# Cmdstanpy WinError 5 Access is denied

**URL:** <https://discourse.mc-stan.org/t/cmdstanpy-winerror-5-access-is-denied/39165>\
**Category:** Interfaces\
**Created:** [March 26, 2025, 8:59pm UTC](https://discourse.mc-stan.org/t/cmdstanpy-winerror-5-access-is-denied/39165 "2025-03-26T20:59:20Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![Nick\_Henscheid](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.mc-stan.org/nick_henscheid/32/20248_2.png) [@Nick\_Henscheid](https://discourse.mc-stan.org/u/Nick_Henscheid)\
**Post date:** [March 26, 2025, 8:59pm UTC](https://discourse.mc-stan.org/t/cmdstanpy-winerror-5-access-is-denied/39165/1 "2025-03-26T20:59:20Z")

</div>

I’ve been having a frustrating intermittent problem with cmdstanpy where I frequently get a “[WinError 5] Access is denied” when trying to call methods on my compiled model object. It works sometimes, usually when I have freshly opened a notebook (Jupyter or VSCode), but if I re-compile after making changes to my model script, I get the error essentially every time.

Example:

```no-highlight
model = stan.CmdStanModel(stan_file='model.stan')
MAP_fit = model.optimize(...) 

```

I’ve tried just about every combination of running as administrator, moving my files to a “safe” location like c:\Users\myusername, etc. I can’t seem to find a pattern. Is cmdstanpy putting a lock on the executable and then forgetting to release it? If so why would it be so intermittent?

My IT admin says he doesn’t see any activity by our security programs (that was my first thought, that our antimalware was killing the .exe).

Thanks ahead for any ideas!

Operating System: Win11 Enterprise  
Interface Version: cmdstanpy 2.36  
Compiler/Toolkit: RTools40

Full error dump:

13:44:35 - cmdstanpy - INFO - Chain [1] start processing

* * *

PermissionError Traceback (most recent call last)  
File ~\AppData\Local\anaconda3\envs\LSTM\lib\site-packages\cmdstanpy\model.py:2061, in CmdStanModel.\_run\_cmdstan(self, runset, idx, show\_progress, show\_console, progress\_hook, timeout)  
2060 fd\_out = open(runset.stdout\_files[idx], ‘w’)  
 → 2061 proc = subprocess.Popen(  
2062 cmd,  
2063 bufsize=1,  
2064 stdin=subprocess.DEVNULL,  
2065 stdout=subprocess.PIPE,  
2066 stderr=subprocess.STDOUT, # avoid buffer overflow  
2067 env=os.environ,  
2068 universal\_newlines=True,  
2069 )  
2070 timer: Optional[threading.Timer]

File ~\AppData\Local\anaconda3\envs\LSTM\lib\subprocess.py:947, in Popen. **init** (self, args, bufsize, executable, stdin, stdout, stderr, preexec\_fn, close\_fds, shell, cwd, env, universal\_newlines, startupinfo, creationflags, restore\_signals, start\_new\_session, pass\_fds, user, group, extra\_groups, encoding, errors, text, umask)  
944 self.stderr = io.TextIOWrapper(self.stderr,  
945 encoding=encoding, errors=errors)  
 → 947 self.\_execute\_child(args, executable, preexec\_fn, close\_fds,  
948 pass\_fds, cwd, env,  
949 startupinfo, creationflags, shell,  
950 p2cread, p2cwrite,  
951 c2pread, c2pwrite,  
952 errread, errwrite,  
953 restore\_signals,  
954 gid, gids, uid, umask,  
955 start\_new\_session)  
956 except:  
957 # Cleanup if the child failed starting.

File ~\AppData\Local\anaconda3\envs\LSTM\lib\subprocess.py:1416, in Popen.\_execute\_child(self, args, executable, preexec\_fn, close\_fds, pass\_fds, cwd, env, startupinfo, creationflags, shell, p2cread, p2cwrite, c2pread, c2pwrite, errread, errwrite, unused\_restore\_signals, unused\_gid, unused\_gids, unused\_uid, unused\_umask, unused\_start\_new\_session)  
1415 try:  
 → 1416 hp, ht, pid, tid = \_winapi.CreateProcess(executable, args,  
1417 # no special security  
1418 None, None,  
1419 int(not close\_fds),  
1420 creationflags,  
1421 env,  
1422 cwd,  
1423 startupinfo)  
1424 finally:  
1425 # Child is launched. Close the parent’s copy of those pipe  
1426 # handles that only the child should have open. You need  
(…)  
1429 # pipe will not close when the child process exits and the  
1430 # ReadFile will hang.

