Concepts¶
These pages explain the mental model behind VIPP.
Start with:
VIPP is easiest to use when you treat every graph edge as a scientific object: an image, mask, label image, table, PSF, or future data type with a specific meaning.
In 0.13, apply the same distinction to execution: the workflow stores a portable CPU / Auto / Prefer GPU / Custom request, while an accepted run records the actual CPU/CuPy/cuCIM implementation and any fallback. See choose and verify compute.