release(0.2.0): scmidas.integrate() + bundled quickstart + Colab notebook#32
Merged
Conversation
…book User-facing entry point overhaul focused on first-impression UX. The previous Quick start in the README was an API sketch with `...` placeholders that nobody could paste-run; new users had to read the full demo2 tutorial before their first MIDAS run. This release bundles a 2.66 MB toy MuData (1600 cells, RNA+ADT mosaic, hand-tuned subset of wnn_mosaic_8batch_mtx) inside the wheel, plus a top-level `scmidas.integrate(mdata)` wrapper with hyperparameters (batch_size=128, max_epochs=65, lr=3e-4) tuned to converge on that toy dataset in about a minute on a mid-range GPU. The README quickstart is now five runnable lines plus a rendered UMAP image; an `Open in Colab` badge points at a pre-executed `examples/quickstart.ipynb` so users can try MIDAS without installing anything locally. The new defaults are loud about their toy-tuned scope — both the docstring `.. warning::` block and the README explicitly say to override `max_epochs` (1000-2000) and `batch_size` (256) for real data. Also folds in the unreleased library-logging refactor (5 modules drop `logging.basicConfig` for module-level `getLogger(__name__)`) that has been on `main` since 4e75e4e.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
User-facing UX overhaul: a 2.66 MB toy MuData ships inside the wheel, a top-level
scmidas.integrate(mdata)wrapper trains it in ~1 min on a mid-range GPU, and the README has a runnable five-line quickstart with rendered UMAP image. AnOpen in Colabbadge points at a pre-executedexamples/quickstart.ipynbso prospective users can try MIDAS without installing.Closes the
...placeholder problem in the previous Quick start.Changes
New
scmidas.integrate(mdata)— toy-tuned one-line wrapper aroundMIDAS.configure_data_from_mdata+train(). Defaults (batch_size=128, max_epochs=65, lr=3e-4) are loud about being toy-only; docstring + README warn to overridemax_epochs(1000-2000) andbatch_size(256) for real data.scmidas.datasets.quickstart()— load the bundled 1600-cell PBMC RNA+ADT mosaic MuData. 4 batches with full mosaic structure (1 RNA-only + 1 ADT-only + 2 paired), 500 HVGs + 224 ADT features.src/scmidas/data/quickstart_pbmc_mosaic.h5mu— bundled wheel data (2.66 MB).examples/quickstart.ipynb— pre-rendered notebook (five cells, UMAP output saved). Linked from README viaOpen in Colabbadge.scripts/build_quickstart_demo.py— reproducible build script for the bundled dataset.Changed
...placeholders with the runnable five-line snippet, embeds the rendered UMAP, and adds an explicit toy-data warning.Open in Colab.pyproject.toml:scmidaspackage-data now includesdata/*.h5mu. Bumps version to0.2.0..gitignore: whitelistsdocs/source/_static/img/*.pngso the new UMAP screenshot can be tracked despite the global*.pngrule.Folded in (was unreleased on main)
config,data,model,nn,utils) droplogging.basicConfig(level=INFO)and adopt module-levellogger = logging.getLogger(__name__). Demo notebooks callbasicConfigthemselves so user-visible output is unchanged.Test plan
pytest tests/37/37 pass locallyscmidas.integrate(scmidas.datasets.quickstart())end-to-end: 60.8s on V100, UMAP shows clean lineage separation + batch mixingexamples/quickstart.ipynbexecutes viajupyter nbconvert --executewith the same cell graph