Leland McInnes
Leland McInnes
So I'm actually working on a (generalized) LargeVis implementation myself, more to explore some other fun things I believe you can do (LargeVis on arbitrary dataframes rather than just vector...
There is a new release of hdbscan on PyPI that will hopefully fix this now.
Is this something that could be worked around via cffi and GMP? Is there significant round trip overhead on going out and back via cffi within numba?
If I get some time I will definitely look into it. Thank you.
Not out of the box no; I think you might need to do some filtering as preprocessing, or possibly come up with a custom distance metric that might do the...
This seems to potentially relate to some changes in Cython and or numpy; it is very hard to know or track down exactly where in the hdbscan code this is...
I think the catch may be having a cluster of size 1; that would definitely break something.
There was a recent sklearn release that changed some internals the hdbscan relied on (which resulted in the 0.8.30 release to attempt to fix those). It's possible that this is...
I'm at a bit of a loss; especially if 0.8.29 is also not building anymore. I can at least reproduce this locally, but it is unclear how to fix things...
Okay, I poked the obvious things in terms of module name resolution issues and it seems to have fixed the problem locally. I don't understand what changed, or, indeed, why...