For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo function

cubic_kernel

def cubic_kernel(x: Float32) -> Float32

Cubic interpolation kernel matching PyTorch/torchvision's BICUBIC filter.

This uses the Catmull-Rom variant (Robidoux cubic) with a = -0.75, which is what PyTorch uses in get_cubic_upsample_coefficients. (Source). This also matches OpenCV's interpolateCubic.

Args:

x (Float32): Distance from the center point.

Returns:

Float32: Weight contribution based on the distance.

def cubic_kernel(x: SIMD) -> SIMD[x.dtype, x.size]

Cubic interpolation kernel matching PyTorch/torchvision's BICUBIC filter.

This uses the Catmull-Rom variant (Robidoux cubic) with a = -0.75, which is what PyTorch uses in get_cubic_upsample_coefficients. (Source). This also matches OpenCV's interpolateCubic.

Args:

x (SIMD): Distance from the center point.

Returns:

SIMD[x.dtype, x.size]: Weight contribution based on the distance.