Xyc/liespline

[XyC0212]

change the type of input and output to LieTensor

[wang-chen]

this will have issues if inputs tensors is on GPU.

[XyC0212]

Have modified the code
`
alpha = torch.arange(4, dtype=time.dtype, device=time.device)

    tt = time[:, None, :] ** alpha[None, :, None]

    B = torch.tensor([[5, 3, -3, 1],
                      [1, 3, 3, -2],
                      [0, 0, 0, 1]], dtype=time.dtype, device=time.device) / 6`

[wang-chen]

Here are some overall comments:

• No data is stored in the module, thus I suggest defining it as a function instead of a module.
• many lines exceeding 90 characters. refer to 2nd point in developer guideline.
• use tensor[..., i] to support batch operation instead of tensor[:, i], refer to 10th point in developer guideline.
• some display issues in the doc, please view this pr doc