Flag Embedding LLM Reranker

The flag embedding llm reranker module is a reranker using BAAI LLM-based-Reranker model for passage reranking.

Module Parameters

• batch : The size of a batch. If you have limited CUDA memory, decrease the size of the batch. (default: 64)

• model_name : The type of model you want to use for reranking. Default is “BAAI/bge-reranker-v2-gemma.”

  • you can check a model list at here

• use_fp16 : Whether to use fp16 or not. (default: False)

Example config.yaml

- module_type: flag_embedding_llm_reranker
  batch: 32
  model_name: BAAI/bge-reranker-v2-gemma