E5rope Base

🚀 E5-RoPE-Base

This project presents E5-RoPE-Base, a model detailed in the paper LongEmbed: Extending Embedding Models for Long Context Retrieval. It aims to facilitate performance comparisons between embedding models using absolute position embeddings (APE) and rotary position embeddings (RoPE), highlighting the superiority of RoPE-based models in handling long contexts.

🚀 Quick Start

Prerequisites

Ensure you have torch and transformers installed.

Example

Below is an example to encode queries and passages from the MS-MARCO passage ranking dataset.

import torch
import torch.nn.functional as F

from torch import Tensor
from transformers import AutoTokenizer, AutoModel

def average_pool(last_hidden_states: Tensor,
                 attention_mask: Tensor) -> Tensor:
    last_hidden = last_hidden_states.masked_fill(~attention_mask[..., None].bool(), 0.0)
    return last_hidden.sum(dim=1) / attention_mask.sum(dim=1)[..., None]
# Each input text should start with "query: " or "passage: ".
# For tasks other than retrieval, you can simply use the "query: " prefix.
input_texts = ['query: how much protein should a female eat',
               'query: summit define',
               "passage: As a general guideline, the CDC's average requirement of protein for women ages 19 to 70 is 46 grams per day. But, as you can see from this chart, you'll need to increase that if you're expecting or training for a marathon. Check out the chart below to see how much protein you should be eating each day.",
               "passage: Definition of summit for English Language Learners. : 1  the highest point of a mountain : the top of a mountain. : 2  the highest level. : 3  a meeting or series of meetings between the leaders of two or more governments."]
tokenizer = AutoTokenizer.from_pretrained('dwzhu/e5rope-base', trust_remote_code=True)
model = AutoModel.from_pretrained('dwzhu/e5rope-base', trust_remote_code=True).cuda()
# Tokenize the input texts
batch_dict = tokenizer(input_texts, max_length=512, padding=True, truncation=True, return_tensors='pt', pad_to_multiple_of=8)
batch_dict = {k: v.cuda() for k, v in batch_dict.items()}
outputs = model(**batch_dict)
embeddings = average_pool(outputs.last_hidden_state, batch_dict['attention_mask'])
# normalize embeddings
embeddings = F.normalize(embeddings, p=2, dim=1)
scores = (embeddings[:2] @ embeddings[2:].T) * 100
print(scores.tolist())

✨ Features

• Layer and Embedding Size: This model has 12 layers and the embedding size is 768.
• Comparison Purpose: It enables performance comparisons between embedding models using APE and RoPE, demonstrating the advantage of RoPE in long - context handling.

📚 Documentation

Training Details

Please refer to our paper at https://arxiv.org/abs/2404.12096.pdf.

Benchmark Evaluation

Check out unilm/e5 to reproduce evaluation results on the BEIR and MTEB benchmark.

Note

Please note that E5-RoPE-Base is not specifically trained for optimized performance. Its purpose is to enable performance comparisons between embedding models that utilize absolute position embeddings (APE) and rotary position embeddings (RoPE). By comparing E5-Base and E5-RoPE-Base, we demonstrate the superiority of RoPE-based embedding models in effectively managing longer context. See our paper LongEmbed: Extending Embedding Models for Long Context Retrieval for more details.

📄 License

This project is licensed under the MIT license.

📚 Citation

If you find our paper or models helpful, please consider cite as follows:

@article{zhu2024longembed,
  title={LongEmbed: Extending Embedding Models for Long Context Retrieval},
  author={Zhu, Dawei and Wang, Liang and Yang, Nan and Song, Yifan and Wu, Wenhao and Wei, Furu and Li, Sujian},
  journal={arXiv preprint arXiv:2404.12096},
  year={2024}
}