Open Models Almanac
Methodsinference & servingserving parallelism

single mention · inference & serving

Data-parallel vision encoding

Uses data parallelism for multimodal vision encoding, as selected by the encoder tensor-parallel mode.

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

Also called --mm-encoder-tp-mode data.

Filed under inference & serving › serving parallelism