ir_datasets
ir_datasets copied to clipboard
allenai
Add msmarco v2.1 trec rag
Dear all,
with a focus on the current TREC Rag task, this is an initial implementation for accessing documents via the corpus_iter and the document store for the document and the segmented document datasets of MS MARCO v2.1.
This is not completely finished (i.e., mainly documentation todos in the corresponding tickets are open), but as the deadline is soon, this might be useful for others even when the documentation is not yet completed.
Hey- can you try out the revision? I was getting errors running the tests before, so I refactored a bit. Now it's using the classes from v2 where possible.
Perfect, nice that we could re-use more code from v2 classes, thanks!
The changes look good, I started the tests, I think they will run for a while, and will report back when they still work. (but I think everything should work)
Hi, an update from the tests, they still run on my machine and also the other scripts that I had still work, so everything seems to be fine from my side.