pnpl.datasets.hdf5.dataset.HDF5Dataset

pnpl.datasets.hdf5.dataset.HDF5Dataset#

class pnpl.datasets.hdf5.dataset.HDF5Dataset(data_path, preprocessing_name=None, tmin=-0.2, tmax=0.6, sfreq=None, subjects=[], sessions=[None], tasks=[None], runs=[None], include_info=False, standardize=True, clipping_factor=None, channel_means=None, channel_stds=None)[source]#

Torch Dataset for Serialized hdf5 files. Please implement get_phonemes method in the subclass.

Parameters:

  • data_path (str)

  • preprocessing_name (str | None)

  • tmin (float)

  • tmax (float)

  • sfreq (float | None)

  • subjects (list[str])

  • sessions (list[str | None])

  • tasks (list[str | None])

  • runs (list[str | None])

  • include_info (bool)

  • standardize (bool)

  • clipping_factor (float | None)

  • channel_means (ndarray | None)

  • channel_stds (ndarray | None)

__init__(data_path, preprocessing_name=None, tmin=-0.2, tmax=0.6, sfreq=None, subjects=[], sessions=[None], tasks=[None], runs=[None], include_info=False, standardize=True, clipping_factor=None, channel_means=None, channel_stds=None)[source]#

data_path: path to serialized dataset. include: Subjects to load (e.g. [‘010002’, ‘010047’]). If empty load all partition: train, val, or test

#included_files_dict: dictionary to indicate which files to load.

Keys are subjects and values are sessions. If empty, load all.

standardize: If True, standardize the data. Calculate means and stds if not provided.

Parameters:

  • data_path (str)

  • preprocessing_name (str | None)

  • tmin (float)

  • tmax (float)

  • sfreq (float | None)

  • subjects (list[str])

  • sessions (list[str | None])

  • tasks (list[str | None])

  • runs (list[str | None])

  • include_info (bool)

  • standardize (bool)

  • clipping_factor (float | None)

  • channel_means (ndarray | None)

  • channel_stds (ndarray | None)

Methods

__init__(data_path[, preprocessing_name, ...])

data_path: path to serialized dataset.

get_phonemes(subject, session, task, run)

Contents