Error while running plot_media_channel.... function

[chsheth]

Hi, I am running "simple_end_to_end_demo" notebook on a GCP instance. All of the code runs fine except that I get the following error when running the code. Quick google search says numba needs to be turned off - I am not sure because all other plot functions are working fine. Any advice? Thanks.

plot.plot_media_channel_posteriors(media_mix_model=mmm)


---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
TypeError: expected dtype object, got 'numpy.dtype[float64]'

The above exception was the direct cause of the following exception:

SystemError                               Traceback (most recent call last)
<ipython-input-55-4b8328576e48> in <module>
----> 1 plot.plot_media_channel_posteriors(media_mix_model=mmm)

/opt/conda/lib/python3.7/site-packages/lightweight_mmm/plot.py in plot_media_channel_posteriors(media_mix_model, channel_names, quantiles, fig_size)
    708           media_channel_posteriors[:, channel_i],
    709           quantiles=quantiles,
--> 710           ax=channel_axis)
    711       axis_label = f"media channel {channel_names[channel_i]}"
    712       channel_axis.set_xlabel(axis_label)

/opt/conda/lib/python3.7/site-packages/arviz/plots/kdeplot.py in plot_kde(values, values2, cumulative, rug, label, bw, adaptive, circular, quantiles, rotated, contour, fill_last, figsize, textsize, plot_kwargs, fill_kwargs, rug_kwargs, contour_kwargs, contourf_kwargs, pcolormesh_kwargs, is_circular, ax, legend, backend, backend_kwargs, show, return_glyph, **kwargs)
    248                 bw = "experimental"
    249 
--> 250         grid, density = kde(values, circular, bw=bw, adaptive=adaptive, cumulative=cumulative)
    251         lower, upper = grid[0], grid[-1]
    252 

/opt/conda/lib/python3.7/site-packages/arviz/stats/density_utils.py in kde(x, circular, **kwargs)
    529         kde_fun = _kde_linear
    530 
--> 531     return kde_fun(x, **kwargs)
    532 
    533 

/opt/conda/lib/python3.7/site-packages/arviz/stats/density_utils.py in _kde_linear(x, bw, adaptive, extend, bound_correction, extend_fct, bw_fct, bw_return, custom_lims, cumulative, grid_len, **kwargs)
    625         x_min, x_max, x_std, extend_fct, grid_len, custom_lims, extend, bound_correction
    626     )
--> 627     grid_counts, _, grid_edges = histogram(x, grid_len, (grid_min, grid_max))
    628 
    629     # Bandwidth estimation

/opt/conda/lib/python3.7/site-packages/arviz/utils.py in __call__(self, *args, **kwargs)
    181         """Call the jitted function or normal, depending on flag."""
    182         if Numba.numba_flag:
--> 183             return self.numba_fn(*args, **kwargs)
    184         else:
    185             return self.function(*args, **kwargs)

SystemError: CPUDispatcher(<function histogram at 0x7f66dc4f9a70>) returned a result with an error set

[pabloduque0]

Hello @chsheth ! Looks like a problem with the arviz library. What version of arviz are you currently using and what python version?

[chsheth]

On the GCP instance, Python is 3.7.8 and arviz is 0.11.2.

[chsheth]

Hi @pabloduque0 -- not sure if you had any other suggestions but I was getting this error today on colab while running the end_to_end_demo_with_multiple notebook --

ImportError: cannot import name '_png' from 'matplotlib' (/usr/local/lib/python3.7/dist-packages/matplotlib/__init__.py)

Any suggestions? Thanks.

[pabloduque0]

Hello @chsheth !

I am unable to reproduce your error on a GCP instance with python 3.7 and the same arviz version.

Could you try pip install --upgrade git+https://github.com/google/lightweight_mmm.git ? Just in case some version got hanged or similar.

As per your other error, if its in colab could you please provide a reproducible colab for me to take a look?

Thanks!

[chsheth]

Thank you @pabloduque0.

For GCP/arviz, I can email you the notebook, cannot add it here -- I do think it has to do with some versioning.

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
<ipython-input-4-9a43dbab01ab> in <module>
      1 # Import the relevant modules of the library
----> 2 from lightweight_mmm import lightweight_mmm
      3 from lightweight_mmm import optimize_media
      4 from lightweight_mmm import plot
      5 from lightweight_mmm import preprocessing

/opt/conda/lib/python3.7/site-packages/lightweight_mmm/lightweight_mmm.py in <module>
     51 from lightweight_mmm import models
     52 from lightweight_mmm import preprocessing
---> 53 from lightweight_mmm import utils
     54 
     55 Prior = Union[

/opt/conda/lib/python3.7/site-packages/lightweight_mmm/utils.py in <module>
     22 import jax.numpy as jnp
     23 import numpy as np
---> 24 import pandas as pd
     25 from scipy import interpolate
     26 from scipy import optimize

