C compiler error while running python module 'pymc'

Paulami

Hi,
I am using pymc module to develop a MMM model in Dataiku. While working on it, I am experiencing this somewhat well-known error on C compiler but couldn't find any solution on Dataiku yet.
Code where the error occured:
import warnings
import arviz as az
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import pymc as pm
import seaborn as sns
import dataiku
from pymc_marketing.mmm import MMM, GeometricAdstock, LogisticSaturation,MichaelisMentenSaturation
from pymc_marketing.prior import Prior
mmm = MMM()
mmm.fit(X, y, nuts_sampler="numpyro", **sampler_kwargs)
Error:
error: ld returned 1 exit su can find the C code in this temporary file: /tmp/pytensor_compilation_error_eusqphwd

Traceback (most recent call last)<ipython-input-107-73aef3a45723> in <cell line: 1>()----> 1 idata = mmm.fit(X, y, nuts_sampler="numpyro", **sampler_kwargs)/datadir/dataiku/DATA_ATHENA/code-envs/python/MMM_TEST/lib/python3.10/site-packages/pymc_marketing/model_builder.py in fit(self, X, y, progressbar, predictor_names, random_seed, **kwargs) 533 534 if not hasattr(self, "model"):--> 535 self.build_model(self.X, self.y) 536 537 sampler_config = self.sampler_config.copy()/datadir/dataiku/DATA_ATHENA/code-envs/python/MMM_TEST/lib/python3.10/site-packages/pymc_marketing/mmm/delayed_saturated_mmm.py in build_model(self, X, y, **kwargs) 482 channel_contributions = pm.Deterministic( 483 name="channel_contributions",--> 484 var=self.forward_pass(x=channel_data_), 485 dims=("date", "channel"), 486 )/datadir/dataiku/DATA_ATHENA/code-envs/python/MMM_TEST/lib/python3.10/site-packages/pymc_marketing/mmm/delayed_saturated_mmm.py in forward_pass(self, x) 342 ) 343 --> 344 return second.apply(x=first.apply(x=x, dims="channel"), dims="channel") 345 346 def build_model(/datadir/dataiku/DATA_ATHENA/code-envs/python/MMM_TEST/lib/python3.10/site-packages/pymc_marketing/mmm/components/base.py in apply(self, x, dims) 511 """ 512 kwargs = self._create_distributions(dims=dims)--> 513 return self.function(x, **kwargs)/datadir/dataiku/DATA_ATHENA/code-envs/python/MMM_TEST/lib/python3.10/site-packages/pymc_marketing/mmm/components/adstock.py in function(self, x, alpha) 188 189 def function(self, x, alpha):--> 190 return geometric_adstock( 191 x, alpha=alpha, l_max=self.l_max, normalize=self.normalize, mode=self.mode 192 )/datadir/dataiku/DATA_ATHENA/code-envs/python/MMM_TEST/lib/python3.10/site-packages/pymc_marketing/mmm/transformers.py in geometric_adstock(x, alpha, l_max, normalize, axis, mode) 233 w = pt.power(pt.as_tensor(alpha)[..., None], pt.arange(l_max, dtype=x.dtype)) 234 w = w / pt.sum(w, axis=-1, keepdims=True) if normalize else w--> 235 return batched_convolution(x, w, axis=axis, mode=mode) 236 237 /datadir/dataiku/DATA_ATHENA/code-envs/python/MMM_TEST/lib/python3.10/site-packages/pymc_marketing/mmm/transformers.py in batched_convolution(x, w, axis, mode) 123 124 if l_max is None: # pragma: no cover--> 125 raise NotImplementedError( 126 "At the moment, convolving with weight arrays that don't have a concrete shape " 127 "at compile time is not supported."NotImplementedError: At the moment, convolving with weight arrays that don't have a concrete shape at compile time is not supported.

Version used for relevant packages:
pymc==5.15.1
arviz==0.18.0
matplotlib==3.6.2
numpy==1.23.5
pandas==1.5.2
jax==0.4.30
numpyro==0.15.1
ipywidgets==7.6.5
pymc-marketing==0.8.0
theano-pymc==1.1.2
pytensor<2.23, >=2.22.1
This is a similar discussion I can find on Google on the exact same error.
Pytensor compilation error - Questions / version agnostic - PyMC Discourse
Also Note: The same code with same package versions is running fine in my local Anaconda.

Paulami

Hi, Any update on this? I am still not able to fix this one.😐️

Turribeach

This is not a Dataiku issue but a pymc issue. You should probably raise it in the pymc forums or in the Github issues. The fact that it works in your local Anaconda doesn't mean it will work in Dataiku.