NEURON subprocessing

Run multiple NEURON setups isolated from each other from a single Python script. This package uses Python's subprocess to run multiple NEURON instances that are completely seperated from eachother, making it easier to executed repeated and parametrized simulations without having to worry about cleaning up the state of the previous run.

Installation

pip install nrn-subprocess

Usage

Write your entire NEURON setup that you'd like to isolate inside of a function, then use the subprocess or isolate approach to execute it in isolation:

subprocess

import nrnsub

def my_sim(param1, opt1=None):
  from neuron import h
  s = h.Section(name="main")
  # ...
  return s.v

for i in range(10):
  nrnsub.subprocess(my_sim, 15, opt1=i)

This will run the subprocesses in series, parallel coming Soon (tm).

isolate

There's also the isolate decorator that will make sure every call to that function is ran as an isolated subprocess:

import nrnsub

@nrnsub.isolate
def my_sim(param1, opt1=None):
  from neuron import h
  s = h.Section(name="main")
  # ...
  return s.v

for i in range(10):
  my_sim(15, opt1=i)

Worker PATH

The worker might have trouble unpacking the serialized objects because it can't find the module they came from on the main process. This might result in dill raising "module not found" errors. To fix them you can pass the _worker_path to subprocess or worker_path to the isolate decorator:

import nrnsub

sys.path.insert(0, "/home/me/my_modules_folder")

@nrnsub.isolate(worker_path=["/home/me/my_modules_folder"])
def f():
  import something_in_my_modules_folder