/opt/conda/lib/python3.7/site-packages/pandas/__init__.py in <module>
    140 from pandas.util._print_versions import show_versions
    141 
--> 142 from pandas.io.api import (
    143     # excel
    144     ExcelFile,

/opt/conda/lib/python3.7/site-packages/pandas/io/api.py in <module>
      6 
      7 from pandas.io.clipboards import read_clipboard
----> 8 from pandas.io.excel import ExcelFile, ExcelWriter, read_excel
      9 from pandas.io.feather_format import read_feather
     10 from pandas.io.gbq import read_gbq

/opt/conda/lib/python3.7/site-packages/pandas/io/excel/__init__.py in <module>
----> 1 from pandas.io.excel._base import ExcelFile, ExcelWriter, read_excel
      2 from pandas.io.excel._odswriter import _ODSWriter
      3 from pandas.io.excel._openpyxl import _OpenpyxlWriter
      4 from pandas.io.excel._util import register_writer
      5 from pandas.io.excel._xlsxwriter import _XlsxWriter

/opt/conda/lib/python3.7/site-packages/pandas/io/excel/_base.py in <module>
     30     get_writer,
     31 )
---> 32 from pandas.io.parsers import TextParser
     33 
     34 _read_excel_doc = (

/opt/conda/lib/python3.7/site-packages/pandas/io/parsers/__init__.py in <module>
----> 1 from pandas.io.parsers.readers import (
      2     TextFileReader,
      3     TextParser,
      4     read_csv,
      5     read_fwf,

/opt/conda/lib/python3.7/site-packages/pandas/io/parsers/readers.py in <module>
     15 import pandas._libs.lib as lib
     16 from pandas._libs.parsers import STR_NA_VALUES
---> 17 from pandas._typing import (
     18     ArrayLike,
     19     DtypeArg,

ImportError: cannot import name 'DtypeArg' from 'pandas._typing' (/opt/conda/lib/python3.7/site-packages/pandas/_typing.py)

For '_png' from 'matplotlib' error, I found on stackoverflow that I would need to downgrade matplotlib version

!python -m pip uninstall matplotlib
!pip install matplotlib==3.1.3

I ran the above before any plot functions and there were no errors.

[pabloduque0]

Glad to hear the other error was solved.

Is the arviz error persisting or has that one been solved as well?

In the repro I dont need to see the data or your specific code, you can use mock data for providing the error as the error looks unrelated to your data. You can create a new notebook/colab.

[chsheth]

Actually I dont know about arviz error, if it got fixed -- the code didnt get to that cell and errored out much earlier. I am running the demo notebooks from here on my GCP instance, so there is nothing that I cannot share... I was thinking of creating an environment, use requirements from here and then run -- maybe that might make the version issue (if there is one) go away.

[pabloduque0]

If you can provide me the basic specifications of your instance I can try to re-create from my side.

[chsheth]

Here are the details. Thanks for your help @pabloduque0

[05jeet]

I am facing the same error while using this code on Google Colab: plot.plot_media_channel_posteriors(media_mix_model=mmm, channel_names=media_cols)

Getting the below error:

**AttributeError Traceback (most recent call last) in <cell line: 2>() 1 # posterior distributions of the media effects. ----> 2 plot.plot_media_channel_posteriors(media_mix_model=mmm, 3 channel_names=media_cols)

AttributeError: 'Figure' object has no attribute 'plot_media_channel_posteriors'**

I have already tried this as well but no resolve: !pip uninstall -y matplotlib !pip install matplotlib==3.1.3

This command: !pip install --upgrade git+https://github.com/google/lightweight_mmm.git is by default installing matplotlib 3.6.1 and giving below error:

\ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts. bigframes 1.0.0 requires matplotlib>=3.7.1, but you have matplotlib 3.6.1 which is incompatible. Successfully installed matplotlib-3.6.1 WARNING: The following packages were previously imported in this runtime: [matplotlib,mpl_toolkits] You must restart the runtime in order to use newly installed versions.

So again did this: !python -m pip uninstall matplotlib !pip install matplotlib==3.7.1 and again used this code: plot.plot_media_channel_posteriors(media_mix_model=mmm, channel_names=media_cols) and got the same error below:

**AttributeError Traceback (most recent call last) in <cell line: 2>() 1 # posterior distributions of the media effects. ----> 2 plot.plot_media_channel_posteriors(media_mix_model=mmm, 3 channel_names=media_cols)

AttributeError: 'Figure' object has no attribute 'plot_media_channel_posteriors'**

Only thing different I have done is I have used: from prophet import Prophet in order to decompose weekly sales with ds and holidys data and added trend, season, holiday under extra features in control variables. Rest I am following steps under: https://github.com/google/lightweight_mmm#introduction

Please let me know if anyone able to resolve this.