This model is fine-tuned from intfloat/multilingual-e5-base and use our dataset It's then converted to onnx format
