Convert from original model to ncnn

[Siziff]

How do you convert the original model to ncnn? If there is no WARP layer in ncnn, then how can it be converted with a dimension of not 3 channels? in the inference function, the flow dimension has 2 channels instead of the required three to replace, as they do here - https://github.com/TNTwise/REAL-Video-Enhancer/wiki/Convert-Rife-Models