A modified Stan model estimates only after a laptop reboot

OK - fixed the data issue and it is apparent I can compile and run the programs in bash. And there is no problem if I recompile (using make --always-make) and re-run. (I do run into problems with my working folder that has a β€œ-” in the name - but if I move everything to a test folder, everything works fine).

So, my problem seems to be an R issue, but let me know if there are non-R tests I should check.

1 Like

This seems to be a M1 problem, I have the same problem with a M1 macbook air.

I am sorry to hear that you are having the same problem, but maybe that will help us figure out a solution - given that it is not necessarily due to my own ineptitude. The frustrating thing is that I’ve been working with the M1 for a few months now, and this was not a problem early on - it is a recent development. So, I guess cmdstanr must have changed something.

If you think that cmdstanr changed something you can always revert to an earlier version of cmdstanr. But I also have access to an M1 and I just grabbed the development version of cmdstanr and compiled and re-compiled a model with no issues. I can’t imagine how this would be an issue, but FWIW here’s what I have in make/local:

CXX=arch -arch arm64e clang++
CXXFLAGS+= -O3 -mcpu=apple-a14

Thanks for the suggestion - I did change the make/local file and unfortunately still getting the same results. I still need to delete the old compiled file in order for it to recompile Not great, but it works. Thanks for trying.