-
Notifications
You must be signed in to change notification settings - Fork 96
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
BUG: Compiler error when calling pm.sample() #952
Comments
] |
We'll need more info, how did you install PyMC and can you provide a minimal reproducible example? |
Hi, I encountered a similar problem. I deleted the previous Python 3.12/PyMC 5.16.2 installation from MacOS 14 Sonoma and reinstalled it, and the problem happened. (I tried in the base environment and also in a different environment, following the PyMC instructions). The previous installation from MacOS 14 worked on MacOS 15. This issue might be related to clang and the latest developer beta (3) of MacOS 15 Sequoia, on a Macbook Air M2
This is the code that I'm running:
and if I sample (pm.sample) with numpyro, blackjax or even the 'pymc' option
and more similar code. Additionally, when installing PyMC, the BLAS library that gets installed is OpenBLAS, not the accelerated version, which needs to be installed manually, and the usual warning "WARNING (pytensor.tensor.blas): Using NumPy C-API based implementation for BLAS functions." |
@notluquis how are you install pymc? Are you using conda? |
I’m installing pymc with the installation instructions on the website. Yes, I’m using conda, and I also tried with mamba/miniforge and pip. |
Going to move this issue to PyTensor. In the meantime, if you want to completely bypass the C backend you can set |
That works thank you! Could this momentary patch cause slower sampling or other secondary problems? I hope this bug doesn't go unnoticed, maybe it's a problem with the new macOS beta and clang |
If you use the default PyMC samplers it will be considerably slower. It may also slow down prior/posterior predictive to some extent but those are usually not the bottleneck compared to posterior sampling |
If I set |
Describe the issue:
Reproduceable code example:
Error message:
No response
PyMC version information:
pymc 5.12.1
Context for the issue:
No response
The text was updated successfully, but these errors were encountered: