which loss function was used?
#5
by
npotts
- opened
Which loss function was used to fine tune this model? Euclidean distance, cosine similarity?
We used cosine similarity, normalizing the output of the model (last token pooling)
npotts
changed discussion status to
closed