Hi everyone, I’m having an issue with getting a heirarchical model to work with pm.Data containers. I’m using data containers because in the model documentation ( https://docs.pymc.io/api/model.html#pymc3.model.set_data ) if you want to predict on new data you need to use pm.set_data() to switch the …

Hi, This is indeed the object of an ongoing PR . In the meantime, you can do pm.intX(pm.Data('group_identifier', group_identifier.values)). This will cast the shared variables as ints again, allowing you to use them as index variables. At first glance, I don’t see any obvious issue with the model (…

Data containers, heirarchical models and minibatches

Questions

willblenkhornjisc April 17, 2020, 9:33am 3

Hi Alex, I’ve just tested that and it works thanks. I managed to train the model, save it, reload and predict on new data with it. Thanks alot.

1 Like

Topic		Replies	Views
Running a model on multiple datasets error Questions	5	942	March 7, 2021
Problems moving from shared variable to data container Questions	3	555	April 13, 2023
Trouble using Data Containers with GPs Questions bug	2	601	February 17, 2021
Data container question Questions	1	428	February 26, 2020
Testing hiearchical model on unseen data Questions	3	821	January 4, 2018

Data containers, heirarchical models and minibatches

Related topics