Skip to content

V0.1.8

Latest

Choose a tag to compare

@EsdenRun EsdenRun released this 06 Oct 07:02
· 16 commits to main since this release
73159d7

Changelog

What’s New

Fixes

  • Resolved an issue where results could not be saved reliably.
    • Added robust save/load pathways:
      • Arrays (source_feat, target_feat, target_pred/_filtered) saved via np.savez_compressed.
      • Metadata maps (pairs, pairs_filter) saved as JSON.
    • (Helpers included) save_dirac_results() and load_dirac_results() for one-line persistence.

Additions

  • Function encapsulation / utilities
    • compute_markers_ctg(adata, label_key, n_genes=50): a thin, documented wrapper around sd.utils.ctg for differential gene discovery per label.

Improvements

  • Cleaner, more modular notebooks and RAW Markdown tutorials:
    • Clear section anchors, segmented steps, and copy-paste code cells.
    • Unified parameter notes and recommended defaults (neighbors, epochs, lamb, scale_loss).
  • Better out-of-the-box UX:
    • Fewer repeated code blocks (use helpers).
    • Safer defaults and clearer error messages.
    • Example layouts and download instructions for datasets.

Backward Compatibility

  • No breaking API changes. Existing notebooks should run unchanged.
  • New helpers are optional but recommended for clarity and reuse.

Acknowledgments

Thanks to users who reported the save-failure bug and requested cleaner utilities for DE and preprocessing. Your feedback directly shaped this update.