multi-qa-MiniLM-L6-cos-v1
| Property | Value |
|---|---|
| Parameter Count | 22.7M |
| Embedding Dimensions | 384 |
| Training Data | 215M QA pairs |
| Model Type | Sentence Transformer |
What is multi-qa-MiniLM-L6-cos-v1?
multi-qa-MiniLM-L6-cos-v1 is a compact but powerful sentence embedding model designed specifically for semantic search applications. It transforms text inputs into 384-dimensional dense vector representations, enabling efficient similarity comparisons between sentences and paragraphs. The model was trained on an extensive dataset of 215 million question-answer pairs from diverse sources including WikiAnswers, Stack Exchange, and MS MARCO.
Implementation Details
The model implements a mean pooling architecture with normalized embeddings and supports multiple frameworks including PyTorch, TensorFlow, and ONNX. It processes text up to 512 word pieces (with optimal performance for texts under 250 word pieces) and produces normalized embeddings that can be compared using dot-product or cosine similarity metrics.
- Optimized for semantic search and question-answering tasks
- Supports multiple deep learning frameworks
- Efficient 384-dimensional embeddings
- Pre-trained on MiniLM-L6-H384-uncased architecture
Core Capabilities
- Semantic search across document collections
- Question-answer matching
- Text similarity comparison
- Dense passage retrieval
- Cross-encoder pre-screening
Frequently Asked Questions
Q: What makes this model unique?
The model's distinctive feature is its optimal balance between size and performance, trained on one of the largest and most diverse question-answer datasets (215M pairs). It produces normalized embeddings that enable fast similarity computations using simple dot products.
Q: What are the recommended use cases?
The model excels in semantic search applications, document retrieval, and question-answering systems. It's particularly effective for applications requiring fast and accurate semantic similarity matching between shorter texts (up to 250 word pieces).
