Fix bugs, preserve .var, bump deps, migrate to uv, add smoke tests#46
Merged
Conversation
Fixes MultiVAE(losses=None) crash (bugs in both _multivae.py and _multivae_torch.py), load_query_data AttributeError on theta when integrate_on is set without nb/zinb loss, and silent zero-batch training in StratifiedSampler on small datasets. Preserves .var metadata across modalities in organize_multimodal_anndatas, suffixing overlapping column names with modality index (issue #4). Adds defensive cont_covs.get() guard (issue #45). Bumps to scvi-tools>=1.4.3 (Python >=3.12), removes numpy<2 and jax<0.6 caps, migrates from hatch to uv with PEP 735 dependency-groups, rewrites CI and RTD configs accordingly. Adds minimal smoke test suite (4 tests, all passing), updates CHANGELOG, api.md, contributing.md, .gitignore, and re-executes both tutorial notebooks. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
…main - pyproject.toml: Merged dependency updates with optional-dependencies and dependency-groups - _preprocessing.py: Integrated .var preservation (issue #4) with improved validation checks - _multivae.py: Fixed condition formatting for losses check - _multivae_torch.py: Kept ParameterList approach for supporting multiple NB/ZINB losses - test_basic.py: Kept comprehensive test suite from HEAD - CHANGELOG.md: Already properly resolved with all fixes documented
Fixed Issues: 1. test_warns_and_makes_unique_if_var_names_duplicate_across_modalities: Updated warning regex to match both old and new AnnData warning messages 2. test_end_to_end_default_losses: Fixed theta parameter handling by: - Changed test assertion to check for empty ParameterList instead of None - Updated _multivae.py to check if theta params have numel() > 0 instead of None - Added defensive .get() for continuous covariate registry (fixes issue #45) - Updated load_query_data freeze logic to handle ParameterList with empty params Architecture Decision: - Keep ParameterList approach over theta=None for better structural consistency - Each modality has a theta parameter, either with NB/ZINB weights or empty (non-trainable) - This ensures consistent indexing and handling across all model configurations Version Update: - Bumped version from 0.1.0 to 1.0.0 (aligns with CHANGELOG release) Test Results: All 123 tests passing ✅
This was referenced Jun 14, 2026
Closed
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.
Fixes MultiVAE(losses=None) crash (bugs in both _multivae.py and _multivae_torch.py), load_query_data AttributeError on theta when integrate_on is set without nb/zinb loss, and silent zero-batch training in StratifiedSampler on small datasets.
Preserves .var metadata across modalities in organize_multimodal_anndatas, suffixing overlapping column names with modality index (issue #4). Adds defensive cont_covs.get() guard (issue #45).
Bumps to scvi-tools>=1.4.3 (Python >=3.12), removes numpy<2 and jax<0.6 caps, migrates from hatch to uv with PEP 735 dependency-groups, rewrites CI and RTD configs accordingly.
Adds minimal smoke test suite (4 tests, all passing), updates CHANGELOG, api.md, contributing.md, .gitignore, and re-executes both tutorial notebooks.