README.md

metadata

language:
  - ko
  - en
  - zh
license: mit
pipeline_tag: feature-extraction
tags:
  - transformers
  - sentence-transformers
  - text-embeddings-inference

upskyy/ko-reranker

ko-reranker는 BAAI/bge-reranker-large 모델에 한국어 데이터를 finetuning 한 model 입니다.

Usage

Using FlagEmbedding

pip install -U FlagEmbedding

Get relevance scores (higher scores indicate more relevance):

from FlagEmbedding import FlagReranker


reranker = FlagReranker('upskyy/ko-reranker', use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation

score = reranker.compute_score(['query', 'passage'])
print(score) # -1.861328125

# You can map the scores into 0-1 by set "normalize=True", which will apply sigmoid function to the score
score = reranker.compute_score(['query', 'passage'], normalize=True)
print(score) # 0.13454832326359276

scores = reranker.compute_score([['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']])
print(scores) # [-7.37109375, 8.5390625]

# You can map the scores into 0-1 by set "normalize=True", which will apply sigmoid function to the score
scores = reranker.compute_score([['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']], normalize=True)
print(scores) # [0.0006287840192903181, 0.9998043646624727]

Using Sentence-Transformers

pip install -U sentence-transformers

Get relevance scores (higher scores indicate more relevance):

from sentence_transformers import SentenceTransformer


sentences_1 = ["경제 전문가가 금리 인하에 대한 예측을 하고 있다.", "주식 시장에서 한 투자자가 주식을 매수한다."]
sentences_2 = ["한 투자자가 비트코인을 매수한다.", "금융 거래소에서 새로운 디지털 자산이 상장된다."]

model = SentenceTransformer('upskyy/ko-reranker')

embeddings_1 = model.encode(sentences_1, normalize_embeddings=True)
embeddings_2 = model.encode(sentences_2, normalize_embeddings=True)
similarity = embeddings_1 @ embeddings_2.T

print(similarity)

Using Huggingface transformers

Get relevance scores (higher scores indicate more relevance):

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer


tokenizer = AutoTokenizer.from_pretrained('upskyy/ko-reranker')
model = AutoModelForSequenceClassification.from_pretrained('upskyy/ko-reranker')
model.eval()

pairs = [['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']]

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)

Citation

@misc{bge_embedding,
      title={C-Pack: Packaged Resources To Advance General Chinese Embedding}, 
      author={Shitao Xiao and Zheng Liu and Peitian Zhang and Niklas Muennighoff},
      year={2023},
      eprint={2309.07597},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

License

FlagEmbedding is licensed under the MIT License. The released models can be used for commercial purposes free of charge.

Reference

• Dongjin-kr/ko-reranker
• reranker-kr
• FlagEmbedding