Colbert Reranker

The colber reranker module is a reranker using ColBERT 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 “colbert-ir/colbertv2.0”.

Example config.yaml

- module_type: colbert_reranker
  batch: 64
  model_name: colbert-ir/colbertv2.0