(utt.fetch_seed())
+        state = rng.uniform()
+        steps = 5
+
+        numpy_values = numpy.array([ state*(2**(k+1)) for k
+                                    in xrange(steps) ])
+        theano_values = my_f(state,steps)
+        assert numpy.allclose(numpy_values,theano_values)
+
+
     # generator network, only one output , type scalar ; no sequence or
     # non sequence arguments
     def test_generator_one_output_scalar(self):