Compiling for 64 bits windows 7 about matlabstan HOT 8 CLOSED

When you get this error, is there a *.stan file in your working directory? Does your working directory have any spaces in the name?

from matlabstan.

Hmm no to both questions. After running a model, a *.stan file was created in the pwd though (anon_model.stan). Subsequent runs gave a pop-up asking for a new file name.

I'm testing this on two different computers, both Windows 7, 64 bits (two different work places). The second computer gives the following error instead:

Error using StanModel/stan_version
Output argument "ver" (and maybe others) not assigned during call to
"C:\Users\Max\Documents\MATLAB\Packages\MatlabStan-    2.7.0.0\StanModel.m>StanModel.stan_version".

Not sure if it's related of course, but the hardware and OS of both systems is identical.

from matlabstan.

Perhaps related (although it wouldn't appear so) is the warning that precedes this error:

Warning: # of chains > # of cores. 
> In StanModel>StanModel.set.chains at 470
  In StanModel>StanModel.StanModel at 224
  In stan at 108 

Even though I'm only running 1 chain.

from matlabstan.

I'm still confused. Would you be so kind as to try running the model outside of Matlab at the command line? The .stan file and data are here:

https://github.com/stan-dev/example-models/tree/master/misc/eight_schools

Thanks.

from matlabstan.

That works like a charm!

from matlabstan.

So I'm basically getting different errors on both systems (they are far apart so difficult to try both at the same time). System one gives the 'no compiled in support x86_64' error, system two gives the problem with getting the Stan version.

I'm currently on system two, where I've now tried:

• changing the cmdstan installation to a path without spaces: no changes
• I can get the stanc version via cygwin, which I used to compile cmdstan with. However, when I use 'stanc --version' in the windows command utility, I get the error 'application unable to start correctly'. I assume this is what MatlabStan uses under the hood, and is where the problem arises.

from matlabstan.

Thanks for the update. For system two, you are correct, I'm calling down to stanc to get the version. I would have assumed that this would prevent you from running the eight_schools or bernoulli examples from the command line as well?

from matlabstan.

I'm sorry for the confusion: I have tried the eight-schools and Bernoulli example using cygwin. Here, the cmdstan tool works as intended (and also gives me the correct version). However, Matlab presumably uses the windows Cmd tool (or something similar), in which this goes awry.

I have now found a dirty hack for system two, which is to simply put
ver = '2.7.0';
in the version() function in StanModel.m. Apparently, stanc performs fine otherwise, because now I can run everything from within Matlab.

I'll return to system one on Monday. I'm suspecting I have some libraries messed up.

Thanks a lot for your support by the way - the tool is great when it works! :-)

from matlabstan.