Rerun model with different data


I’ve defined a model that fits my data fairly well:

def model(data):
   blah blah blah...
   trace = pm.sample()
   return trace

The data is an array of (~300,2) floats. I would like to run this model over and over with different data. However, there’s a lot of overhead that seems to happen before the actual sampling. Is there any way to cache that and just rerun?

Check out the Data Container object; this example might help Using shared variables (Data container adaptation) — PyMC3 3.11.4 documentation

1 Like