In pystan 3, I see that the supported workflow is:
posterior = stan.build(schools_code, data=schools_data) fit = posterior.sample(num_chains=4, num_samples=1000)
whereas in the supported workflow in pystan 2 we could easily cache a compiled Stan model and reuse it with a different data set, or even pickle the compiled model and train it on a different machine (with the same installation).
Is there currently a work-around to enable compiled model cacheing? Or any plans to support the ‘legacy’ workflow in pystan 3?