Cast this graph to use float64 (double-precision) for any floating-point edge and node feature data.
A shallow copy is returned so that the original graph is not modified. Feature tensors that are not floating-point will not be modified.
Clone of graph with the feature data converted to float64.
- Return type