Can't log mmm model with pymc_marketing.mlflow.log_mmm

Questions v5
,
1

pymc version: 5.27.1

pymc marketing version: 0.18.2

I am getting the following error on pymc_marketing.mlflow.log_mmm

TypeError: log_model() got an unexpected keyword argument 'name'

 # Log model configuration (saturation, adstock, priors, etc.)
pymc_marketing.mlflow.log_mmm_configuration(mmm_all)

# Log model derived info - pass the PyMC model
pymc_marketing.mlflow.log_model_derived_info(mmm_all.model)

# Log sample diagnostics - pass the idata
pymc_marketing.mlflow.log_sample_diagnostics(mmm_all.idata, tune=FINAL_TUNE)

mlflow.log_params({"train_test_split_date": TRAIN_TEST_SPLIT_DATE})

# Log the full MMM model as deployable artifact
pymc_marketing.mlflow.log_mmm(
    mmm=mmm_all,
    include_last_observations=True,
    original_scale=True
)
2

Hi, thanks for pointing this out. I believe you can fix it by updating your mlflow to 3.0.1 better. What version are you on, mlflow.__version__?

We had some protection against this sort of bug but I think it might have been misconfigured. I’ve opened a PR to avoid this sort of thing bug : mlflow uses the name param on versions 3.0.0 and greater by daniel-saunders-phil · Pull Request #2339 · pymc-labs/pymc-marketing · GitHub. If you cannot update mlflow for whatever reason, you might watch that PR for updates.

3

PR’s been merged by the way, so it should disappear on the next release.

4

I’m running 2.22.0