The definitive vLLM runtime for dual RTX 2080 Ti 22GB + NVLink, delivering 27B/31B local inference with 100+ tok/s single-request decode and native 262K context.
Latest commits.
Builders behind this project.
