HPD MiniLM F128

🚀 HPD-MiniLM-F128

This repository provides pre-trained models for semantic retrieval sentence representation compression, offering high efficiency and quality.

🚀 Quick Start

This repository contains the pre-trained models for our paper Compressing Sentence Representation for Semantic Retrieval via Homomorphic Projective Distillation. The sentence embedding model contains only 23M parameters and the model size is only 87MB.

✨ Features

We propose Homomorphic Projective Distillation (HPD) to learn compressed sentence embeddings. Our method augments a small Transformer encoder model with learnable projection layers to produce compact representations while mimicking a large pre-trained language model to retain the sentence representation quality.

📦 Installation

Using this model becomes easy when you have sentence-transformers installed:

pip install -U sentence-transformers

💻 Usage Examples

Basic Usage

After installing the package, you can simply load our model

from sentence_transformers import SentenceTransformer
model = SentenceTransformer('Xuandong/HPD-MiniLM-F128')

Advanced Usage

Then you can use our model for encoding sentences into embeddings

sentences = ['He plays guitar.', 'A street vendor is outside.']
sentence_embeddings = model.encode(sentences)

for sentence, embedding in zip(sentences, sentence_embeddings):
    print("Sentence:", sentence)
    print("Embedding:", embedding)
    print("")

📚 Documentation

This is a sentence-transformers model: It maps sentences & paragraphs to a 128 dimensional dense vector space and can be used for tasks like clustering or semantic search.

The teacher model is princeton-nlp/sup-simcse-roberta-large and the student model is nreimers/MiniLM-L6-H384-uncased.

Evaluation Results

We evaluate our model on semantic textual similarity (STS) tasks. The results are:

Training

Please refer to the github repo (https://github.com/XuandongZhao/HPD) for the details about the training.

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
  (2): Dense({'in_features': 384, 'out_features': 128, 'bias': True, 'activation_function': 'torch.nn.modules.activation.Tanh'})
)

📄 License

This project is licensed under the Apache-2.0 license.

📖 Citation

Please cite our paper if you use HPD in your work:

@article{zhao2022compressing,
  title={Compressing Sentence Representation for Semantic Retrieval via Homomorphic Projective Distillation},
  author={Zhao, Xuandong and Yu, Zhiguo and Wu, Ming and Li, Lei},
  journal={arXiv preprint arXiv:2203.07687},
  year={2022}
}