A few weird errors

Questions v5
,
1

I have encountered a few annoying errors which made me restart my model fitting over and over again. Just wondering if anyone has seen similar issues.

  1. When I’m fitting models on Jupiter notebook (through vs code on mac), no matter how many cores I use, it runs fine. But when I use the mac terminal to fit models, if I specify cores > 1, I get BrokenPipeError: [Errno 32] Broken pipe.
  2. When I’m fitting models on Jupiter notebook (through vs code on mac), sometimes my kernel just crashes by itself when the fitting is mid-way through.

Canceled future for execute_request message before replies were done

The Kernel crashed while executing code in the the current cell or a previous cell. Please review the code in the cell(s) to identify a possible cause of the failure. Click here for more info. View Jupyter log for further details.

2
  1. Are you certain that you are running in the same environment/versions/etc.? That sounds like 2 different setups to me.
  2. The error message is not informative because you don’t (necessarily) get full traceback messages when running inside a notebook (and VS Code may hide even more, I’m not sure). I suggest you pull the relevant code into a script of just dump it into a terminal to get a better idea of what is going on.
3

Yes I used anaconda to create a virtual environment for this project and checked that pymc version is 4.1.3 before running. For the first error, it’s an error on the terminal actually. It wasn’t an error on the notebook. I saw another post where Junpeng simply recommended using only one core when running on the terminal. But that was a couple of years ago so not sure if this issue was fixed. Regarding my second issue, I think it was replicated on the terminal as well, where I got even fewer error messages, simply “terminal killed” for no reason… Maybe there is a time limit set by either pymc or python on how long can a program stay running?

4

I just meant that the environment you are using in the terminal might be different than the environment you are using from inside VS Code/Jupyter.

Definitely not from PyMC. Are you possibly reaching the limit of your machine’s memory?

5

running out of memory might be a possible cause! I have been puzzling over it during multi-core fitting as well. My computer auto-restarts because memory runs out. It’s weird because if I use prior Normal(0, 1), my memory is enough at least when I set core = 2, but if I use Normal(0, 10), everything else held constant, the memory is all of sudden not enough.