This repo is a simple example of converting a Brian script into pure Python using numba.
We start from the CUBA example included in Brian.
Run the code using Brian via Google Colab.
Next I manually converted this example to do exactly the same thing as Brian does, but in pure Python accelerated using the numba Python JIT compiler.
Run the code without Brian via Google Colab.
Or here's a version that uses vectorisation and NumPy instead of numba.
Run the code without Brian via Google Colab.
For this last one, you can copy/paste the content of the cell from the notebook into a new notebook using JupyterLite to run it locally in the browser rather than on Colab. This lets you run this example without a Google account.
I hope this might be somewhat helpful in understanding how Brian works behind the scenes, and in general what is involved in simulating a spiking neural network.