ENH: add the lite_data dataset (JupyterLite split 1/5) - #14128
ENH: add the lite_data dataset (JupyterLite split 1/5)#14128natinew77-creator wants to merge 11 commits into
Conversation
A small curated archive holding only the files the browser documentation reads, taken from sample, kiloword, erp_core, mtrf and eegbci. Those ship as separate multi-GB archives, so without it the docs build downloads several gigabytes to serve a handful of files.
|
@teonbrooks PR 1 of the split is up. |
natinew77-creator
changed the title
teonbrooks
left a comment
teonbrooks
left a comment
There was a problem hiding this comment.
this looks good. i think we should just clarify the documentations and docstrings
|
|
||
| This exists for the documentation build; for analysis, use the individual | ||
| dataset fetchers above. | ||
|
|
There was a problem hiding this comment.
I would add a note that it currently does not support the somatosensory examples
There was a problem hiding this comment.
Done. Both the dataset docs and the docstring now say somato is not included, so those examples are not available in the browser.
| ================ | ||
| :func:`mne.datasets.lite_data.data_path` | ||
|
|
||
| A small curated archive holding only the files the browser documentation reads, |
There was a problem hiding this comment.
I would frame it as the files to run the examples and tutorials or something along that line that is more descriptive of the dataset use case
There was a problem hiding this comment.
Reworded to lead with the use case the files needed to run the tutorials and examples.
| The full MNE datasets (``sample``, ``kiloword``, ``erp_core``, ``mtrf``, | ||
| ``eegbci``) ship as separate multi-GB archives, so the docs build would download | ||
| several gigabytes just to serve a handful of files to the browser notebooks. | ||
| ``lite_data`` is a small curated archive holding only those files -- same data, |
There was a problem hiding this comment.
I would move this section first in the docstring description.
The explanation of why not the full MNE dataset could come later
There was a problem hiding this comment.
Moved. It now opens with what lite_data is, and the size explanation follows.
Lead with the data files needed to run the tutorials and examples, move the multi-GB rationale below it, and note that somato is not included.
2 participants
Closes part of #14041.
Adds
mne.datasets.lite_data, a small curated archive holding only the files the browser documentation reads, taken fromsample,kiloword,erp_core,mtrfandeegbci(same files, same checksums). It extracts toMNE-lite-data/with each file under its original dataset folder, so paths match the full datasets.Those ship as separate multi-GB archives, so without this the docs build downloads several gigabytes to serve a handful of files.
The approach was agreed in #14041, where @teonbrooks signed off on the fetcher and @larsoner kindly hosted the archive on OSF.
Nothing under
mne/uses it yet; the consumer is the JupyterLite docs build in #13925. Keeping it separate lets it land on its own.Follows the usual dataset layout and adds a metadata-only test that needs no network.