Bge M3 En Ru

🚀 BGE-M3 Model for English and Russian

This model is designed for sentence similarity tasks, offering feature extraction capabilities for both English and Russian languages. It's a tokenizer - shrinked version of [BAAI/bge - m3](https://huggingface.co/BAAI/bge - m3), which retains only English and Russian tokens in the vocabulary. As a result, the vocabulary size is 21% of the original, and the number of parameters in the whole model is 63.3% of the original, without any loss in the quality of English and Russian embeddings.

🚀 Quick Start

✨ Features

• Tokenizer - shrinked version for English and Russian.
• Maintains embedding quality with reduced vocabulary and parameters.

📦 Installation

If you want to use this model, you need to install sentence - transformers first:

pip install -U sentence-transformers

💻 Usage Examples

Basic Usage

Using the sentence - transformers library:

from sentence_transformers import SentenceTransformer
sentences = ["This is an example sentence", "Each sentence is converted"]

model = SentenceTransformer('TatonkaHF/bge-m3_en_ru')
embeddings = model.encode(sentences)
print(embeddings)

Advanced Usage

Without the sentence - transformers library, using the transformers library directly:

from transformers import AutoTokenizer, AutoModel
import torch


#Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output[0] #First element of model_output contains all token embeddings
    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)


# Sentences we want sentence embeddings for
sentences = ['This is an example sentence', 'Each sentence is converted']

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('TatonkaHF/bge-m3_en_ru')
model = AutoModel.from_pretrained('TatonkaHF/bge-m3_en_ru')

# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')

# Compute token embeddings
with torch.no_grad():
    model_output = model(**encoded_input)

# Perform pooling. In this case, mean pooling.
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])

print("Sentence embeddings:")
print(sentence_embeddings)

📚 Documentation

Specs

Other bge - m3 models are also shrinked.

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: XLMRobertaModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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})
)

🔧 Technical Details

The model is based on the concept of self - knowledge distillation. The tokenizer shrinking process is inspired by [LaBSE - en - ru](https://huggingface.co/cointegrated/LaBSE - en - ru) and [https://discuss.huggingface.co/t/tokenizer - shrinking - recipes/8564/1](https://discuss.huggingface.co/t/tokenizer - shrinking - recipes/8564/1).

📄 License

This model is licensed under the MIT license.

Reference

Jianlv Chen, Shitao Xiao, Peitian Zhang, Kun Luo, Defu Lian, Zheng Liu. BGE M3 - Embedding: Multi - Lingual, Multi - Functionality, Multi - Granularity Text Embeddings Through Self - Knowledge Distillation.