How to stop pystan from caching the models/results

Interfaces PyStan
1

I’m running my script on a HPC, and I only have a certain amount of disk space available to me. When I sample from my posterior using PyStan it seems like Stan is caching the results/model, and the resulting file is over 50Gb somehow. This causes the program to be killed by the HPC.

Is there a way to stop stan from caching the samples/models?

I’m using the most up to date version of PyStan and httpstan, working on Ubuntu 20.

2

If there is temporary storage space on your server, you can make pystan/httpstan save the results there. Set XDG_CACHE_HOME to, say, /tmp/httpstan-cache and things will be written there. The default value for XDG_CACHE_HOME varies by platform. On linux it’s ~/.cache.

(I haven’t tested this, but it should work.)