Tensor parallelism

Tensor parallelism: Splitting individual layers of a model across devices so one large operation runs in parallel on several chips at once.

Browse the full glossary →