ms-marco-MiniLM-L-6-v2 (Cross-Encoder) — ONNX

ONNX export of ms-marco-MiniLM-L-6-v2, a cross-encoder model trained on MS MARCO passage ranking data. Scores query-document pairs for search result reranking.

Mirrored for use with inference4j, an inference-only AI library for Java.

Original Source

• Repository: Xenova (originally cross-encoder by sentence-transformers)
• License: apache-2.0

Usage with inference4j

try (MiniLMReranker reranker = MiniLMReranker.fromPretrained("models/ms-marco-MiniLM-L-6-v2")) {
    float score = reranker.score("What is Java?", "Java is a programming language.");
    System.out.printf("Relevance score: %.4f%n", score);
}

Model Details

Property	Value
Architecture	MiniLM-L6 cross-encoder (6 layers, 384 hidden)
Task	Query-document relevance scoring / search reranking
Training data	MS MARCO passage ranking
Max sequence length	512
Output	Single relevance score per query-document pair
Original framework	PyTorch (sentence-transformers cross-encoder)
ONNX export	By Xenova (Transformers.js)

License

This model is licensed under the Apache License 2.0. Original model by sentence-transformers, ONNX export by Xenova.