Feature/classic mil - #34
Merged
Merged
Conversation
Contributor
|
Caution The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased. |
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (5)
🚧 Files skipped from review as they are similar to previous changes (4)
📝 WalkthroughWalkthroughThe change introduces slide-label-only bag datasets and datamodules, extracts shared MIL behavior into ChangesClassic MIL workflows
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant Dataset as SLLabeledBagOfEmbeddingsDataset
participant DataModule as ClassicBagOfTilesDataModule
participant Model as ProstateCancerClassicMIL
participant Base as ProstateCancerMILBase
Dataset->>DataModule: provide slide-labeled bags
DataModule->>Model: collate slide labels and embeddings
Model->>Base: compute attention and slide prediction
Base-->>Model: return slide logits
Model-->>DataModule: log slide loss and metrics
Possibly related PRs
Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (2)
prostate_cancer/datamodule/datasets/bag_of_embeddings_dataset.py (1)
32-32: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valuePlace
Generic[T]as the last base class.Ruff reports
PYI059becauseGeneric[T]is not last. Reorder fromclass BagOfEmbeddingsDataset(Dataset[T], Generic[T], ABC):toclass BagOfEmbeddingsDataset(Dataset[T], ABC, Generic[T]):.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@prostate_cancer/datamodule/datasets/bag_of_embeddings_dataset.py` at line 32, Update the base-class order in BagOfEmbeddingsDataset so Generic[T] is the last base class, preserving Dataset[T] and ABC before it.Source: Linters/SAST tools
prostate_cancer/mil_model_base.py (1)
97-133: 🩺 Stability & Availability | 🔵 Trivial | 💤 Low valueGuard against NaN when a bag has zero valid tiles.
maskat Line 104 can be all-zero if every tile in a bag is treated as padding. In that case,raw_attn.masked_fill(~mask.bool(), float("-inf"))at Line 110 sets every attention logit to-inf, andtorch.softmaxover an all--infrow producesNaN(0/0). ThatNaNthen propagates intosl_pred_rawand the downstream loss/metrics for the whole batch.The class docstring states a bag is "all embeddings from a single slide," so this is unlikely in practice today, but there is no explicit safeguard if an upstream data issue ever produces an empty bag. Consider asserting
mask.sum(dim=1) > 0per bag, or clamping the softmax denominator.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@prostate_cancer/mil_model_base.py` around lines 97 - 133, Update forward to explicitly handle bags with zero valid tiles before softmax, using a per-bag validity check such as mask.sum(dim=1) > 0 and the project’s established failure-handling convention. Ensure attention_weights never comes from an all--inf row, preventing NaN propagation into sl_pred_raw and downstream outputs.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@prostate_cancer/mil_model_base.py`:
- Around line 135-149: Apply sigmoid once to the slide-level and tile-level
outputs before passing them to binary metric update calls, and reuse those
probabilities for the returned values where applicable. Update test_step in
prostate_cancer/mil_model_base.py#L135-L149, both affected metric-update sites
in prostate_cancer/attention_mil_model.py#L59-L60 and `#L99-L100`, and both sites
in prostate_cancer/classic_mil_model.py#L25 and `#L40`; preserve masking and
existing metric/logging behavior.
In `@prostate_cancer/typing.py`:
- Line 46: Update the comment describing the UnlabeledTileSampleBatch alias to
state “Images | Metadata” instead of “Images | labels | Metadata,” matching its
Tensor and TileMetadataBatch contents.
---
Nitpick comments:
In `@prostate_cancer/datamodule/datasets/bag_of_embeddings_dataset.py`:
- Line 32: Update the base-class order in BagOfEmbeddingsDataset so Generic[T]
is the last base class, preserving Dataset[T] and ABC before it.
