Add/multimodal rnn + soft fix mask absolute value bug

docs/api/models/pyhealth.models.RNN.rst

@@ -10,6 +10,11 @@ The separate callable RNNLayer and the complete RNN model.
     :show-inheritance:
 
 .. autoclass:: pyhealth.models.RNN
+    :members:
+    :undoc-members:
+    :show-inheritance:
+
+.. autoclass:: pyhealth.models.MultimodalRNN
     :members:
     :undoc-members:
     :show-inheritance:

examples/drug_recommendation/drug_recommendation_eicu_transformer.py

@@ -0,0 +1,74 @@
+"""
+Drug Recommendation on eICU with Transformer
+
+This example demonstrates how to use the modernized eICUDataset with the 
+DrugRecommendationEICU task class for drug recommendation using a Transformer model.
+
+Features:
+- Uses the new BaseDataset-based eICUDataset with YAML configuration
+- Uses the new DrugRecommendationEICU BaseTask class
+- Demonstrates the standardized PyHealth workflow
+"""
+
+import tempfile
+
+from pyhealth.datasets import eICUDataset
+from pyhealth.datasets import split_by_patient, get_dataloader
+from pyhealth.models import Transformer
+from pyhealth.tasks import DrugRecommendationEICU
+from pyhealth.trainer import Trainer
+
+
+if __name__ == "__main__":
+    # Use tempfile to automate cleanup
+    cache_dir = tempfile.TemporaryDirectory()
+
+    # STEP 1: Load dataset
+    # Replace with your eICU dataset path
+    base_dataset = eICUDataset(
+        root="/srv/local/data/physionet.org/files/eicu-crd/2.0",
+        tables=["diagnosis", "medication", "physicalexam"],
+        cache_dir=cache_dir.name
+    )
+    base_dataset.stats()
+
+    # STEP 2: Set task using the new DrugRecommendationEICU class
+    task = DrugRecommendationEICU()
+    sample_dataset = base_dataset.set_task(task)
+    sample_dataset.stats()
+
+    # STEP 3: Split and create dataloaders
+    train_dataset, val_dataset, test_dataset = split_by_patient(
+        sample_dataset, [0.8, 0.1, 0.1]
+    )
+    train_dataloader = get_dataloader(train_dataset, batch_size=32, shuffle=True)
+    val_dataloader = get_dataloader(val_dataset, batch_size=32, shuffle=False)
+    test_dataloader = get_dataloader(test_dataset, batch_size=32, shuffle=False)
+
+    # STEP 4: Define model
+    model = Transformer(
+        dataset=sample_dataset,
+        feature_keys=["conditions", "procedures", "drugs_hist"],
+        label_key="drugs",
+        mode="multilabel",
+    )
+
+    # STEP 5: Train
+    trainer = Trainer(model=model)
+    trainer.train(
+        train_dataloader=train_dataloader,
+        val_dataloader=val_dataloader,
+        epochs=50,
+        monitor="pr_auc_samples",
+    )
+
+    # STEP 6: Evaluate
+    print(trainer.evaluate(test_dataloader))
+
+    # Cleanup
+    sample_dataset.close()
+
+
+
+
+

examples/length_of_stay/length_of_stay_eicu_rnn.py

@@ -0,0 +1,88 @@
+"""
+Length of Stay Prediction on eICU with RNN
+
+This example demonstrates how to use the modernized eICUDataset with the 
+LengthOfStayPredictioneICU task class for predicting ICU length of stay
+using an RNN model.
+
+Length of stay is categorized into 10 classes:
+- 0: < 1 day
+- 1: 1 day  
+- 2: 2 days
+- 3: 3 days
+- 4: 4 days
+- 5: 5 days
+- 6: 6 days
+- 7: 7 days
+- 8: 1-2 weeks
+- 9: > 2 weeks
+
+Features:
+- Uses the new BaseDataset-based eICUDataset with YAML configuration  
+- Uses the LengthOfStayPredictioneICU BaseTask class
+- Demonstrates the standardized PyHealth workflow
+"""
+
+import tempfile
+
+from pyhealth.datasets import eICUDataset
+from pyhealth.datasets import split_by_patient, get_dataloader
+from pyhealth.models import RNN
+from pyhealth.tasks import LengthOfStayPredictioneICU
+from pyhealth.trainer import Trainer
+
+
+if __name__ == "__main__":
+    # Use tempfile to automate cleanup
+    cache_dir = tempfile.TemporaryDirectory()
+
+    # STEP 1: Load dataset
+    # Replace with your eICU dataset path
+    base_dataset = eICUDataset(
+        root="/srv/local/data/physionet.org/files/eicu-crd/2.0",
+        tables=["diagnosis", "medication", "physicalexam"],
+        cache_dir=cache_dir.name
+        dev=True,
+    )
+    base_dataset.stats()
+
+    # STEP 2: Set task using LengthOfStayPredictioneICU
+    task = LengthOfStayPredictioneICU()
+    sample_dataset = base_dataset.set_task(task)
+    sample_dataset.stats()
+
+    # STEP 3: Split and create dataloaders
+    train_dataset, val_dataset, test_dataset = split_by_patient(
+        sample_dataset, [0.8, 0.1, 0.1]
+    )
+    train_dataloader = get_dataloader(train_dataset, batch_size=32, shuffle=True)
+    val_dataloader = get_dataloader(val_dataset, batch_size=32, shuffle=False)
+    test_dataloader = get_dataloader(test_dataset, batch_size=32, shuffle=False)
+
+    # STEP 4: Define model
+    model = RNN(
+        dataset=sample_dataset,
+        feature_keys=["conditions", "procedures", "drugs"],
+        label_key="los",
+        mode="multiclass",
+    )
+
+    # STEP 5: Train
+    trainer = Trainer(model=model)
+    trainer.train(
+        train_dataloader=train_dataloader,
+        val_dataloader=val_dataloader,
+        epochs=20,
+        monitor="accuracy",
+    )
+
+    # STEP 6: Evaluate
+    print(trainer.evaluate(test_dataloader))
+
+    # Cleanup
+    sample_dataset.close()
+
+
+
+
+