PermissionError: [WinError 5] Access is denied

The above exception was the direct cause of the following exception:

RuntimeError Traceback (most recent call last)  
Cell In[21], line 19  
2 init\_values = {  
3 ‘theta\_S0’: 0.5,  
4 ‘theta\_r’: 0.3,  
(…)  
15 ‘omega\_subj\_r’: 0.3  
16 }  
18 # Run optimization  
—\> 19 MAP\_fit = conrado\_model.optimize(  
20 data=stan\_data\_train,  
21 inits=init\_values,  
22 algorithm = “lbfgs”,  
23 iter=10000,  
24 tol\_grad=1e-12,  
25 tol\_param=1e-12,  
26 tol\_obj=1e-12,  
27 init\_alpha=1e-3,  
28 refresh=50,  
29 show\_console=True, # equivalent to verbose=True  
30 output\_dir = “C:/Users/XXXX/stan\_models/tmp\_output”  
31 )

File ~\AppData\Local\anaconda3\envs\LSTM\lib\site-packages\cmdstanpy\model.py:644, in CmdStanModel.optimize(self, data, seed, inits, output\_dir, sig\_figs, save\_profile, algorithm, init\_alpha, tol\_obj, tol\_rel\_obj, tol\_grad, tol\_rel\_grad, tol\_param, history\_size, iter, save\_iterations, require\_converged, show\_console, refresh, time\_fmt, timeout, jacobian)  
642 dummy\_chain\_id = 0  
643 runset = RunSet(args=args, chains=1, time\_fmt=time\_fmt)  
 → 644 self.\_run\_cmdstan(  
645 runset,  
646 dummy\_chain\_id,  
647 show\_console=show\_console,  
648 timeout=timeout,  
649 )  
650 runset.raise\_for\_timeouts()  
652 if not runset.\_check\_retcodes():

File ~\AppData\Local\anaconda3\envs\LSTM\lib\site-packages\cmdstanpy\model.py:2110, in CmdStanModel.\_run\_cmdstan(self, runset, idx, show\_progress, show\_console, progress\_hook, timeout)  
2108 except OSError as e:  
2109 msg = ‘Failed with error {}\n’.format(str(e))  
 → 2110 raise RuntimeError(msg) from e  
2111 finally:  
2112 fd\_out.close()

RuntimeError: Failed with error [WinError 5] Access is denied

---

<div class="post-metadata">

**Author:** ![WardBrian](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.mc-stan.org/wardbrian/32/12078_2.png) [@WardBrian](https://discourse.mc-stan.org/u/WardBrian)\
**Post date:** [March 27, 2025, 1:43pm UTC](https://discourse.mc-stan.org/t/cmdstanpy-winerror-5-access-is-denied/39165/2 "2025-03-27T13:43:07Z")

</div>

> [@Nick\_Henscheid](#):
>
> Is cmdstanpy putting a lock on the executable and then forgetting to release it? If so why would it be so intermittent?

We don’t do this manually, though it’s possible that win32 does when running. Usually running multiple processes from the same exe is no issue, though.

Can you describe a bit more about your set-up? Are you running this on a local folder or a network drive? Can you get the issue to happen when running non-interactively (e.g., from the command line)?

---

<div class="post-metadata">

**Author:** ![Nick\_Henscheid](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.mc-stan.org/nick_henscheid/32/20248_2.png) [@Nick\_Henscheid](https://discourse.mc-stan.org/u/Nick_Henscheid)\
**Post date:** [March 28, 2025, 4:17pm UTC](https://discourse.mc-stan.org/t/cmdstanpy-winerror-5-access-is-denied/39165/3 "2025-03-28T16:17:41Z")

</div>

I ended up switching my whole setup to WSL to avoid having to deal with this, but for reference here is what my setup looked like:

OS: Win 11 Enterprise 24H2 Build 26100.3476  
Files located under c:/Users/username (to avoid any OneDrive issues or any group permissions settings - my IT Admin suggested this)  
Python/conda: Conda env with python 3.9.0 (can provide more details if needed)  
IDE: VSCode 1.98.2 but issue persisted with Jupyter

I did not explore running via command line because I switched to WSL, so can’t comment on that.

Apparently some colleagues were having a similar issue with NONMEM (.exes getting locked) so it’s entirely possible this is an issue with our IT group policies or some other obscure Windows issue that is not specific to Stan or cmdstanpy. We have all switched to WSL and this has fixed our problems, so not too bothered by it.