Use outputs list to set CTC Op's outputs in make_node

theano/tensor/nnet/ctc.py

@@ -109,15 +109,16 @@ class ConnectionistTemporalClassification(gof.COp, gof.OpenMPOp):
             raise TypeError('Label lengths must use the int32 type!')
 
         costs = T.fvector(name="ctc_cost")
-
+        outputs = [costs]
         if self.compute_grad:
             gradients = T.ftensor3(name="ctc_grad")
+            outputs += [gradients]
 
         # Return only the cost. Gradient will be returned by grad()
         self.default_output = 0
 
         return gof.Apply(self, inputs=[t_activations, t_labels, t_input_lengths],
-                         outputs=[costs, gradients])
+                         outputs=outputs)
 
     def L_op(self, inputs, outputs, output_grads):
         if not ctc_enabled: