vLLM enables LoRA for GTE embedding and reranker models
November 18, 2026
vLLM now supports LoRA adapters for GTE embedding and reranker models, including bert_with_rope. This allows for more efficient fine-tuning and deployment of retrieval models within the runtime.
HOW THIS AFFECTS YOU
●
builderYou can use LoRA to specialize your embedding and reranking models in vLLM.