Modulation¶

`BaseModulation`	Transform a fixed tensor using a latent vector.
`AdditiveModulation`	Apply `W' = W + alpha * z` across the rows of a matrix.
`AffineModulation`	Apply per-column latent scale and shift to a matrix.
`LowRankModulation`	Decode `z` into factors and apply `W' = W + alpha/rank * AB`.