global_tensor = torch.arange(8).view(4, 2)
mesh_2d = init_device_mesh(
    self.device_type, (2, 2), mesh_dim_names=("DP", "TP")
)
tp_mesh = mesh_2d["TP"]
dtensor_tp = distribute_tensor(
    global_tensor, tp_mesh, placements=[Shard(0)]
)

# This is expected, as we don't have DTensor op strategy registered for aten.diagonal.default
# Error -- NotImplementedError: Operator aten.diagonal.default does not have a sharding strategy registered.
# b= torch.diagonal(dtensor_tp)

# This is an example that `_propagate_tensor_meta(op_schema)` get called first and errors out, 
# so it doesn't correctly show the no sharding strategy registered error. 
# Error -- torch._subclasses.fake_tensor.DynamicOutputShapeException: aten.nonzero.default
a = torch.nonzero(dtensor_tp)