IBLBrainWideBenchTS3

class ts3.IBLBrainWideBenchTS3(root, regime, dirname='ibl_brain_wide_bench_2026', transform=None, **kwargs)

Bases: core.dataset.WholeSessionSpikeDataset

TS3’s view of the benchmark: whole sessions after neural QC.

Shares its units with unit-level pretraining: the units TS3 scores are the units those models saw. The class exists to state that: the QC policy is fixed rather than a parameter, so no consumer can evaluate on a population other than the suite’s, and a contract failure says which suite asked.

Parameters:

• root (str) – The root directory of the dataset.

• regime (Literal['pretrain', 'eval']) – The regime of the dataset (pretrain, eval).

• dirname (str) – The name of the dataset (and the directory containing its data).

• transform (Optional[Callable]) – The transform(s) to apply to the data.