Ruri: Japanese General Text Embeddings

Notes: v3 models are out!
We recommend using the following v3 models going forward.

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers fugashi sentencepiece unidic-lite

Then you can load this model and run inference.

import torch.nn.functional as F
from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("cl-nagoya/ruri-small", trust_remote_code=True)

# Don't forget to add the prefix "クエリ: " for query-side or "文章: " for passage-side texts.
sentences = [
    "クエリ: 瑠璃色はどんな色？",
    "文章: 瑠璃色（るりいろ）は、紫みを帯びた濃い青。名は、半貴石の瑠璃（ラピスラズリ、英: lapis lazuli）による。JIS慣用色名では「こい紫みの青」（略号 dp-pB）と定義している[1][2]。",
    "クエリ: ワシやタカのように、鋭いくちばしと爪を持った大型の鳥類を総称して「何類」というでしょう?",
    "文章: ワシ、タカ、ハゲワシ、ハヤブサ、コンドル、フクロウが代表的である。これらの猛禽類はリンネ前後の時代(17~18世紀)には鷲類・鷹類・隼類及び梟類に分類された。ちなみにリンネは狩りをする鳥を単一の目(もく)にまとめ、vultur(コンドル、ハゲワシ)、falco(ワシ、タカ、ハヤブサなど)、strix(フクロウ)、lanius(モズ)の4属を含めている。",
]

embeddings = model.encode(sentences, convert_to_tensor=True)
print(embeddings.size())
# [4, 768]

similarities = F.cosine_similarity(embeddings.unsqueeze(0), embeddings.unsqueeze(1), dim=2)
print(similarities)
# [[1.0000, 0.9453, 0.6860, 0.7225],
#  [0.9453, 1.0000, 0.6852, 0.7005],
#  [0.6860, 0.6852, 1.0000, 0.8567],
#  [0.7225, 0.7005, 0.8567, 1.0000]]

Benchmarks

JMTEB

Evaluated with JMTEB.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: cl-nagoya/ruri-pt-small
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768
• Similarity Function: Cosine Similarity
• Language: Japanese
• License: Apache 2.0
• Paper: https://arxiv.org/abs/2409.07737

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: DistilBertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

Training Details

Framework Versions

• Python: 3.10.13
• Sentence Transformers: 3.0.0
• Transformers: 4.41.2
• PyTorch: 2.3.1+cu118
• Accelerate: 0.30.1
• Datasets: 2.19.1
• Tokenizers: 0.19.1

Citation

@misc{
  Ruri,
  title={{Ruri: Japanese General Text Embeddings}}, 
  author={Hayato Tsukagoshi and Ryohei Sasano},
  year={2024},
  eprint={2409.07737},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2409.07737}, 
}

License

This model is published under the Apache License, Version 2.0.