I am new to Pystan, and have been having incredible trouble getting simple models to sample.
I receive the error: ModuleNotFoundError: No module named ‘stanfit4anon_model…’
And then my Jupyter Notebook cell runs in perpetuity, until I restart the kernel.
Any advice helps! Thank you.
Hi, what OS do you have?
What does gcc --version say?
What Python and PyStan versions do you use?
Are you able to compile models?
Does sampling work with n_jobs=1
Thank you so much!
OS: Mac 10.15.4
gcc version:
Python version:
Is it conflicting that it says Clang 10.0.0 here? And Clang 12.0 in the gcc version?
Pystan version: 2.19.1.1
Able to compile models? - Yes!
Sampling with n_jobs = 1 – this worked! Thank you
What would be the best way to ensure I can run multiple chains at once?
Hi can you try this
import multiprocessing
multiprocessing.set_start_method("fork")
This worked!
Thank you.
This works! Thanks~
It worked!! Make sure to restart the notebook and then run:
import multiprocessing
multiprocessing.set_start_method(“fork”)
before running other commands.
Thank you!!!
I’m having the same problem as the OP, but my context is loading a pickled Stan model from S3 (that was fitted in another notebook). Restarting the notebook and importing multiprocessing does not fix the issue.
I’m using PyStan 2.19 in a Databricks notebook. I can fit the model just fine (using my_stan_model.sampling()) and pickle it to save in S3. But when I try to load the pickled model in a second notebook, I get the No module named 'stanfit4anon_model... error.
I’m able to load a pickled compiled model (from pystan.StanModel) just fine, so the problem seems to be with the fitted model in particular.
In PyStan 2 you also need to save (and then load first) the model object with your fit.
So next time save a list
[model, fit]
If you don’t have the model object saved anymore, you could try the following “hack”
https://pystan2.readthedocs.io/en/latest/unpickling_fit_without_model.html
Thanks, that solved it!