feat: extract runs to functions, to avoid instantiation on import

psycop/projects/forced_admission_inpatient/model_eval/model_description/feature_importance/gain_table.py

@@ -64,6 +64,6 @@ def generate_feature_importance_table(pipeline_run: PipelineRun) -> pl.DataFrame
 
 if __name__ == "__main__":
     top_100_features = generate_feature_importance_table(
-        pipeline_run=get_best_eval_pipeline,
+        pipeline_run=get_best_eval_pipeline(),
     )
     pass

psycop/projects/forced_admission_inpatient/model_eval/model_description/performance/auroc.py

@@ -25,4 +25,4 @@ def fa_auroc_plot(run: PipelineRun) -> pn.ggplot:
 
 
 if __name__ == "__main__":
-    fa_auroc_plot(run=get_best_eval_pipeline)
+    fa_auroc_plot(run=get_best_eval_pipeline())

psycop/projects/forced_admission_inpatient/model_eval/model_description/performance/confusion_matrix_pipeline.py

@@ -39,4 +39,4 @@ def fa_confusion_matrix_plot(run: PipelineRun) -> pn.ggplot:
 
 
 if __name__ == "__main__":
-    fa_confusion_matrix_plot(run=get_best_eval_pipeline)
+    fa_confusion_matrix_plot(run=get_best_eval_pipeline())

psycop/projects/forced_admission_inpatient/model_eval/model_description/performance/performance_by_ppr.py

@@ -88,4 +88,4 @@ def fa_output_performance_by_ppr(run: PipelineRun):
         get_best_eval_pipeline,
     )
 
-    fa_output_performance_by_ppr(run=get_best_eval_pipeline)
+    fa_output_performance_by_ppr(run=get_best_eval_pipeline())

psycop/projects/forced_admission_inpatient/model_eval/model_description/performance/sensitivity_by_time_to_event.py

@@ -109,4 +109,4 @@ def fa_sensitivity_by_time_to_event(pipeline_run: PipelineRun) -> pn.ggplot:
         get_best_eval_pipeline,
     )
 
-    fa_sensitivity_by_time_to_event(pipeline_run=get_best_eval_pipeline)
+    fa_sensitivity_by_time_to_event(pipeline_run=get_best_eval_pipeline())

psycop/projects/forced_admission_inpatient/model_eval/selected_runs.py

@@ -9,17 +9,21 @@
     PipelineRun,
 )
 
-get_best_dev_pipeline = PipelineRun(
-    group=DEVELOPMENT_GROUP,
-    name=DEVELOPMENT_GROUP.get_best_runs_by_lookahead()[
-        0,
-        2,
-    ],  # [0,2] for best logistic regression and [1,2] for best xgboost
-    pos_rate=BEST_POS_RATE,
-    create_output_paths_on_init=False,
-)
 
-get_best_eval_pipeline = test_selected_model_pipeline(
-    pipeline_to_test=get_best_dev_pipeline,
-    datasets_for_evaluation=["val_with_washout"],
-)
+def get_best_dev_pipeline() -> PipelineRun:
+    return PipelineRun(
+        group=DEVELOPMENT_GROUP,
+        name=DEVELOPMENT_GROUP.get_best_runs_by_lookahead()[
+            0,
+            2,
+        ],  # [0,2] for best logistic regression and [1,2] for best xgboost
+        pos_rate=BEST_POS_RATE,
+        create_output_paths_on_init=False,
+    )
+
+
+def get_best_eval_pipeline() -> PipelineRun:
+    return test_selected_model_pipeline(
+        pipeline_to_test=get_best_dev_pipeline(),
+        datasets_for_evaluation=["val_with_washout"],
+    )

psycop/projects/forced_admission_inpatient/model_eval/table_one/table_one.py

@@ -13,16 +13,20 @@
 
 model_train_df = pl.concat(
     [
-        get_best_eval_pipeline.inputs.get_flattened_split_as_lazyframe(split="train"),
+        get_best_eval_pipeline().inputs.get_flattened_split_as_lazyframe(split="train"),
     ],
     how="vertical",
 ).with_columns(dataset=pl.format("0. train"))
 
 
-val_dataset = get_best_eval_pipeline.inputs.get_flattened_split_as_lazyframe(
-    split="val",
-).with_columns(
-    dataset=pl.format("val"),
+val_dataset = (
+    get_best_eval_pipeline()
+    .inputs.get_flattened_split_as_lazyframe(
+        split="val",
+    )
+    .with_columns(
+        dataset=pl.format("val"),
+    )
 )
 
 flattened_combined = pl.concat([model_train_df, val_dataset], how="vertical").rename(
@@ -156,9 +160,9 @@
 ############
 combined = pd.concat([visit_table_one, patient_table_one])
 
