Best practice and pythonic way to create an instance of a specific class based on a configuration file

I am writing, as an exercice to improve my coding habits, a package that will implement some neural network architectures using PyTorch and that will allow the user to train, test and exploit those architectures.

I have a exploit.py file that handles command line arguments provided by the user. Among the arguments, the user can provide a configuration file. I parse this configuration file using configparser. And in this configuration, the user can enter the network architecture that he desires to use.

What I would like to do is to create an instance of the right architecture depending on the user input. I have defined a module named architectures.py in which I define my architectures. Below is a pseudo-code example of this architecture.py file :

from torch import nn
class myArchitecture(nn.module):
    def __init__(self):
        super.__init__()

    def forward(self,x)
        pass

class myNextArchitecture(nn.module):
    def __init__(self):
        super.__init__()

    def forward(self,x)
        pass

so the user could put in the configuration file architecture1 or architecture2 (user-friendly name let's imagine) and I would need to handle the loading of the corresponding instance. Obviously, I could do something like this in the exploit.py file if I have already parsed the architecture in an arch variable:

import architecture
if arch == 'architecture1':
    net = architecture.myArchitecture()
elif arch == 'architecture2':
    net = architecture.myNextArchitecture()
else :
    print("Architecture not handled")
    # which kind of error should I raise here ?? Should I made a custom one ??

but I am wondering :

• is this a good practice to handle such cases like that ?
• How should I raise the exception in case the user inputs a wrong architecture name ? Should I use a built-in exception and if yes which one ? Or should I create my own ?

Or should I somehow write a more explicit method like architecture.load_model(arch) ? This method I imagine would be located in architecture.py and would do something like :

def load_model(name):
    if name == 'architecture1' :
        return myArchitecture() # I assume this would work since located in `architecture.py`