How to train and set the training data ？
#2
by
koshinryuu
- opened
how to construct the training data pair and loss function , could you provide more details? Thanks.
How to set the similarity scores for the training sentence pairs, could you provide more details? In my opinion, if two sentences have the same meaning, the similarity is 1. But if all the similarity scores are 1.0, the training will result in error. Your examples only show sentence pairs with similarity of 1.0. Do you have any ideas of constructing labels below 1.0 or giving some examples? Thanks.
same question, how to obtain the negative samples, i.e., the sentence pair with low scores ?