Fix bug in get_scalar_constant value of Subtensor(Shape(Rebroadcast(v)))

The inner loop that would update the broadcastable pattern according to the rebroadcast dict would overwrite the idx extracted from the Subtensor, thus always selecting the last dim.

Fix bug in get_scalar_constant value of Subtensor(Shape(Rebroadcast(v)))
d27f2398 · Arnaud Bergeron · 02c218fa · d27f2398
--- a/theano/tensor/basic.py
+++ b/theano/tensor/basic.py
@@ -712,11 +712,9 @@ def get_scalar_constant_value(orig_v, elemwise=True,
                    ndim = grandparent.type.ndim
                    if grandparent.owner and isinstance(grandparent.owner.op,
                                                        Rebroadcast):
-                        l = []
+                        ggp_broadcastable = grandparent.owner.inputs[0].broadcastable
-                        for idx, (b1, b2) in enumerate(
+                        l = [b1 or b2 for b1, b2 in zip(ggp_broadcastable,
-                                zip(grandparent.owner.inputs[0].broadcastable,
+                                                        gp_broadcastable)]
-                                    gp_broadcastable)):
-                            l.append(b1 or b2)
                        gp_broadcastable = tuple(l)
                    assert ndim == len(gp_broadcastable)