-get_best_eval_pipeline.paper_outputs.paths.tables.mkdir(parents=True, exist_ok=True)
+get_best_eval_pipeline().paper_outputs.paths.tables.mkdir(parents=True, exist_ok=True)
 combined.to_csv(
-    get_best_eval_pipeline.paper_outputs.paths.tables / "descriptive_stats_table.csv",
+    get_best_eval_pipeline().paper_outputs.paths.tables / "descriptive_stats_table.csv",
 )
 
 # %%

psycop/projects/t2d/paper_outputs/aggregate_eval/supplementary_full_eval.py

@@ -106,7 +106,7 @@ def full_eval_for_supplementary(
         )
 
         # Do not add the main pipeline's eval to supplementary
-        if run.name != get_best_eval_pipeline.name:
+        if run.name != get_best_eval_pipeline().name:
             artifacts += run_artifacts
 
     combined_supplementary_md = create_supplementary_from_markdown_artifacts(

psycop/projects/t2d/paper_outputs/dataset_description/describe_features.py

@@ -28,11 +28,11 @@
 )
 from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
-out_dir = get_best_eval_pipeline.paper_outputs.paths.tables / "feature_description"
+out_dir = get_best_eval_pipeline().paper_outputs.paths.tables / "feature_description"
 out_dir.mkdir(parents=True, exist_ok=True)
 
 df = generate_feature_description_df(
-    df=get_best_eval_pipeline.inputs.get_flattened_split_as_pd(split="train"),
+    df=get_best_eval_pipeline().inputs.get_flattened_split_as_pd(split="train"),
     predictor_specs=selected_specs,  # type: ignore
 )
 
@@ -89,7 +89,7 @@ def prettify_feature_description_df(
 from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
 predictor_description_path = (
-    get_best_eval_pipeline.paper_outputs.paths.tables / "predictor_description.csv"
+    get_best_eval_pipeline().paper_outputs.paths.tables / "predictor_description.csv"
 )
 
 prettified.to_csv(predictor_description_path)
@@ -119,7 +119,7 @@ def prettify_feature_description_df(
 ).get_markdown()
 
 with (
-    get_best_eval_pipeline.paper_outputs.paths.tables / "predictor_description.md"
+    get_best_eval_pipeline().paper_outputs.paths.tables / "predictor_description.md"
 ).open(
     "+w",
 ) as f:

psycop/projects/t2d/paper_outputs/dataset_description/hba1c_within_lookahead.py

@@ -96,7 +96,7 @@ def get_dataset(self, run: T2DPipelineRun) -> pl.DataFrame:
         )
 
         lookahead_days = (
-            get_best_eval_pipeline.inputs.cfg.preprocessing.pre_split.min_lookahead_days
+            get_best_eval_pipeline().inputs.cfg.preprocessing.pre_split.min_lookahead_days
         )
 
         hba1c_timestamps = hba1c()
@@ -239,8 +239,8 @@ def get_plot(self, run: T2DPipelineRun) -> pn.ggplot:
 
 
 if __name__ == "__main__":
-    pipeline = get_best_eval_pipeline
-    plot = MeasurementsWithinLookaheadPlot().get_plot(run=get_best_eval_pipeline)
+    pipeline = get_best_eval_pipeline()
+    plot = MeasurementsWithinLookaheadPlot().get_plot(run=get_best_eval_pipeline())
     size = (6.5, 8)
 
     plot.save(

psycop/projects/t2d/paper_outputs/dataset_description/incidence_by_time.py

@@ -26,10 +26,10 @@
 # %%
 from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
-get_best_eval_pipeline.paper_outputs.paths.figures.mkdir(parents=True, exist_ok=True)
+get_best_eval_pipeline().paper_outputs.paths.figures.mkdir(parents=True, exist_ok=True)
 
 save_path = (
-    get_best_eval_pipeline.paper_outputs.paths.figures
+    get_best_eval_pipeline().paper_outputs.paths.figures
     / "diabetes_incidence_by_time.png"
 )
 

psycop/projects/t2d/paper_outputs/dataset_description/selection_flowchart.py

@@ -6,7 +6,7 @@
 msg = Printer(timestamp=True)
 
 if __name__ == "__main__":
-    pipeline_inputs = get_best_eval_pipeline.inputs
+    pipeline_inputs = get_best_eval_pipeline().inputs
 
     flattened_dataset = pl.concat(
         [

psycop/projects/t2d/paper_outputs/dataset_description/table_one/table_one.py

@@ -11,16 +11,20 @@
 
 model_train_df = pl.concat(
     [
-        get_best_eval_pipeline.inputs.get_flattened_split_as_lazyframe(split="train"),
-        get_best_eval_pipeline.inputs.get_flattened_split_as_lazyframe(split="val"),
+        get_best_eval_pipeline().inputs.get_flattened_split_as_lazyframe(split="train"),
+        get_best_eval_pipeline().inputs.get_flattened_split_as_lazyframe(split="val"),
     ],
     how="vertical",
 ).with_columns(dataset=pl.format("0. train"))
 
