| Property | Value |
|---|---|
| Author | sentence-transformers |
| Vector Dimension | 768 |
| Paper | Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks |
| Status | Deprecated |
What is bert-base-nli-stsb-mean-tokens?
This is a deprecated sentence transformer model that maps sentences and paragraphs to 768-dimensional dense vector spaces. While historically used for tasks like clustering and semantic search, it is no longer recommended due to its relatively low-quality embeddings.
Implementation Details
The model architecture consists of a BERT-based transformer followed by a pooling layer. It uses mean pooling over token embeddings and can process sequences up to 128 tokens. The model can be implemented using either the sentence-transformers library or HuggingFace Transformers.
- Built on BERT-base architecture
- Uses mean pooling strategy
- Maximum sequence length of 128 tokens
- Implements Siamese network architecture
Core Capabilities
- Sentence and paragraph embedding generation
- Semantic similarity comparison
- Text clustering
- Semantic search operations
Frequently Asked Questions
Q: What makes this model unique?
This model was one of the early implementations of Sentence-BERT architecture, using mean token pooling strategy. However, it has been superseded by more modern alternatives available on SBERT.net.
Q: What are the recommended use cases?
Due to its deprecated status, it is NOT recommended for any production use cases. Users should instead refer to the newer models listed on SBERT.net for better performance in sentence embedding tasks.