Skip to content

vllm.models.hy_v4_flash.nvidia.engram

Flash frozen n-gram embeddings with TP head sharding and V2 lookback.