Skip to content

Update scvi 1x#88

Merged
alitinet merged 5 commits into
mainfrom
update-scvi-1x
Jun 7, 2026
Merged

Update scvi 1x#88
alitinet merged 5 commits into
mainfrom
update-scvi-1x

Conversation

@alitinet

@alitinet alitinet commented Jun 7, 2026

Copy link
Copy Markdown
Collaborator

No description provided.

alitinet and others added 5 commits June 7, 2026 12:31
Replaces scvi-tools 0.20.3 with >=1.4 (Lightning 2.x), dropping use_gpu
in favour of accelerator/devices, adapting to the new _initialize_model /
_get_loaded_data / SaveCheckpoint / parse_device_args APIs. Moves
GroupDataSplitter to sample-level splitting via external_indexing so that
every validation bag is complete, and fixes a floating-point undercount
that produced zero-length validation sets. Also fixes a single-layer
regressor shape bug (class_input_dim vs z_dim), a .squeeze() dim for the
1-sample edge case, numpy-2 dtype identity check, and pandas-2 groupby
observed=True. Switches the package manager to uv (pyproject.toml +
uv.lock), bumps requires-python to >=3.12, and adds pytest smoke tests
(tests/test_mil.py) plus tutorial-parity scripts (tests/scripts/).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Replace mamba/conda env setup with uv, bump Python requirement to 3.12,
and update contributing guide to use uv sync / uv run pytest.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- CI and ReadTheDocs: bump Python 3.10 → 3.12/3.13 to match requires-python
- test.yaml: replace 3.10 matrix entry with 3.13; keep 3.12 + pre-release entries
- build.yaml + .readthedocs.yaml: 3.10 → 3.12, ubuntu-20.04 → 22.04
- CHANGELOG.md: add Unreleased section for scvi-tools 1.x migration; fix
  misformatted "## Fixed" → "### Fixed" in 0.3.2 section
- docs/api.md: add nn (MLP, Aggregator) and utils (get_sample_representations,
  score_top_cells, setup_ordinal_regression, plt_plot_losses) sections
- mil_classification.ipynb: fix pandas 2.x incompatibilities (.cat.unique() →
  .unique(), groupby observed=True) before re-running

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Classification: 96/91/93% accuracy across 3 splits (early stopping fired
at epochs ~60, ~60, ~55). Regression: R² of -0.26, -0.12, +0.41.

Also fixed two pandas 2.x incompatibilities in mil_classification.ipynb:
.cat.unique() → .unique(), groupby(observed=True).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
@review-notebook-app

Copy link
Copy Markdown

Check out this pull request on  ReviewNB

See visual diffs & provide feedback on Jupyter Notebooks.


Powered by ReviewNB

@alitinet alitinet merged commit 881f5c8 into main Jun 7, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant