torchcubicspline icon indicating copy to clipboard operation
torchcubicspline copied to clipboard

Interpolating natural cubic splines. Includes batching, GPU support, support for missing values, evaluating derivatives of the spline, and backpropagation.

Results 7 torchcubicspline issues
Sort by recently updated
recently updated
newest added

Hello, Thanks for this very useful package. I have a question, Does this package support the clamped cubic spline, i.e., s'(t_0)=a, s'(t_T)=b? Qunxi

The function `_validate_input` seems to be incorrect. The code tries to block non-monotonic data, but since `prev_t_i` is not updated, it doesn't seem to work. And in a Cuda environment,...

Hi. As you warn in the code, `tridiagonal_solve()` is quite slow. I've compared to plain `torch.solve()`, which is much faster, so I'll be using that in my application, but I...

Hi I wonder if this code works for 2 or 3 dimensional data! Thanks.

Hi, Is it possible to apply the cubic spline interpolation on one dimension of a data with the data format as in PyTorch `[BatchSize, Channel, Height, Width]`? Thanks

Hi, I've got a tensor of the shape batch_size x T x nr_channels and I created a cubic spline accordingly. Let's say I have created this tensor with shape (2,...

Hi Patrick, Would it be possible to have torchcubicspline available on pypi? I have a package that uses it (https://github.com/rorymaizels/velvet) and I would like to make this package available for...