Add for-else (#49521)

* add for-else

* add * for unpacking

python/paddle/fluid/layers/math_op_patch.py

@@ -381,22 +381,20 @@ def monkey_patch_variable():
             lhs_dtype = safe_get_dtype(self)
             if not isinstance(other_var, Variable):
                 if reverse:
-                    has_batch_size = False
                     for elem in self.shape:
                         if elem < 0:
-                            has_batch_size = True
+                            other_var = create_tensor_with_batchsize(
+                                self, other_var, lhs_dtype
+                            )
                             break
-                    if not has_batch_size:
+                    else:
+                        # when break is not triggered, enter the else branch
                         other_var = create_tensor(
                             current_block(self),
                             other_var,
                             dtype=lhs_dtype,
                             shape=self.shape,
                         )
-                    else:
-                        other_var = create_tensor_with_batchsize(
-                            self, other_var, lhs_dtype
-                        )
                 else:
                     # add fill_op to current_block
                     other_var = create_scalar(

python/paddle/fluid/tests/unittests/test_imperative_auto_prune.py

@@ -372,9 +372,9 @@ class TestImperativeAutoPrune(unittest.TestCase):
             loss = model.embed_linear0(indices)
             loss.backward()
             _, params_grads = optimizer.minimize(loss)
-            for items in params_grads:
-                assert items[0].name is not model.embed1.weight.name
-                assert items[0].name is not model.linear_1.weight.name
+            for (items_0, *items_len) in params_grads:
+                assert items_0.name is not model.embed1.weight.name
+                assert items_0.name is not model.linear_1.weight.name
             assert model.embed1.weight._grad_ivar() is None
             assert model.linear_1.weight._grad_ivar() is None