The error information is as follows:
RuntimeError: eig_ backward: Backward calculation does not support complex eigenvalues at the moment.

The details are as follows:
One step in my loss function is to calculate the eigenvalues of the matrix. I used the torch.eig function, but later when I used the reverse derivation of the loss function, the program was stuck in the position of the torch.eig function, because sometimes the matrix has complex eigenvalues. Under what circumstances, the torch.eig function does not support reverse derivation! How can I solve it???

Please help me！！！

cc @ezyang @albanD @zou3519 @gqchen @pearu @nikitaved @soulitzer @anjali411 @dylanbespalko @mruberry @jianyuh @heitorschueroff @walterddr @IvanYashchuk