AllenNeuralDynamics / aind-ephys-pipeline-kilosort25

Code Ocean pipeline for ephys processing with Kilosort2.5
MIT License
2 stars 2 forks source link

device names in NWB broken #21

Open bjhardcastle opened 4 hours ago

bjhardcastle commented 4 hours ago

@alejoe91

Up until this week, the units table included device names like this:

>>> s.nwb_zarr["units/device_name"][:]
array(['ProbeA', 'ProbeA', 'ProbeA', ..., 'ProbeF', 'ProbeF', 'ProbeF'],
      dtype=object)

Now they've changed to this:

>>> s.nwb_zarr["units/device_name"][:]
array(['Neuropixels 1.0', 'Neuropixels 1.0', 'Neuropixels 1.0', ...,
       'Neuropixels 1.0', 'Neuropixels 1.0', 'Neuropixels 1.0'],
      dtype=object)

...and the names in the devices table look like this:

>>> tuple(s.nwb_zarr["general/devices"].keys()) 
('Neuropixels 1.0', 'Neuropixels 1.0-1', 'Neuropixels 1.0-2', 'Neuropixels 1.0-3', 'Neuropixels 1.0-4', 'Neuropixels 1.0-5')

This broke some downstream processing for us (https://github.com/AllenInstitute/npc_sessions/issues/127) and makes it really difficult to link units to probes. Could you please either fix the names of the devices or, if this was necessary, consider adding an electrode_group_name column to the units table.

Could you also please help us re-run the nwb-generation step of the pipeline for affected sessions?

alejoe91 commented 3 hours ago

Sure! It was definitely an unintended change! I'll fix it on Monday