Fix bad name used.

theano/tensor/nnet/nnet.py

@@ -1138,7 +1138,7 @@ def crossentropy_to_crossentropy_with_softmax_with_bias(fgraph):
                     new_nll, new_sm, new_am = crossentropy_softmax_argmax_1hot_with_bias(x, b,
                             one_of_n)
                     fgraph.replace_all_validate([(nll, new_nll), (sm, new_sm)],
-                            reason="crossentropy_to_crossentropy_with_softmax")
+                reason="crossentropy_to_crossentropy_with_softmax_with_bias")
                     return True
 
         return False