Add usage example for the LabTOP model

examples/labtop_usage.py

@@ -0,0 +1,45 @@
+"""
+Example: Lab Test Outcome Prediction with LabTOP
+
+This example shows how to use LabTOP model for predicting
+lab test values on MIMIC-IV dataset.
+
+Requirements:
+- MIMIC-IV dataset access
+- PyHealth installed
+"""
+
+from pyhealth.datasets import MIMIC4Dataset
+from pyhealth.models import LabTOP
+from pyhealth.trainer import Trainer
+
+# Step 1: Load dataset
+dataset = MIMIC4Dataset(
+    root="/path/to/mimic4",
+    tables=["labevents", "patients", "admissions"],
+)
+
+# Step 2: Create train/val/test splits
+train_dataset, val_dataset, test_dataset = dataset.split(ratios=[0.7, 0.15, 0.15])
+
+# Step 3: Initialize model
+model = LabTOP(
+    dataset=dataset,
+    feature_keys=["demographics", "lab_history"],
+    label_key="lab_value",
+    mode="regression",
+    embedding_dim=768,
+)
+
+# Step 4: Train
+trainer = Trainer(model=model, device="cuda")
+trainer.train(
+    train_dataset=train_dataset,
+    val_dataset=val_dataset,
+    epochs=10,
+    batch_size=32,
+)
+
+# Step 5: Evaluate
+metrics = trainer.evaluate(test_dataset)
+print(f"Test MAE: {metrics['mae']:.4f}")