huggingface / huggingface/datasets

Request: Review pending neuroimaging PRs (#7886 BIDS loader, #7887 lazy loading)

Open
#7,904 1 comment 1 reaction 0 assignees View on GitHub
Dominant language
Python
Stars
22k
Forks
3.4k
Avg merge
5d 7h
Merged PRs (30d)
17

Description

## Summary

I'm building production neuroimaging pipelines that depend on `datasets` and would benefit greatly from two pending PRs being reviewed/merged.

## Pending PRs

| PR | Description | Status | Open Since |
|----|-------------|--------|------------|
| [#7886](https://github.com/huggingface/datasets/pull/7886) | BIDS dataset loader | Open | Nov 29 |
| [#7887](https://github.com/huggingface/datasets/pull/7887) | Lazy loading for NIfTI | Open | Nov 29 |

## Use Case

The neuroimaging community uses the BIDS (Brain Imaging Data Structure) standard for organizing MRI/fMRI data. These PRs would enable:

1. **#7886**: `load_dataset('bids', data_dir='/path/to/bids')` - Load local BIDS directories directly
2. **#7887**: Memory-efficient NIfTI handling (single 4D fMRI file can be 1-2GB)

## Current Workaround

Without these, users must either:
- Upload to Hub first, then consume (works but slow iteration)
- Hand-roll BIDS parsing (duplicates effort)

## Request

Could a maintainer review these PRs? Happy to address any feedback. The BIDS loader has tests passing and was end-to-end tested with real OpenNeuro data.

Thank you for the great work on `Nifti()` support - these PRs build on that foundation.

## Related

- Contributes to #7804 (Support scientific data formats)
- Built on @TobiasPitters's Nifti feature work

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.