In `@prostate_cancer/mil_model_base.py`:
- Around line 97-133: Update forward to explicitly handle bags with zero valid
tiles before softmax, using a per-bag validity check such as mask.sum(dim=1) > 0
and the project’s established failure-handling convention. Ensure
attention_weights never comes from an all--inf row, preventing NaN propagation
into sl_pred_raw and downstream outputs.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 2ed322df-24aa-4869-8280-3d5402420d27
📒 Files selected for processing (64)
configs/datamodule/classic_bag_of_embeddings_datamodule.yamlconfigs/datamodule/datasets/embeddings/bag_of_embeddings_sl_labeled.yamlconfigs/experiment/predict/bag_of_embeddings/base.yamlconfigs/experiment/predict/bag_of_embeddings/classic_base.yamlconfigs/experiment/predict/bag_of_embeddings/hybrid_base.yamlconfigs/experiment/predict/bag_of_embeddings/pgp/classic_base.yamlconfigs/experiment/predict/bag_of_embeddings/pgp/classic_slide_threshold.yamlconfigs/experiment/predict/bag_of_embeddings/pgp/hybrid_base.yamlconfigs/experiment/predict/bag_of_embeddings/pgp/hybrid_slide_threshold.yamlconfigs/experiment/predict/bag_of_embeddings/pgp/mil_predictions/classic_base.yamlconfigs/experiment/predict/bag_of_embeddings/pgp/mil_predictions/classic_mmci_sl_group_1.yamlconfigs/experiment/predict/bag_of_embeddings/pgp/mil_predictions/classic_mmci_tl_val.yamlconfigs/experiment/predict/bag_of_embeddings/pgp/mil_predictions/hybrid_base.yamlconfigs/experiment/predict/bag_of_embeddings/pgp/mil_predictions/hybrid_mmci_sl_group_1.yamlconfigs/experiment/predict/bag_of_embeddings/pgp/mil_predictions/hybrid_mmci_tl_val.yamlconfigs/experiment/predict/bag_of_embeddings/virchow2/classic_base.yamlconfigs/experiment/predict/bag_of_embeddings/virchow2/classic_slide_threshold.yamlconfigs/experiment/predict/bag_of_embeddings/virchow2/hybrid_base.yamlconfigs/experiment/predict/bag_of_embeddings/virchow2/hybrid_slide_threshold.yamlconfigs/experiment/predict/bag_of_embeddings/virchow2/mil_predictions/classic_base.yamlconfigs/experiment/predict/bag_of_embeddings/virchow2/mil_predictions/classic_mmci_sl_group_1.yamlconfigs/experiment/predict/bag_of_embeddings/virchow2/mil_predictions/classic_mmci_tl_val.yamlconfigs/experiment/predict/bag_of_embeddings/virchow2/mil_predictions/hybrid_base.yamlconfigs/experiment/predict/bag_of_embeddings/virchow2/mil_predictions/hybrid_mmci_sl_group_1.yamlconfigs/experiment/predict/bag_of_embeddings/virchow2/mil_predictions/hybrid_mmci_tl_val.yamlconfigs/experiment/test/bag_of_embeddings/base.yamlconfigs/experiment/test/bag_of_embeddings/classic_base.yamlconfigs/experiment/test/bag_of_embeddings/hybrid_base.yamlconfigs/experiment/test/bag_of_embeddings/pgp/classic_base.yamlconfigs/experiment/test/bag_of_embeddings/pgp/classic_tile_threshold.yamlconfigs/experiment/test/bag_of_embeddings/pgp/hybrid_base.yamlconfigs/experiment/test/bag_of_embeddings/pgp/hybrid_tile_threshold.yamlconfigs/experiment/test/bag_of_embeddings/pgp/test_suite/classic_base.yamlconfigs/experiment/test/bag_of_embeddings/pgp/test_suite/classic_mmci_tl_test.yamlconfigs/experiment/test/bag_of_embeddings/pgp/test_suite/hybrid_base.yamlconfigs/experiment/test/bag_of_embeddings/pgp/test_suite/hybrid_mmci_tl_test.yamlconfigs/experiment/test/bag_of_embeddings/virchow2/classic_base.yamlconfigs/experiment/test/bag_of_embeddings/virchow2/classic_tile_threshold.yamlconfigs/experiment/test/bag_of_embeddings/virchow2/hybrid_base.yamlconfigs/experiment/test/bag_of_embeddings/virchow2/hybrid_tile_threshold.yamlconfigs/experiment/test/bag_of_embeddings/virchow2/test_suite/classic_base.yamlconfigs/experiment/test/bag_of_embeddings/virchow2/test_suite/classic_mmci_tl_test.yamlconfigs/experiment/test/bag_of_embeddings/virchow2/test_suite/hybrid_base.yamlconfigs/experiment/test/bag_of_embeddings/virchow2/test_suite/hybrid_mmci_tl_test.yamlconfigs/experiment/train/bag_of_embeddings/base.yamlconfigs/experiment/train/bag_of_embeddings/classic_base.yamlconfigs/experiment/train/bag_of_embeddings/classic_pgp.yamlconfigs/experiment/train/bag_of_embeddings/classic_virchow2.yamlconfigs/experiment/train/bag_of_embeddings/hybrid_base.yamlconfigs/experiment/train/bag_of_embeddings/hybrid_pgp.yamlconfigs/experiment/train/bag_of_embeddings/hybrid_virchow2.yamlconfigs/model/checkpoints/mmci_tl/pgp_classic_mil.yamlconfigs/model/checkpoints/mmci_tl/virchow2_classic_mil.yamlconfigs/model/pgp_classic_mil_model.yamlconfigs/model/virchow2_classic_mil_model.yamlprostate_cancer/attention_mil_model.pyprostate_cancer/classic_mil_model.pyprostate_cancer/datamodule/__init__.pyprostate_cancer/datamodule/bag_of_tiles_data_module.pyprostate_cancer/datamodule/datasets/__init__.pyprostate_cancer/datamodule/datasets/bag_of_embeddings_dataset.pyprostate_cancer/datamodule/samplers/bag_of_tiles_stratified_weighted_random_sampler.pyprostate_cancer/mil_model_base.pyprostate_cancer/typing.py
💤 Files with no reviewable changes (2)
- configs/experiment/test/bag_of_embeddings/base.yaml
- configs/experiment/train/bag_of_embeddings/base.yaml
Adames4
approved these changes
Aug 3, 2026
vejtek
approved these changes
Aug 4, 2026
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
3 participants
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.
This PR adds MIL with SL supervision
Summary by CodeRabbit
New Features
Improvements