Nomic Embed Text v1.5 — ONNX
ONNX export of nomic-ai/nomic-embed-text-v1.5, a long-context English embedding model. Maps text to 768-dimensional dense vectors using mean pooling, with Matryoshka training so the vectors can be truncated to 512, 256, 128 or 64 dimensions. Inputs need a task prefix such as search_query: or search_document: .
Mirrored for use with inference4j, an inference-only AI library for Java.
Original Source
- Repository: Nomic AI
- License: apache-2.0
Usage with inference4j
try (SentenceTransformerEmbedder model = SentenceTransformerEmbedder.builder()
.modelId("inference4j/nomic-embed-text-v1.5")
.textPrefix("search_query: ")
.normalize()
.build()) {
float[] embedding = model.encode("What is the capital of France?");
System.out.println("Dimension: " + embedding.length); // 768
}
Model Details
| Property | Value |
|---|---|
| Architecture | NomicBERT (12 layers, 768 hidden, rotary positions, SwiGLU) |
| Task | Sentence embeddings / semantic search |
| Output dimension | 768 (Matryoshka: 512, 256, 128, 64) |
| Pooling | Mean |
| Normalization | L2 |
| Task prefixes | search_query: , search_document: , clustering: , classification: |
| Max sequence length | 8192 |
| Original framework | PyTorch (HuggingFace Transformers) |
License
This model is licensed under the Apache License 2.0. Original model and ONNX export by Nomic AI.