UnitQCPolicy

class core.dataset.UnitQCPolicy(keep_qc_neural=('PASS', 'WARNING', 'FAIL'), keep_qc_neural_alignment=('PASS', 'WARNING', 'FAIL'), min_firing_rate=None, min_label=None)

Bases: object

Which sessions and units a whole-session dataset keeps.

Distinct from UnitFiltering, which is what the build already did: this is the cut the consumer makes on top of it. Every dataset still names one, in its own directory, so a model that deviates does it on purpose.

Every field asks whether a unit’s data is usable, never whether it has a usable target, which is the task’s business. Every field defaults to cutting nothing, so a policy states only what it removes and UnitQCPolicy() is the raw population.

keep_mask(units)

The units this policy keeps, over one recording’s unit table.

Return type:

ndarray