Can't resolve numpy version conflicts in COlab

[alew46]

I'm trying to step through the install steps of the simple end-to-end demo. After I import Lightweight, jax and numpyro:

!pip install --upgrade git+https://github.com/google/lightweight_mmm.git

'import jax.numpy as jnp
import numpyro'

And then attempt to import the MMM elements:

from lightweight_mmm import lightweight_mmm
from lightweight_mmm import optimize_media
from lightweight_mmm import plot
from lightweight_mmm import preprocessing
from lightweight_mmm import utils

I get an error on the plot import that implicates a mismatched numpy version:

[/usr/local/lib/python3.11/dist-packages/matplotlib/transforms.py](https://localhost:8080/#) in <module>
     47 
     48 from matplotlib import _api
---> 49 from matplotlib._path import (
     50     affine_transform, count_bboxes_overlapping_bbox, update_path_extents)
     51 from .path import Path

ImportError: numpy.core.multiarray failed to import

After researching, I've tried following the advisories to upgrade numpy, downgrade matplotlib, etc. I can't find the blend that results in an error-free install.

Does anyone have this working in Colab currently?

[ikeda-AIPJ]

Set library version in google colab

!pip install numpy==1.25.2
!pip install jax==0.4.20
!pip install jaxlib==0.4.20
!pip install numpyro==0.12.1
!pip install scipy==1.10.1

After this, restart the session.

I was able to get this to work
But it's unclear whether it's optimal