ONNX export sentence-transformers

onnx-export.py

from sentence_transformers import (
    SentenceTransformer,
    export_optimized_onnx_model,
    export_dynamic_quantized_onnx_model,
)

# 1. Load the model to be optimized with the ONNX backend
model = SentenceTransformer(
    "IEITYuan/Yuan-embedding-2.0-en",
    backend="onnx",
)

# 2. Export the model with O4 optimization level
export_optimized_onnx_model(
    model,
    optimization_config="O4",
    model_name_or_path="thomasht86/Yuan-embedding-2.0-en-ONNX",
    push_to_hub=True,
)

for quantization_config in ["arm64", "avx2", "avx512", "avx512_vnni"]:
    # 2. Export the model with static quantization
    export_dynamic_quantized_onnx_model(
        model,
        quantization_config=quantization_config,
        model_name_or_path="thomasht86/Yuan-embedding-2.0-en-ONNX",
        push_to_hub=True,
    )