Error when running klusta sorter in container mode

[chyumin]

I'm getting this error when trying to run klusta sorter in docker or singularity mode

The sorter runs well, this happens when SI tries to extract the container's output

TypeError                                 Traceback (most recent call last)
Input In [4], in <cell line: 1>()
----> 1 sorting = si.run_sorter(
      2     'klusta',
      3     recording,
      4     verbose=True,
      5     delete_output_folder=False,
      6     singularity_image=True
      7 )
      8 sorting

File ~/CatalystNeuro/spikeinterface/spikeinterface/sorters/runsorter.py:132, in run_sorter(sorter_name, recording, output_folder, remove_existing_folder, delete_output_folder, verbose, raise_error, docker_image, singularity_image, with_output, **sorter_params)
    130         else:
    131             container_image = singularity_image
--> 132     return run_sorter_container(
    133         container_image=container_image,
    134         mode=mode,
    135         **common_kwargs,
    136     )
    138 return run_sorter_local(**common_kwargs)

File ~/CatalystNeuro/spikeinterface/spikeinterface/sorters/runsorter.py:565, in run_sorter_container(sorter_name, recording, mode, container_image, output_folder, remove_existing_folder, delete_output_folder, verbose, raise_error, with_output, extra_requirements, **sorter_params)
    563 else:
    564     if with_output:
--> 565         sorting = SorterClass.get_result_from_folder(output_folder)
    567 if delete_output_folder:
    568     shutil.rmtree(output_folder)

File ~/CatalystNeuro/spikeinterface/spikeinterface/sorters/basesorter.py:281, in BaseSorter.get_result_from_folder(cls, output_folder)
    277 if bool(log['error']):
    278     raise SpikeSortingError(
    279         "Spike sorting failed. You can inspect the runtime trace in spikeinterface_log.json")
--> 281 sorting = cls._get_result_from_folder(output_folder)
    283 recording = load_extractor(output_folder / 'spikeinterface_recording.json')
    284 if recording is not None:
    285     # can be None when not dumpable

File ~/CatalystNeuro/spikeinterface/spikeinterface/sorters/klusta/klusta.py:155, in KlustaSorter._get_result_from_folder(cls, output_folder)
    153 @classmethod
    154 def _get_result_from_folder(cls, output_folder):
--> 155     sorting = KlustaSortingExtractor(file_or_folder_path=Path(output_folder) / 'recording.kwik')
    156     return sorting

File ~/CatalystNeuro/spikeinterface/spikeinterface/extractors/klustaextractors.py:82, in KlustaSortingExtractor.__init__(self, file_or_folder_path, exclude_cluster_groups)
     79         cs_to_exclude.append(ec.lower())
     81 for channel_group in kf_reader.get('/channel_groups'):
---> 82     chan_cluster_id_arr = kf_reader.get(f'/channel_groups/{channel_group}/spikes/clusters/main')[()]
     83     chan_cluster_times_arr = kf_reader.get(f'/channel_groups/{channel_group}/spikes/time_samples')[()]
     84     chan_cluster_ids = np.unique(chan_cluster_id_arr)  # if clusters were merged in gui,

TypeError: 'NoneType' object is not subscriptable

[alejoe91]

@chyumin does this happen for all recordings? Maybe something just went wrong!

[chyumin]

@alejoe91 This happenned on toy data when running tests from spikeinterface-dockerfiles, but I also tested to run separately on a notebook and I got the same error. With real data is working well

[alejoe91]

@chyumin can you try to install the sorter locally and see if the same error appears? How long is the test recording?