Recieving ValueError when running extract_synthesis_inputs.py

[riddick-the-furyan]

When running the .py file I get "ValueError: zero-size array to reduction operation maximum which has no identity." The error seems to be in line 61 of simple_diametrizer.py indicated by lengths /= lengths.max(). The source code of extract_synthesis_inputs.py is syntactically different to what I am used to in python so I am not throwing out the probability that the error is with a change I made. Though I assume the slash is the error but I made a change to the source code of extract_synthesis_inputs.py. Which was only adding the directory to the reconstructions of neurons on my pc.

Here is the source code for extract_synthesis_inputs.py:

import json
from pathlib import Path

import neurots
from neurots import extract_input

def run(output_dir, data_dir):
    """Run the example for extracting inputs for synthesis."""
    # Generate distribution from directory of neurons
    distr = extract_input.distributions(
        data_dir / "neurons", feature="path_distances", diameter_model="default"
    )

if __name__ == "__main__":
    result_dir = Path("results_extract_synthesis_inputs")
    result_dir.mkdir(parents=True, exist_ok=True)

    run(result_dir, Path("C:\\Users\\user\\Documents\\Experiment_01\\Data\\Human\\Neocortex_interneurons"))
    print("DONE!")

Package versions neurots==3.3.1, tmd==2.3.0, diameter-synthesis==0.5.3 and morphio==3.3.4. None of these packages have been edited in anyway.

[adrien-berchet]

Hello again @riddick-the-furyan I think that you have to provide a value to the neurite_types parameter. The default value for this parameter is ["basal_dendrite", "apical_dendrite", "axon"], but if none of your morphologies has axon for example, then you should provide neurite_types=["basal_dendrite", "apical_dendrite"] to the neurots.extract_input.distributions function. Nevertheless I think we should support this case, WDYT @arnaudon ?

[arnaudon]

Yes, I got this error once as well but I didn't make it crash more gracefully. I can send a PR for this later today if you want.

[riddick-the-furyan]

@adrien-berchet when you say morphologies are you still referring to the variable neurite_types or the morphologies parameter when talking about if it has axon??

[adrien-berchet]

"morphologies" refers to the data in the files contained in the data_dir / "neurons" (or neurons if you prefer)

[riddick-the-furyan]

I see. I misread what you meant.

[riddick-the-furyan]

By any chance do you have an idea of the timeframe in which the next micro version of NeuroTS can support morphologies without axons? Sorry in advance if this sounds forward.

[arnaudon]

Hello, it is already, in the function distribution, there is a neurite_types argument which you didn't use in your script above. Just give it the neurite types you want to synthesize. I suppose for you it would be distr = extract_input.distributions(..., neurite_types=["basal_dendrite", "apical_dendrite"]), or without apical_dendrites if you don't have any.

[riddick-the-furyan]

Hi, thank you so much! The extract_synthesis_inputs.py ran with no errors and the parameters, distributions and morphologies were created so I am really grateful for your help. Though after the program terminated, I got this message "Selected step size 1.000000 is too big for bars of size 0.681029" in red in jupyter notebook. Should I be worried about it?

[arnaudon]

Great! For this warning, it should not be an issue, but maybe @lidakanari could tell you more on this one. It is related to a bar being smaller than the distance between two points. I see this sometimes, and the created morphologies are ok.

[adrien-berchet]

Hi, thank you so much! The extract_synthesis_inputs.py ran with no errors and the parameters, distributions and morphologies were created so I am really grateful for your help. Though after the program terminated, I got this message "Selected step size 1.000000 is too big for bars of size 0.681029" in red in jupyter notebook. Should I be worried about it?

Glad to know it worked! I close this issue as it seems solved but don't hesitate to reopen it if you need more help on this topic (or open a new one if it's about about something else).