This repository implements an Interp1d class that overrides torch.autograd.Function, enabling linear 1D interpolation on the GPU for Pytorch.

class Interp1d(torch.autograd.Function):
    def forward(ctx, x, y, xnew, out=None)

This function returns interpolated values of a set of 1-D functions at the desired query points xnew.

It works similarly to Matlab™ or scipy functions with the linear interpolation mode on, except that it parallelises over any number of desired interpolation problems and exploits CUDA on the GPU

• x : a (N, ) or (D, N) Pytorch Tensor: Either 1-D or 2-D. It contains the coordinates of the observed samples.

• y : (N,) or (D, N) Pytorch Tensor. Either 1-D or 2-D. It contains the actual values that correspond to the coordinates given by x. The length of y along its last dimension must be the same as that of x

• xnew : (P,) or (D, P) Pytorch Tensor. Either 1-D or 2-D. If it is not 1-D, its length along the first dimension must be the same as that of whichever x and y is 2-D. x-coordinates for which we want the interpolated output.

• out : (D, P) Pytorch Tensor` Tensor for the output. If None: allocated automatically.

a Pytorch tensor of shape (D, P), containing the interpolated values.

The CUDA interp1d function depends on the torchsearchsorted repository.

If not installed, you must:

1. Clone that repo through git clone [email protected]:aliutkus/torchsearchsorted.git
2. Go in the corresponding directory of that repo and launch python setup.py install.

Then, type pip install -e . in the root folder of this repo.

Try out python test.py in the examples folder.

Solving 100000 interpolation problems: each with 100 observations and 30 desired values
CPU: 8060.260ms, GPU: 70.735ms, error: 0.000000%.