take care of case when n_steps is None

theano/sandbox/scan_module/scan_utils.py

@@ -187,13 +187,16 @@ def canonical_arguments(sequences,
     parameters = [tensor.as_tensor_variable(x) for x in to_list(non_sequences)]
 
     inputs = []
+    if n_steps is not None:
+        negative_n_steps = tensor.lt(tensor.as_tensor_variable(n_steps), 0)
+
     for input in to_list(sequences):
         if not isinstance(u, dict):
             nw_input = tensor.as_tensor_variable(input)
             if go_backwards:
                 nw_input = nw_input[::-1]
             if n_steps is not None:
-                nw_input = tensor.switch(n_steps < 0, nw_input[::-1],
+                nw_input = tensor.switch(negative_n_steps, nw_input[::-1],
                                          nw_input)
             inputs.append(tensor.as_tensor_variable(nw_input))
         elif input.get('taps', True) is None:
@@ -201,7 +204,7 @@ def canonical_arguments(sequences,
             if go_backwards:
                 nw_input = nw_input[::-1]
             if n_steps is not None:
-                nw_input = tensor.switch(n_steps < 0, nw_input[::-1],
+                nw_input = tensor.switch(negative_n_steps, nw_input[::-1],
                                          nw_input)
             inputs.append(nw_input)
         elif input.get('taps', None):
@@ -211,7 +214,7 @@ def canonical_arguments(sequences,
             if go_backwards:
                 orig_input = orig_input[::-1]
             if n_steps is not None:
-                orig_input = tensor.switch(n_steps < 0, orig_input[::-1],
+                orig_input = tensor.switch(negative_n_steps, orig_input[::-1],
                                          org_input)
             for k in input['taps']:
                 # We cut the sequence such that seq[i] to correspond to