Is there interest in ggml_upscale_to_shape supporting non integer scaling factors?

[balisujohn]

So I have a tensor code_emb shaped [43,1024], and I need to upscale it to be [187, 1024], to match the torch code. (dimensions listed here left to right ggml style)

expanded_code_emb = F.interpolate(code_emb, size=187, mode='nearest')

Best I can tell, the ggml_upscale only allows integer scaling factors. A clean way to implement this could be to create another op
ggml_upscale_to_shape which upscales a tensor optionally across all 4 dimensions.

Would there be interest in up-streaming this if I work on it?