Maps sentences and short paragraphs to 384-dimensional vectors for semantic search, clustering and similarity. Small and fast.
About
A six-layer MiniLM model fine-tuned by the Sentence Transformers project on over a billion sentence pairs. It turns text into a 384-dimensional vector so that similar meanings land close together — the usual starting point for semantic search and retrieval-augmented generation. Input longer than 256 word pieces is truncated.
Use it
from sentence_transformers import SentenceTransformer
model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
vectors = model.encode(["A man is eating food.", "Someone is having a meal."])
The ONNX export (onnx/model.onnx) runs without PyTorch, for example with ONNX Runtime; pair it with tokenizer.json and apply mean pooling and normalisation yourself.
Files
Files link to the original repository on Hugging Face. English only.
Licence
Apache-2.0, from the Sentence Transformers project.