PyStan 3.0 Release Candidate 1

PyStan 3.0 Release Candidate 1 is now available on PyPI. Please test this release.

New in PyStan 3.0 Release Candidate 1:

New in PyStan 3:

  • Automatic caching of compiled Stan models
  • Automatic caching of samples from Stan models
  • Open source software: ISC License
  • CVODES support out-of-the-box.
  • Timely releases. PyStan 3 releases appear within 48 hours of Stan releases with probability greater than ⅔.

Documentation: https://pystan.readthedocs.io/
Requirements: Python 3.7 or higher; macOS or Linux.
Installation instructions: python3 -m pip install --pre pystan
Release date for PyStan 3.0 Final: 25. March

Upgrading to PyStan 3 from PyStan 2:

Known issues:

  • Only the recommended sampler (No-U-Turn Sampler) is currently supported.
  • Recent Linux version required. Linux users with older operating systems may have problems using wheels from PyPI. Ubuntu 20.04 and Debian Bullseye should work. Users with an older version of GCC or Clang need to install httpstan from source before installing pystan.
  • Using PyStan 3 in a Jupyter Notebook requires an extra configuration step. See the FAQ.
  • Windows is not supported at this time.
3 Likes

Hey,

I’m not sure whether this is the appropriate place to ask, but I remember using PyStan “back in the days” and some kind of summary of the samples being available?

Something like this, from Google’s cache?

http://webcache.googleusercontent.com/search?q=cache:RF70Wdd5pssJ:pystan.readthedocs.io/en/latest/_modules/pystan/misc.html+&cd=1&hl=en&ct=clnk&gl=de

Is there a way to get something similar? The page does no longer exist for the latest version:

https://pystan.readthedocs.io/en/latest/_modules/pystan/misc.html

The old documentation is here: https://pystan2.readthedocs.io/

As for a stansummary for PyStan 3, there are no plans to write such a function for PyStan. Adding stansummary to stan::services in Stan C++ and then exposing it in PyStan would be fine. I’m not sure there’s anyone who has expressed interest in doing this.

2 Likes

this is available in CmdStanPy - it wraps CmdStan’s bin/stansummary and returns it as a pd.DataFrame - API Reference — CmdStanPy 0.9.68 documentation

2 Likes

Thank you both!

To get ‘summary’ there is possibility to use

arviz.summary(fit)

3 Likes