-test_dataset = get_best_eval_pipeline.inputs.get_flattened_split_as_lazyframe(
-    split="test",
-).with_columns(
-    dataset=pl.format("test"),
+test_dataset = (
+    get_best_eval_pipeline()
+    .inputs.get_flattened_split_as_lazyframe(
+        split="test",
+    )
+    .with_columns(
+        dataset=pl.format("test"),
+    )
 )
 
 flattened_combined = pl.concat([model_train_df, test_dataset], how="vertical").rename(
@@ -209,9 +213,9 @@
 ############
 combined = pd.concat([visit_table_one, patient_table_one])
 
-get_best_eval_pipeline.paper_outputs.paths.tables.mkdir(parents=True, exist_ok=True)
+get_best_eval_pipeline().paper_outputs.paths.tables.mkdir(parents=True, exist_ok=True)
 combined.to_csv(
-    get_best_eval_pipeline.paper_outputs.paths.tables / "descriptive_stats_table.csv",
+    get_best_eval_pipeline().paper_outputs.paths.tables / "descriptive_stats_table.csv",
 )
 
 # %%

psycop/projects/t2d/paper_outputs/intervention_eval/hba1c.py

@@ -35,7 +35,7 @@ def time_from_first_pos_pred_to_next_hba1c(
 if __name__ == "__main__":
     from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
-    pipeline = get_best_eval_pipeline
+    pipeline = get_best_eval_pipeline()
     eval_ds = pipeline.pipeline_outputs.get_eval_dataset()
 
     positive_predictions = (

psycop/projects/t2d/paper_outputs/model_description/feature_importance/gain/gain_table.py

@@ -58,6 +58,6 @@ def generate_feature_importance_table(pipeline_run: T2DPipelineRun) -> pl.DataFr
 
 if __name__ == "__main__":
     top_100_features = generate_feature_importance_table(
-        pipeline_run=get_best_eval_pipeline,
+        pipeline_run=get_best_eval_pipeline(),
     )
     pass

psycop/projects/t2d/paper_outputs/model_description/feature_importance/shap/get_shap_values.py

@@ -153,7 +153,7 @@ def get_top_i_features_by_mean_abs_shap(
     from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
     shap_bundle = get_shap_bundle_for_best_run(
-        run=get_best_eval_pipeline,
+        run=get_best_eval_pipeline(),
         n_rows=1_000,
         cache_ver=0.1,
     )

psycop/projects/t2d/paper_outputs/model_description/feature_importance/shap/main.py

@@ -24,7 +24,7 @@
     plotting_df = pl.from_pandas(long_shap_df)
 
     shap_figures_path = (
-        get_best_eval_pipeline.paper_outputs.paths.figures / "shap_plot.png"
+        get_best_eval_pipeline().paper_outputs.paths.figures / "shap_plot.png"
     )
     shap_figures_path.mkdir(exist_ok=True, parents=True)
 

psycop/projects/t2d/paper_outputs/model_description/performance/auroc.py

@@ -22,4 +22,4 @@ def t2d_auroc_plot(run: T2DPipelineRun) -> pn.ggplot:
 if __name__ == "__main__":
     from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
-    t2d_auroc_plot(run=get_best_eval_pipeline)
+    t2d_auroc_plot(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_description/performance/confusion_matrix_pipeline.py

@@ -36,4 +36,4 @@ def t2d_confusion_matrix_plot(run: T2DPipelineRun) -> pn.ggplot:
 if __name__ == "__main__":
     from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
-    t2d_confusion_matrix_plot(run=get_best_eval_pipeline)
+    t2d_confusion_matrix_plot(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_description/performance/incidence_by_time_until_diagnosis.py

@@ -57,8 +57,8 @@ def t2d_first_pred_to_event(run: T2DPipelineRun) -> pn.ggplot:
 if __name__ == "__main__":
     from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
-    t2d_first_pred_to_event(run=get_best_eval_pipeline).save(
-        get_best_eval_pipeline.paper_outputs.paths.figures
+    t2d_first_pred_to_event(run=get_best_eval_pipeline()).save(
+        get_best_eval_pipeline().paper_outputs.paths.figures
         / "time_from_pred_to_event.png",
         width=5,
         height=5,

psycop/projects/t2d/paper_outputs/model_description/performance/main_performance_figure.py

@@ -32,4 +32,4 @@ def t2d_create_main_performance_figure(run: T2DPipelineRun) -> None:
 if __name__ == "__main__":
     from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
-    t2d_create_main_performance_figure(run=get_best_eval_pipeline)
+    t2d_create_main_performance_figure(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_description/performance/performance_by_ppr.py

