What happened?

The get_raw_values function belonging to anemoi.graphs.nodes.attributes.AreaWeights has an argument named nodes of type torch_geometric.data.storage.NodeStorage.

This torch_geometric.data.storage.NodeStorage class has a member x which is a two-dimensional array with values corresponding to the latitude and longitude values. This NodeStorage class member is a Torch Tensor.

In the get_raw_values function, we have:

def get_raw_values(self, nodes: NodeStorage, *args, **kwargs) -> np.ndarray:
    ...
    latitudes, longitudes = nodes.x[:, 0], nodes.x[:, 1]
    points = latlon_rad_to_cartesian(latitudes, longitudes)
    sv = SphericalVoronoi(points, self.radius, self.centre)

The scipy.spatial.SphericalVoronoi class requires the points value to be a numpy array.

What are the steps to reproduce the bug?

N/A

Version

0.2.2

Platform (OS and architecture)

Linux x86_64

Relevant log output

No response

Accompanying data

No response

Organisation

ECMWF