Open Models Almanac
Methodsmodel architecture

single source · model architecture

Spatially shared FiLM modulation across vision-transformer patch embeddings

A FiLM-like modulation applying corresponding embedding-dimension values across visual patch embeddings.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See model architecture for the methods filed beside it.

Also called spatially-agnostic modulation approach.

Filed under model architecture