Closed joellama closed 2 years ago
This isn't an exoplanet or M1 problem really - it's a PyMC3/Theano issue. You should be able to fix it by pinning numpy<1.22
, which PyMC3 should do, but doesn't!
Thanks @dfm! The trick it seems on M1 is then you get float128
isn't defined, that is get-aroundable by pegging pymc3==3.8. Thanks again!
You shouldn't need to restrict to Python 3.8. This is now an issue with the conda versions of PyMC3 (only PyMC3 v3.9 is available). Instead of pinning the Python version, you can just run pip install -U pymc3
to get a version of PyMC3 that works with more recent Pythons.
Unfortunately I think that the numpy pin does remain necessary because Theano is no longer maintained.
Hi @dfm, I'm trying to install on an M1 Mac. I did this to create a new environment:
conda create -c conda-forge -n pymc3_env python=3.9 pymc3
Thenpip install exoplanet
When I then tryimport exoplanet as xo
I getAttributeError: module 'numpy.distutils.__config__' has no attribute 'blas_opt_info'
. Numpy is version '1.22.2' which seems to be a pymc3 requirement(?)Edit: I also tried
conda install -c conda-forge exoplanet
and got
<frozen importlib._bootstrap>:228: RuntimeWarning: scipy._lib.messagestream.MessageStream size changed, may indicate binary incompatibility. Expected 56 from C header, got 64 from PyObject
Any thoughts?