AI Briefingall-MiniLM-L6-v2: 250 Million Monthly Downloads: Ultra-Lightweight Embedding Capturing Sentence Meaning with 384-Dimensional Vectors
sentence-transformers/all-MiniLM-L6-v2
About the project
It converts sentences or short paragraphs into 384-dimensional dense vectors for semantic similarity calculation and semantic search. Since input text exceeding 256 word pieces is truncated, it is optimized for sentence-level processing rather than long documents.
It was trained using self-supervised contrastive learning on a large-scale dataset consisting of over 1.1 billion sentence pairs. By learning from diverse sources such as Reddit comments, Stack Exchange, and MS MARCO, it understands a wide range of contexts from everyday conversation to technical documentation.
With a lightweight architecture of 22.7M parameters, it supports multiple frameworks including PyTorch, TensorFlow, and ONNX. Recording over 250 million downloads per month, it is used as a standard when fast inference speed is required in the retrieval stage of RAG pipelines or clustering tasks.
sentence-transformers/all-MiniLM-L6-v2
The original page has no description.
sentence-similarity
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.