Sentence Transformers
Library for embedding sentences and passages into vectors.
Обзор
Sentence Transformers provides the training and inference code for bi-encoders and cross-encoders, so a text can be turned into a vector with one call. It solves the problem that a base language model produces a token-level representation which is poor as a single sentence embedding without the right training objective. It exists as the component most retrieval stacks use without naming, and it also hosts the cross-encoders used for reranking.
Граф технологий
Как Sentence Transformers связана с остальным стеком.
Использует и зависит от
Применяется к
- Contrastive Learning technique Algorithms Contrastive Learning used for Sentence Transformers High ✓ source
- Embedding concept Techniques Sentence Transformers used for Embedding High ✓ source
Инструменты
ПО, которое реализует Sentence Transformers или интегрируется с ним.