What is the best way to deal with pymc3 in MLFLOW models in databricks?

[Siebert_Looije]

Last week, we started with using mlflow within databricks. The bayesian models that we are using right now are the pymc3 models (https://docs.pymc.io/en/v3/index.html).

We could use the experiment feature of databricks/mlflow to save the models as an artifact and then load them upon predicting.

However it would also be good to use the models feature of databricks/mlflow. We could not find a way (and it seems not to be supported right now) to use this feature for pymc3. Anyone has an idea how we could still use this?

Thanks!

[Anonymous]

Hi @Siebert Looije​ 

Great to meet you, and thanks for your question! 

Let's see if your peers in the community have an answer to your question first. Or else bricksters will get back to you soon.

Thanks.