New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix for NumPy 1.26 #7101
Fix for NumPy 1.26 #7101
Conversation
Is it possible that @seberg, I hope it's okay to ping you in the hope that you have some insight or know whom to ask about this. 🙏 E.g. might numpy/numpy@7508afd be responsible? |
I'll try my luck with |
An even more minimal reproducing example: import numpy as np
from concurrent.futures import ThreadPoolExecutor
assert np.__version__ == '1.26.0b1'
rng = np.random.default_rng(32)
matrices = (
rng.random((5, 10, 10, 3, 3)),
rng.random((5, 10, 10, 3, 3)),
# rng.random((5, 10, 10, 3, 3)),
)
with ThreadPoolExecutor(max_workers=None) as ex:
list(ex.map(lambda m: np.linalg.eigvalsh(m), matrices,)) on the pre-build NumPy 1.26.0b1, I get one of these errors:
When building NumPy 1.26.0b1 from source with Maybe something todo with CBLAS etc.? Not sure how to read the output of |
Reported upstream in numpy/numpy#24512. |
Just noticed, the failing test in eb97c5e is a different one
compared to the one mentioned in #6970 (comment). But it's the same function |
Sorry, just on the process of coming back from vacation. I am very sure that the lapack lite stuff is not threadsafe, but if you install wheels they shouldn't be using it, the Next, the warning state may be unsafe, but this doesn't seem the problem (fixed on NumPy |
Ah, hand't seen the NumPy issue where some nice sleuthing found that it is indeed the lapack-lite being used (due to build problems) which were fixed. |
eb97c5e
to
0da3f5b
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
🎉
Follow up to #6970.
See