config_sentence_transformers.json : sentence_similarity_fn_name

#2
by nadshe - opened

Is there are reason the default similarity name is "cosine" and not "dot"? Was this model not trained with the dot function for scoring?

{
  "model_type": "SparseEncoder",
  "__version__": {
    "sentence_transformers": "5.0.0",
    "transformers": "4.52.4",
    "pytorch": "2.6.0+cu124"
  },
  "prompts": {
    "query": "",
    "document": ""
  },
  "default_prompt_name": null,
  "similarity_fn_name": "cosine"
}
NeuML org

Hello. It's designed to work with cosine similarity (or dot similarity with normalized vectors).

Sign up or log in to comment