E5 Large En Ru

🚀 E5-large-en-ru

This is a vocabulary pruned model based on intfloat/multilingual-e5-large, which only uses Russian and English tokens. It offers comparable performance to the original model while having a smaller size.

📚 Documentation

Model info

This is a vocabulary pruned version of intfloat/multilingual-e5-large, using only Russian and English tokens.

Size

Performance

Equal performance on SberQuAD dev benchmark.

💻 Usage Examples

Basic Usage

import torch.nn.functional as F
from torch import Tensor
from transformers import XLMRobertaTokenizer, XLMRobertaModel


def average_pool(last_hidden_states: Tensor, attention_mask: Tensor) -> Tensor:
    last_hidden = last_hidden_states.masked_fill(~attent

Usage Tips

⚠️ Important Note

• Use dot product distance for retrieval.
• Use "query: " and "passage: " correspondingly for asymmetric tasks such as passage retrieval in open QA, ad - hoc information retrieval.
• Use "query: " prefix for symmetric tasks such as semantic similarity, bitext mining, paraphrase retrieval.
• Use "query: " prefix if you want to use embeddings as features, such as linear probing classification, clustering.

📄 License

This model is released under the MIT license.