PTT5-base Reranker finetuned on both English and Portuguese MS MARCO

Introduction

ptt5-base-msmarco-en-pt-100k-v2 is a T5-based model pretrained in the BrWac corpus, fine-tuned on both English and Portuguese translated version of MS MARCO passage dataset. In the v2 version, the Portuguese dataset was translated using Google Translate. This model was finetuned for 100k steps. Further information about the dataset or the translation method can be found on our mMARCO: A Multilingual Version of MS MARCO Passage Ranking Dataset and mMARCO repository.

Usage


from transformers import T5Tokenizer, T5ForConditionalGeneration

model_name = 'unicamp-dl/ptt5-base-msmarco-en-pt-100k-v2'
tokenizer  = T5Tokenizer.from_pretrained(model_name)
model      = T5ForConditionalGeneration.from_pretrained(model_name)

Citation

If you use ptt5-base-msmarco-en-pt-100k-v2, please cite:

@misc{bonifacio2021mmarco,
  title={mMARCO: A Multilingual Version of MS MARCO Passage Ranking Dataset}, 
  author={Luiz Henrique Bonifacio and Vitor Jeronymo and Hugo Queiroz Abonizio and Israel Campiotti and Marzieh Fadaee and  and Roberto Lotufo and Rodrigo Nogueira},
  year={2021},
  eprint={2108.13897},
  archivePrefix={arXiv},
  primaryClass={cs.CL}

}