Windows error in neural_simulations.py > load_mechs_from_folder()

[p2jones]

The function load_mechs_from_folder() in brainsignals/neural_simulations.py has not been tested on Windows (according to a to-do comment). I tested it and ran into a couple simple problems.

To Reproduce
On Windows 11, with all requirements installed, run the first cell of Figure-7.11.ipynb

Errors

1. In load_mechs_from_folder, warn is not imported. I resolved this by importing warn and reinstalling brainsignals

2. In load_mechs_from_folder, mod_pth is not defined. In other functions, there are different definitions, so I'm not sure which one to use. I possibly solved this by mod_pth = join(hay_folder, "mod/")
   As the warning suggests, I ran mknrndll from NEURON bash in the folder L5bPCmodelsEH

3. Next error occured at cell = return_hay_cell(tstop=tstop, dt=dt, make_passive=False)
   "ValueError: setting an array element with a sequence."

Copilot suggests that LFPy isn't yet compatible with python 3.14, so I should python 3.11.

Setup

• OS: Windows 11
• Python 3.14
• Neuron 9.0

[torbjone]

Thank you for notifying us. I will look into it. I don't think copilot is correct in this case, but I will let you know once I have tested this on Windows myself.

[torbjone]

A workaround is to use the hay model implementation from this example:
https://github.com/LFPy/LFPy/blob/master/examples/LFPy-example-04.ipynb

does this run? After you have made the cell model as in this example, the rest of the code should probably work.

[p2jones]

I tried LFPy-example-04.ipnyb and got the same error. Copied it below. Still with python 3.14 - I haven't tried an earlier version yet.

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
TypeError: only 0-dimensional arrays can be converted to Python scalars

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

ValueError                                Traceback (most recent call last)
Cell In[5], line 5
      1 # delete old sections from NEURON namespace
      2 LFPy.cell.neuron.h("forall delete_section()")
      3 
      4 # Initialize cell instance, using the LFPy.Cell class
----> 5 cell = LFPy.TemplateCell(**cellParameters)
      6 cell.set_rotation(x=4.729, y=-3.166)
      7 
      8 # Override passive reversal potential, AP is generated

File [~\AppData\Roaming\Python\Python314\site-packages\LFPy\templatecell.py:150](http://localhost:8888/lab/tree/notebooks/~/AppData/Roaming/Python/Python314/site-packages/LFPy/templatecell.py#line=149), in TemplateCell.__init__(self, templatefile, templatename, templateargs, verbose, **kwargs)
    147             neuron.h.load_file(template)
    149 #initialize the cell object
--> 150 Cell.__init__(self, **kwargs)

File [~\AppData\Roaming\Python\Python314\site-packages\LFPy\cell.py:286](http://localhost:8888/lab/tree/notebooks/~/AppData/Roaming/Python/Python314/site-packages/LFPy/cell.py#line=285), in Cell.__init__(self, morphology, v_init, Ra, cm, passive, passive_parameters, extracellular, tstart, tstop, dt, nsegs_method, lambda_f, d_lambda, max_nsegs_length, delete_sections, custom_code, custom_fun, custom_fun_args, pt3d, celsius, verbose, **kwargs)
    284 #Gather geometry, set position and rotation of morphology
    285 if self.pt3d:
--> 286     self._update_pt3d()
    287 else: # self._update_pt3d itself makes a call to self._collect_geometry()
    288     self._collect_geometry()

File [~\AppData\Roaming\Python\Python314\site-packages\LFPy\cell.py:1894](http://localhost:8888/lab/tree/notebooks/~/AppData/Roaming/Python/Python314/site-packages/LFPy/cell.py#line=1893), in Cell._update_pt3d(self)
   1892     neuron.h.define_shape()
   1893 #must recollect the geometry, otherwise we get roundoff errors!
-> 1894 self._collect_geometry()

File [~\AppData\Roaming\Python\Python314\site-packages\LFPy\cell.py:705](http://localhost:8888/lab/tree/notebooks/~/AppData/Roaming/Python/Python314/site-packages/LFPy/cell.py#line=704), in Cell._collect_geometry(self)
    703 elif self.somaidx.size == 1:
    704     self.somapos = np.zeros(3)
--> 705     self.somapos[0] = self.xmid[self.somaidx]
    706     self.somapos[1] = self.ymid[self.somaidx]
    707     self.somapos[2] = self.zmid[self.somaidx]

ValueError: setting an array element with a sequence.

[espenhgn]

Hi @p2jones,
If you're familiar with it, Docker may be a way to set up a virtual Linux environment with all dependencies in place on your machine. These codes, including LFPy, were not much developed/tested on Windows, mainly Linux/MacOS.
I did provide a Dockerfile to build such an environment locally here: https://github.com/LFPy/ElectricBrainSignals/blob/main/Dockerfile, but some of the version pins may need updating. I did not push the build anywhere, unfortunately.
Alternatively, if you have access to a Linux-based cloud compute environment with Jupyterlab or similar, you can try and set up a virtual environment there (e.g. via conda), using versions matching what we used. I think NEURON must be version 8.0.2, and Python 3.10-3.13.

We need to do some updates to "officially" support Python 3.14 and Windows, cf. LFPy/LFPy#476

[espenhgn]

Hi again. Did some more testing including changes introducing more rigorous testing using Python 3.14 and the latest NEURON 9.0.1 here and everything appears to work with LFPy itself.

However, it looks like it is Numpy>=2.x that introduces the TypeError seen here, as it no longer allows implicit conversion of 1-element arrays into Python scalars (try: int(np.array([5]))).

The quick fix may be to uninstall numpy and reinstall an older version your environment (e.g., pip uninstall numpy; pip install numpy==1.26.4), until we can patch things here (in addition to fix other deprecations that may have occurred).

As for "official" Windows support for LFPy, I don't think we will do this until NEURON is pip installable on Windows (neuronsimulator/nrn#3431). As of now, only Linux/Mac wheels are available (https://pypi.org/project/neuron/#files).