I’m trying to fit a nonlinear model on not particularly many observations, and it occasionally crashes my computer forcing me to reset. Running on 1 core seems to always work, 2 cores sometimes crashes, and 3 cores crashes fairly regularly: it’s stochastic, which is most annoying of all. I would …

Also, the 10th gen Intel added some new multi core speed boost stuff. If you are doing any over clocking on the CPU or memory (or even if you aren’t), you may want to head into the BIOS and turn the CPU and memory speed down 10% or so to ensure stability.

One thing that might be worth trying is try running the model directly in cmdstan (in the command line) or using cmdstanpy like @ahartikainen mentioned. That would help narrow down where the problem lies (R or cmd(stan).

Thanks so much for the help. I’m pretty rusty on my python, but managed to get it running (and did it in Jupyter before I saw your edit with the code). I got everything working with the stan file and the json that I prepared for Rok yesterday. It went on for about 3/4 minutes before my system fro…

Same outcome with cmdstan. On the command line, I’m compiling, and then sampling using the following: for i in {1..6} do ./cmdstanr_test sample data file=cmdstanr_test_data- \ output file=output_${i}.csv & done I get the same bunch of warnings (like the following - but maybe 5…

@rok_cesnovar I bet this is something in Stan that is misbehaving. Do we have some example repo where we could inject couts to see where it gets stuck. edit. wait remove macOS, haha

great, did you use the same cmdstan installation or did you create a new one?

I wonder if we are doing something dumb in Stan when the user has a ton of non/infinite values? @rok_cesnovar @mathesong it may be helpful to run this with valgrind and heaptrack like valgrind --leak-check=full --show-leak-kinds=all ./cmdstanr_test sample data file=cmdstanr_test_data- output file…

The thing that is the most weird here is that we are not able to reproduce. I am not familiar with Pop!OS. Going to try to install it as a virtual machine if that would help.

[image] mathesong: Same installation Ok, then the error is probably in Stan part. Could you try to install a new cmdstan with CmdStanPy? python -m cmdstanpy.install_cmdstan And then run the model again. (You can remove the cmdstan from ~/.cmdstanpy folder after the test)

STAN on multiple cores occasionally crashing Linux without overwhelming memory

General

andrjohns September 3, 2020, 2:30pm 26

Do you get the same behaviour for models other than your own? A quick test would be to run the Bernoulli example provided with cmdstan

Topic		Replies	Views
Error in unserialize(socklist[[n]]) : error reading from connection Error in serialize(data, node$con, xdr = FALSE) : error writing to connection RStan	52	19696	December 14, 2021
Model makes R crash Modeling fitting-issues , performance	21	2268	September 24, 2017
Rstan error in RStudio - "Cannot allocate memory of 16921.6 GB" RStan	10	1821	November 10, 2019
R crashing with Rstan/brms RStan installation	13	1589	December 29, 2020
Problems with crashing R session Interfaces rstan , performance	65	9801	March 24, 2022

STAN on multiple cores occasionally crashing Linux without overwhelming memory

Related topics