It is a TT-compressed model of bert-base-uncased stsb model.
Model was trained on STSB corpus with 0.87 combined score, and TTM-compressed with additional finetuning up to 58% (64 mln params) of original size with 0.843 score.
How to use
python
1 tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")2 model = AutoModelForSequenceClassification.from_pretrained("s-nlp/bert-base-uncased-stsb-TTM", trust_remote_code=True)
