bowdpr_wiki_triviaft

This is a fine-tuned retriever on the TriviaQA Task (without distillation). We introduce a novel pre-training paradigm, Bag-of-Word Prediction, for dense retrieval. This retriever is initialized from a base-sized pre-trained model, bowdpr/bowdpr_wiki. Please refer to our paper for detailed pre-training and fine-tuning settings.

Finetuning on QA datasets involves a two-stage pipeline

• s1: BM25 negs
• s2: BM25 negs + Mined negatives from s1

Usage (Sentence-Transformers)

This is a sentence-transformers model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search.

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

pip install -U sentence-transformers

Then you can use the model like this:

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

model = SentenceTransformer('bowdpr/bowdpr_wiki_triviaft')
embeddings = model.encode(sentences)
print(embeddings)

Usage (HuggingFace Transformers)

Without sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.

from transformers import AutoTokenizer, AutoModel
import torch


def cls_pooling(model_output, attention_mask):
    return model_output[0][:,0]


# 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('bowdpr/bowdpr_wiki_triviaft')
model = AutoModel.from_pretrained('bowdpr/bowdpr_wiki_triviaft')

# 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, cls pooling.
sentence_embeddings = cls_pooling(model_output, encoded_input['attention_mask'])

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

Full Model Architecture

SentenceTransformerforCL(
  (0): Transformer({'max_seq_length': 256, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
)

Citing & Authors

If you are interested in our work, please consider citing our paper.

@misc{ma2024bow_pred,
      title={Drop your Decoder: Pre-training with Bag-of-Word Prediction for Dense Passage Retrieval}, 
      author={Guangyuan Ma and Xing Wu and Zijia Lin and Songlin Hu},
      year={2024},
      eprint={2401.11248},
      archivePrefix={arXiv},
      primaryClass={cs.IR}
}