Sebastiano Vigna

I'm sorry for the very late answer. This somehow slipped through. Yes, the RecSplit size should be counted in for coherence with the other data structures. The `bitCount()` you propose,...

I see the milestone have been pushed to 3.1. This does not seem such a difficult PR—one year no hold? 😂 Here are some other papers computing the harmonic diameter...

Well, there's a lot of literature talking about the "effective diameter" which is the 90th percentile of distances, even if it not a diameter. And as remarked above in functional...

In my case, however, there's no difference between 7B and the other models. The bias is still there.

Fantastic. Artificial intelligence baffled by a trailing space 😂.

I'm not against the idea: I'm using it now and it works very well. I'm just saying this should go into thevspecification so there's commitment to the syntax. Any suggestions...

Oops. We discovered the Lender crate which is actually much more developed than ours. So I changed the PR into adding Lender support.

Well, if you consider having all Iterator functions (map, etc.) a little gain I don't think I can convince you otherwise. Lance is a bit whimsical in his documentation, but...

No, you can't because of the orphan rule.

This PR makes tests fail (didn't you check?). Apparently there is a change of behavior of the list used to check the behavior of big lists. I don't have time...