Create predict.py

utils/predict.py

@@ -0,0 +1,33 @@
+from pathlib import Path
+import joblib
+from sentence_transformers import SentenceTransformer
+
+# ---------------------------
+# Path setup
+# ---------------------------
+BASE_DIR = Path(__file__).resolve().parents[1]
+MODEL_DIR = BASE_DIR / "model"
+
+# ---------------------------
+# Load models
+# ---------------------------
+def load_model():
+    logistic_model = joblib.load(MODEL_DIR / "logistic_model.pkl")
+
+    s2v_model = SentenceTransformer(
+        "Pachinee/sentence2vec-brd"
+    )
+
+    return logistic_model, s2v_model
+
+
+# ---------------------------
+# Predict
+# ---------------------------
+def predict_label(texts, logistic_model, s2v_model):
+    embeddings = s2v_model.encode(
+        list(texts),
+        convert_to_numpy=True
+    )
+    preds = logistic_model.predict(embeddings)
+    return ["Clear" if p == 1 else "Unclear" for p in preds]