BGE-Reranker-v2-M3-ONNX-O4 Open-Source Text Reranking Model - Supports Multilingual Text Pair Relevance Scoring

Bge Reranker V2 M3 Onnx O4

Developed by hooman650

The ONNX O4 version of BGE-RERANKER-V2 is an optimized text reordering model that supports relevance scoring for multilingual text pairs.

Text Classification

Transformers

Open Source License:Apache-2.0 #Text Reordering #Multilingual Support #ONNX Optimization

Downloads 39

Release Time : 6/27/2024

Model Overview

This model is an ONNX quantized version based on BAAI/bge-reranker-v2-m3, primarily used for calculating relevance scores of text pairs, suitable for scenarios such as information retrieval and question-answering systems.

Model Features

ONNX Quantization Optimization

The model has been converted to ONNX format and optimized with O4 quantization, improving inference efficiency

Multilingual Support

Capable of processing text pairs in multiple languages, including Chinese, English, Spanish, and more

High-Precision Relevance Scoring

Provides accurate relevance scoring for text pairs, useful for information retrieval and question-answering systems

Model Capabilities

Text Pair Relevance Scoring

Multilingual Text Processing

Information Retrieval Ranking

Use Cases

Information Retrieval

Search Result Reordering

Reordering search engine results based on relevance

Improves the relevance and accuracy of search results

Question-Answering Systems

Candidate Answer Ranking

Ranking multiple candidate answers generated by a question-answering system based on relevance

Improves the ranking position of the best answer

🚀 ONNX O4 Version of BGE-RERANKER-V2

This project provides an ONNX O4 version of the BGE-RERANKER-V2 model, which can be used for text classification tasks. It demonstrates how to use the model for inference and compares the results with the original non - quantized model.

🚀 Quick Start

Prerequisites

Make sure you have installed the necessary libraries such as optimum, transformers, and torch.

Installation

You can install the required libraries using pip:

pip install optimum transformers torch

💻 Usage Examples

Basic Usage

pairs = [['Odio comer manzana.','I reallly like eating apple'],['I reallly like eating apple', 'Realmente me gusta comer manzana.'], ['I reallly like eating apple', 'I hate apples'],['Las manzanas son geniales.','Realmente me gusta comer manzana.']]

from optimum.onnxruntime import ORTModelForFeatureExtraction,ORTModelForSequenceClassification
from transformers import AutoTokenizer

model_checkpoint = "onnxO4_bge_reranker_v2_m3"

ort_model = ORTModelForSequenceClassification.from_pretrained(model_checkpoint)
tokenizer = AutoTokenizer.from_pretrained(model_checkpoint)

# ONNX Results
import torch


with torch.no_grad():
    inputs = tokenizer(pairs, padding=True, truncation=True, return_tensors='pt', max_length=512)
    scores = ort_model(**inputs, return_dict=True).logits.view(-1, ).float()
    print(scores)

## tensor([ -9.5081,  -3.9569, -10.8632,   0.3756])

Comparison with Original Model

# Original non quantized

from transformers import AutoModelForSequenceClassification, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-reranker-v2-m3')
model = AutoModelForSequenceClassification.from_pretrained('BAAI/bge-reranker-v2-m3')
model.eval()

with torch.no_grad():
    inputs = tokenizer(pairs, padding=True, truncation=True, return_tensors='pt', max_length=512)
    scores = model(**inputs, return_dict=True).logits.view(-1, ).float()
    print(scores)

## tensor([ -9.4973,  -3.9538, -10.8504,   0.3660])

📄 License

This project is licensed under the Apache-2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご