Exporting Data

SpikeLab provides export functions for saving SpikeData objects to several common electrophysiology formats. All exporters assume spike times are in milliseconds and convert to the target format’s native time unit automatically.

The export functions live in spikelab.data_loaders.data_exporters.

To HDF5

The HDF5 exporter supports the same four storage styles as the loader:

Ragged (default) – a flat spike times array plus an index array.
Raster – a (N, T) spike count matrix.
Group – one HDF5 group per unit.
Paired – two parallel arrays of unit indices and spike times.

from spikelab.data_loaders.data_exporters import export_spikedata_to_hdf5

# Default ragged style (spike_times + spike_times_index)
export_spikedata_to_hdf5(sd, "output.h5")

# Raster style
export_spikedata_to_hdf5(
    sd, "output.h5",
    style="raster",
    raster_dataset="raster",
    raster_bin_size_ms=1.0,
)

# Group-per-unit style
export_spikedata_to_hdf5(
    sd, "output.h5",
    style="group",
    group_per_unit="units",
    group_time_unit="s",
)

# Paired style
export_spikedata_to_hdf5(
    sd, "output.h5",
    style="paired",
    idces_dataset="idces",
    times_dataset="times",
    times_unit="ms",
)

Ragged vs raster style

The ragged style is space-efficient for sparse data: it stores one concatenated array of spike times and a separate index array that marks unit boundaries. This is the same layout used by NWB files.

The raster style stores a full (N, T) matrix of spike counts. This is convenient for downstream tools that expect a matrix, but can be large for long recordings with fine time bins. When using raster style, you must specify raster_bin_size_ms.

The `spike_times_unit` parameter

For ragged style, the spike_times_unit parameter (default 's') controls the unit of the exported spike times. Set to 'ms' to write times in milliseconds instead of seconds. Analogous parameters exist for the other styles (group_time_unit, times_unit).

Exporting raw data

If the SpikeData object has raw voltage traces attached (e.g. from from_thresholding()), you can include them in the HDF5 file:

export_spikedata_to_hdf5(
    sd, "output.h5",
    raw_dataset="raw_data",
    raw_time_dataset="raw_time",
    raw_time_unit="ms",
)

To NWB

Export to Neurodata Without Borders format. Spike times are saved in seconds. Unit IDs, electrode assignments, and electrode locations are written when available in neuron_attributes.

from spikelab.data_loaders.data_exporters import export_spikedata_to_nwb

export_spikedata_to_nwb(sd, "session.nwb")

# With custom dataset names
export_spikedata_to_nwb(
    sd, "session.nwb",
    spike_times_dataset="spike_times",
    spike_times_index_dataset="spike_times_index",
    group="units",
)

To KiloSort

Export to the KiloSort/Phy folder format. This writes spike_times.npy and spike_clusters.npy files, and optionally channel_map.npy if electrode information is available.

The fs_Hz parameter is required to convert spike times from milliseconds to the target time unit (samples by default).

from spikelab.data_loaders.data_exporters import export_spikedata_to_kilosort

spike_times_path, spike_clusters_path = export_spikedata_to_kilosort(
    sd,
    "output_folder/",
    fs_Hz=30000,
)

# With custom options
export_spikedata_to_kilosort(
    sd,
    "output_folder/",
    fs_Hz=30000,
    time_unit="samples",
    spike_times_file="spike_times.npy",
    spike_clusters_file="spike_clusters.npy",
    cluster_ids=None,
)

To pickle

Save a SpikeData object as a pickle file. Supports optional upload to Amazon S3.

from spikelab.data_loaders.data_exporters import export_to_pickle

# Local file
path = export_to_pickle(sd, "my_data.pkl")

# Upload to S3
s3_url = export_to_pickle(
    sd,
    "s3://my-bucket/my_data.pkl",
    s3_upload=True,
    aws_access_key_id="...",
    aws_secret_access_key="...",
)

The function returns the file path (local) or S3 URL as a string.