Roberto

I still get OOM from this line (`vexp = 2 * Xg @ Xc.T - (Xc**2).sum(1)`) even if clear_cache=True.

Is the problem arising specifically at `vexp = 2 * Xg @ Xc.T - (Xc**2).sum(1)`? If you used my pull request at which line does it happen?

Interesting, might be related to this SpikeInterface/spikeinterface#3332 . I also noticed that running kilosort in a loop sometimes causes weird behaviors.

Just tested it and it works. Amazing! Thanks for fixing it so quickly.

Maybe this is useful, I have a little func to extract infos from the bonsai file, for NPX1 it is useful to get the LFP and AP gain without hardcoding...

They look all like this:

this is the code I use, I don´t explicitly set sparsity: ``` sortings = [] for probe_n in tqdm(range(len(raw_rec.get_probes()))): sortings.append(read_phy(Path(f"{path_recording_folder}/spike_interface_output/probe{probe_n}/sorter_output"))) sorting = aggregate_units(sortings) analyzer = create_sorting_analyzer(sorting, raw_rec, sparse=True, format="binary_folder", folder=f"{path_recording_folder}/spike_interface_output/sorting_analyzer_post_curation",...

I sort with this: ``` path_recording, rec_file_name = find_rec_file(path_recording_folder) timestamps = get_timestamps_from_rec(path_recording_folder) raw_rec = read_spikegadgets(path_recording) raw_rec.set_times(timestamps) # this is useful for other parts of the code split_preprocessed_recording = raw_rec.split_by("group") for...

Actually there is some distance on the x axis (0 column in df) and the x coords are rightly assigned between groups. Channels belonging to group 0 have the lower...

Regarding sparsity: