Version checking error in custom preprocessor

[oaaij-gnahz]

Hi,

I am trying to implement a custom preprocessor class for a project and encountered error with version checking.

Description:
I follow the style of common_reference and implemented classese that inherit BasePreprocessor and BasePreprocessorSegment classes and the get_traces() method. I did not put the custom preprocessor module in my installed SpikeInterface folder, but instead just in another helper file of my project folder. Then when I run the code, it gives me the error with version checking. I guess it is because SpikeInterface assumes that the preprocessor module is in SpikeInterface and automatically checks the version of its parent module, which in my case does not have a version. My question is, is there some way to use SpikeInterface base classes to implement custom preprocessors, without having to register them into my actual SpikeInterface folder and potentially risking messing up my installation?

To give more context, here is the error message I got:

File "/home/sdfafsd/miniconda3/envs/si_env/lib/python3.10/site-packages/spikeinterface/core/base.py", line 1070
, in _check_same_version
    current_version = parse(imported_module.__version__)
AttributeError: module 'utils' has no attribute '__version__'

Where utils is the folder I put my custom preprocessor in.
I am using version 0.99.1

Appreciate any comments! I searched for "custom" in the discussion panel and did not see relevant questions. Still, apologies if it has already been answered somewhere.

[samuelgarcia]

Hi,
If you design your own preprocessing class, then the class need to live somewhere in an installed python module.
Otherwise it can not be loaded back when json or pickle.
In short if this class is in a local .py file this will not work unless all script are in the same module.
You can easly make a simple installable python module and use pip install -e /path/to/cistul/module instead, in this case it should solve the issue.
Good luck.
And of course feel free to propose your class in spikeinterface!

[oaaij-gnahz]

Thanks for the quick reply. I will try to make it an installable module.

What I am currently doing is adding __vesrion__ = spikeinterface.__version__ in my local .py file. So far it is not giving me error but I don't know if I will shoot myself in the foot later on. Would you please comment on that?

Appreciate your patience!

[alejoe91]

I think that is probably ok :)

[oaaij-gnahz]

Thanks!