texttron / tevatron

Tevatron - A flexible toolkit for neural retrieval research and development.
http://tevatron.ai
Apache License 2.0
531 stars 100 forks source link

Reproducing repllama on document retrieval #151

Open hengran opened 2 months ago

hengran commented 2 months ago

Hi, @MXueguang~ I am currently reproducing repllama on document retrieval of MSMARCO. In order to reproduce successfully, I need a training set that is consistent with the paper. I would like to ask if you can provide me with the training set for document retrieval, espcially hard negative documents from CoCondenser-MaxP.

hengran commented 2 months ago

Hi~ I find the document link that you have providedhttps://github.com/texttron/tevatron/issues/103#issuecomment-1905430706. I wonder if this link is hard negative for Repllama or Rankllama?

MXueguang commented 2 months ago

HI @hengran , sorry for the late response, the above link is for repllama

hengran commented 2 months ago

HI @hengran , sorry for the late response, the above link is for repllama

thanks for you response.