cellmil.utils.train.evals.utils¶

Functions

`compute_slide_cell_counts`(dataset, slides)	Estimate per-slide cell counts for balancing folds.
`is_survival_model`(lit_model)	Check if the model is a survival analysis model.

Classes

Report(trainer, lit_model)

Base class for evaluation reports.

cellmil.utils.train.evals.utils.is_survival_model(lit_model: LightningModule) → bool[source]¶: Check if the model is a survival analysis model.

cellmil.utils.train.evals.utils.compute_slide_cell_counts(dataset: Union[CellMILDataset, CellGNNMILDataset, PatchGNNMILDataset, PatchMILDataset], slides: ndarray[Any, Any])[source]¶: Estimate per-slide cell counts for balancing folds.

class cellmil.utils.train.evals.utils.Report(trainer: Trainer, lit_model: LightningModule)[source]¶

Bases: object

Base class for evaluation reports.

__init__(trainer: Trainer, lit_model: LightningModule) → None[source]¶

_load_best_model() → None[source]¶: Load the best model checkpoint if available.

generate(dataloader: Union[DataLoader[Any], DataLoader]) → dict[str, Any][source]¶: Generate the report.

_get_classification_report(dataloader: Union[DataLoader[Any], DataLoader]) → dict[str, Any][source]¶: Get classification report with precision, recall, f1-score.

_get_survival_report(dataloader: Union[DataLoader[Any], DataLoader]) → dict[str, Any][source]¶: Get survival analysis report with C-index