Fix test in DebugMode. Make GpuAdvancedIncSubtensor1 allow to broadcast

theano/sandbox/cuda/basic_ops.py

@@ -2604,10 +2604,10 @@ class GpuAdvancedIncSubtensor1(tensor.AdvancedIncSubtensor1, GpuOp):
         out, = out_
         if not self.inplace:
             x = x.copy()
+        assert y.ndim <= x.ndim   # Should be guaranteed by `make_node`
         if self.set_instead_of_inc:
             # CudaNdarray __setitem__ doesn't do broadcast nor support
             # list of index.
-            assert y.ndim <= x.ndim   # Should be guaranteed by `make_node`
             if y.ndim == x.ndim:
                 assert len(y) == len(idx)
                 for (j, i) in enumerate(idx):
@@ -2619,11 +2619,15 @@ class GpuAdvancedIncSubtensor1(tensor.AdvancedIncSubtensor1, GpuOp):
             # If `y` has as many dimensions as `x`, then we want to iterate
             # jointly on `x` and `y`. Otherwise, it means `y` should be
             # broadcasted to fill all relevant rows of `x`.
-            assert y.ndim <= x.ndim   # Should be guaranteed by `make_node`
             if y.ndim == x.ndim:
-                assert len(y) == len(idx)
-                for (j, i) in enumerate(idx):
-                    x[i] += y[j]
+                if len(y) == 1:
+                    # Allow broadcasting of y[0]
+                    for i in idx:
+                        x[i] += y[0]
+                else:
+                    assert len(y) == len(idx)
+                    for (j, i) in enumerate(idx):
+                        x[i] += y[j]
             else:
                 for i in idx:
                     x[i] += y