single mention · inference & serving
Disaggregated serving across 4+4 GPUs
Uses a four-GPU prefill pool and a four-GPU decode pool in a disaggregated deployment.
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 Disaggregated 4+4.
Filed under inference & serving › serving parallelism