@@ -87,4 +87,4 @@ def t2d_output_performance_by_ppr(run: T2DPipelineRun):
 if __name__ == "__main__":
     from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
-    t2d_output_performance_by_ppr(run=get_best_eval_pipeline)
+    t2d_output_performance_by_ppr(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_description/performance/sensitivity_by_time_to_event_pipeline.py

@@ -103,4 +103,4 @@ def t2d_sensitivity_by_time_to_event(run: T2DPipelineRun) -> pn.ggplot:
 if __name__ == "__main__":
     from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
-    t2d_sensitivity_by_time_to_event(run=get_best_eval_pipeline)
+    t2d_sensitivity_by_time_to_event(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_description/robustness/main_robustness_figure.py

@@ -38,4 +38,4 @@ def t2d_create_main_robustness_figure(run: T2DPipelineRun) -> None:
 
 
 if __name__ == "__main__":
-    t2d_create_main_robustness_figure(run=get_best_eval_pipeline)
+    t2d_create_main_robustness_figure(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_description/robustness/robustness_by_age.py

@@ -32,4 +32,4 @@ def t2d_auroc_by_age(run: T2DPipelineRun) -> pn.ggplot:
 
 
 if __name__ == "__main__":
-    t2d_auroc_by_age(run=get_best_eval_pipeline)
+    t2d_auroc_by_age(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_description/robustness/robustness_by_calendar_time.py

@@ -30,4 +30,4 @@ def t2d_auroc_by_quarter(run: T2DPipelineRun) -> pn.ggplot:
 
 
 if __name__ == "__main__":
-    t2d_auroc_by_quarter(run=get_best_eval_pipeline)
+    t2d_auroc_by_quarter(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_description/robustness/robustness_by_cyclic_time.py

@@ -47,5 +47,5 @@ def t2d_auroc_by_month_of_year(run: T2DPipelineRun) -> pn.ggplot:
 
 
 if __name__ == "__main__":
-    t2d_auroc_by_day_of_week(run=get_best_eval_pipeline)
-    t2d_auroc_by_month_of_year(run=get_best_eval_pipeline)
+    t2d_auroc_by_day_of_week(run=get_best_eval_pipeline())
+    t2d_auroc_by_month_of_year(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_description/robustness/robustness_by_n_hba1c.py

@@ -35,4 +35,4 @@ def t2d_auroc_by_n_hba1c(
 
 
 if __name__ == "__main__":
-    t2d_auroc_by_n_hba1c(run=get_best_eval_pipeline)
+    t2d_auroc_by_n_hba1c(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_description/robustness/robustness_by_sex.py

@@ -31,4 +31,4 @@ def t2d_auroc_by_sex(run: T2DPipelineRun) -> pn.ggplot:
 
 
 if __name__ == "__main__":
-    t2d_auroc_by_sex(run=get_best_eval_pipeline)
+    t2d_auroc_by_sex(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_description/robustness/robustness_by_time_from_first_visit.py

@@ -60,4 +60,4 @@ def t2d_auroc_by_time_from_first_visit(run: T2DPipelineRun) -> pn.ggplot:
 
 
 if __name__ == "__main__":
-    t2d_auroc_by_time_from_first_visit(run=get_best_eval_pipeline)
+    t2d_auroc_by_time_from_first_visit(run=get_best_eval_pipeline())

psycop/projects/t2d/paper_outputs/model_permutation/boolean_features.py

@@ -66,7 +66,7 @@ def modify_features(
     )
 
     evaluate_pipeline_with_modified_dataset(
-        run=get_best_eval_pipeline,
+        run=get_best_eval_pipeline(),
         feature_modifier=CreateBooleanDataset(),
     )
 

psycop/projects/t2d/paper_outputs/model_permutation/snoozing/try_snoozing.py

@@ -9,7 +9,7 @@
 from psycop.projects.t2d.paper_outputs.selected_runs import get_best_eval_pipeline
 
 if __name__ == "__main__":
-    evaluation_dataset = get_best_eval_pipeline.pipeline_outputs.get_eval_dataset()
+    evaluation_dataset = get_best_eval_pipeline().pipeline_outputs.get_eval_dataset()
 
     eval_df = pd.DataFrame(
         {

psycop/projects/t2d/paper_outputs/run_pipeline_on_train.py

@@ -74,4 +74,4 @@ def test_pipeline(
 if __name__ == "__main__":
     from psycop.projects.t2d.paper_outputs.selected_runs import get_best_dev_pipeline
 
-    eval_run = test_pipeline(pipeline_to_test=get_best_dev_pipeline)
+    eval_run = test_pipeline(pipeline_to_test=get_best_dev_pipeline())