SE(3) Exp Torch only working with cpu not with cuda

[saraybakker1]

Hi! Thanks for providing this amazing toolbox!

When using the SE(3) torch version of the function Exp(), I get the error that not all tensors are on the same device (some are on cuda:0 and some on cpu), probably because the device is not specified for all tensors.

For example, in wedge, I think that:
b1 = torch.tensor([0, 0, 0, 0], dtype = xi.dtype).reshape(1, 1, 4)
should be replaced by:
b1 = torch.tensor([0, 0, 0, 0], dtype = xi.dtype, device=xi.device).reshape(1, 1, 4)

and in left_jacobian in so3.py: torch.eye(3, 3).

Thanks! :)

[angadbajwa]

hi, I'm glad you're finding pymlg useful. we're adding CUDA support here: #25. it should be merged soon-ish, but you can work off of device_complex_hotfix while we change things