dgl.to_double¶

dgl.to_double(g)[source]¶

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.