search

UKPLab / sentence-transformers

State-of-the-Art Text Embeddings
https://www.sbert.net
Apache License 2.0
14.78k stars 2.43k forks source link

Difference between 'stsb-roberta-large' and 'roberta-large-nli-stsb-mean-tokens' #656

Open jiachangliu opened 3 years ago

jiachangliu commented 3 years ago

What is the difference between the two pre-trained models 'stsb-roberta-large' and 'roberta-large-nli-stsb-mean-tokens'?

I believe both models are trained on NLI and then fine-tuned on stsb. Additionally, both models use mean pooling. Am I right?

nreimers commented 3 years ago

These are identical models.

Starting with version 4, I changed the naming to put the task name first and the model type last.

Before that, the model name was first followed by the task and the pooling method.

Swty13 commented 1 year ago

Hi @nreimers ,

What is stsb in 'stsb-roberta-large' ?

Thanks in advance !