mContriever pre-built index for Mr.TyDi datasets

[cramraj8]

Hi @lintool ,

I see mContriever pre-built index available for MIRACL benchmark, but not for My.TyDi benchmark. However, mDPR is available for both. Any particular reasons why mContriever for TyDi not available ? If you have already built them, would you mind sharing them.

Thanks,
Ramraj

[lintool]

Hi @cramraj8 - we just haven't had time to build them, but there's no reason why you couldn't, and I don't think it'd be too difficult. PRs welcome!

[cramraj8]

Got it. Thank you.

[cramraj8]

@lintool
I tried to encode BEIR & Mr.TyDi/ MIRACL datasets using Contriever or mContriever models for indexing. But pyserini.encode throws the following error
__main__.py encoder: error: argument --encoder-class: invalid choice: 'contriever' (choose from 'dpr', 'bpr', 'tct_colbert', 'ance', 'sentence-transformers', 'auto')
And I don't see the contriver script inside the encode folder as well. Are there any replacements for this or I should use Tevatron ?