Utilities for PyStan for

• caching model compilation in a smart way
• caching sampling results in a smart way
• checking Stan warnings
• making corner plots of scalar parameters

$ pip install stan-utility

import stan_utility

model = stan_utility.compile_model('myscript.stan')
data = dict(mean=1)

results = stan_utility.sample_model(model, data, chains=2, iter=1000)
print(results.posterior)  ## a arviz.InferenceData object

# create mytest_fit_corner.pdf:
stan_utility.plot_corner(samples, outprefix="mytest_fit")

On the second run of this code,

• compile_model will retrieve the compiled model from cache
• sample_model will retrieve the results from cache. (change the seed or parameters if you want a fresh run).

See demos:

• https://github.com/JohannesBuchner/stan_utility/blob/master/examples/rosen2d.ipynb

• https://github.com/JohannesBuchner/stan_utility/blob/master/examples/rosenNd.ipynb

  import stan_utility

  model = stan_utility.compile_model_code(""" data { // maybe later } parameters { real<lower=0,upper=1> x; real y[10]; } model { y ~ normal(1, 2); } """) data = dict(something=1)

  results = stan_utility.sample_model(model, data) print(results.posterior)

  stan_utility.plot_corner(samples)

Editing the comments, adding and removing lines will not require recompilation of the model, because empty lines and comments are stripped out.

Derived originally from Stan_utility by Michael Betancourt and Sean Talts.

• @grburgess
• @cescalara
• @JohannesBuchner

Contributions are welcome.