Why is there 2 out_features in the classifier ?
Hi,
I was surprised to see that the model has 2 out_features : (classifier): Linear(in_features=768, out_features=2, bias=True)
Code :
from llama_index.indices.postprocessor import SentenceTransformerRerank
reranker = SentenceTransformerRerank(top_n=4, model="amberoad/bert-multilingual-passage-reranking-msmarco")
reranker._model.dict
What is the meaning of the second score ?
same question here
same, I print(model), and see that the output the the last linear layer is 2.
@jmaronasm
I haven't touched the model in a long time, and my documentation from back then is not very extensive.
The model is by now heavily outdated when it comes to speed and quality.
You should better switch to Colbert https://github.com/stanford-futuredata/ColBERT '
We are currently thinking about making some of the experiements from colbert open source (see old thread here: https://github.com/stanford-futuredata/ColBERT/issues/12)
thanks so much for the reply