Pierre Yger
Pierre Yger
Can you show print(recording) and print(probe), just to be sure data/probe are correctly loaded?
Yes, indeed, in spyking circus 2, there used to be a hardcoded limit for the find_peaks procedure (if not using wobble). Are you using the latest version from main ?...
Thanks for your feedback, I'll have a look quickly since indeed we would like everything to properly scale up with large number of units. Just to generate dummy data, remind...
I don't think you need to share anything. Based on your comments, there seems to be an issue in compute_similarity_with_templates_array because of internal data structure that should be too large....
If we suppose that the problem is in the creation of the mask in compute_similarity_with_templates_array , you can try to see if #4152 is solving the issue. This should decrease...
Thanks I'll keep digging then!
I guess the error comes from the fact that you should have templates_from_svd=True as a default argument in run_sorter(). The bug is thus slightly different, and I'll fix it also....
The behaviour templates_from_svd=False should be fixed now in this branch also. But as said, I think you should turn this option to True both for speed and memory considerations
I'll make a separate PR, thanks for spotting all these issues triggered when scaling up the number of templates
The memory footprint for overlaps is also fixed in #4157