Leland McInnes
Leland McInnes
I think there is some potential benefit to providing binary wheels. A catch is that compilation specific to the processors of the host machine can be pretty beneficial, and binary...
I'll try to look into the AOT a little more and see if I can have umap-learn-aot and pynndescent-aot packages available on PyPI; hopefully that would resolve some of the...
No further news I am afraid. I'm really dependent on numba's AOT compilation to make this possible. I understand there is work being done on this, but it is not...
These operations are in numba compiled functions, so loops are not actually slow, and can often be faster than numpy. Can you provide benchmarks demonstrating that this is actually any...
I believe you need python 3.8 or newer and then importlib.metadata is in the stdlib.
I understand that import time can be annoying. On the other hand I'm not convinced that punting it to first use necessarily improves the experience. Long term I would like...
You do make a decent case. On the other hand it will result in a significantly slower first run of UMAP because it will be deferring the import to that...
That is a little baffling to me to be honest. To just crash like that I think it needs to be related to something inside numba compiled functions interacting strangely....
I think there are some documentation build errors going on. I'll see if I can get that sorted out. Thanks for the heads up.
I'm afraid that looks to be down in numba bytecode, which is beyond the realm of things I can help with.