A mirror of the two files Memoravox's semantic index resolves at runtime: onnx/model_quantized.onnx and tokenizer.json. Held in the memoravox org so the product does not depend on a third-party account for a file it needs on first run.

Upstream: Xenova/paraphrase-multilingual-MiniLM-L12-v2, itself an ONNX export of sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2 (Apache-2.0). No weights were modified.