Ajout du print pour vérifier que le modèle ne se recharge qu'une fois

inference.py

@@ -132,6 +132,7 @@ label_encoder = joblib.load("label_encoder.pkl")
 class_centroids = joblib.load("class_centroids.pkl")
 best_threshold = joblib.load("oos_threshold.pkl")
 
+print("Model charging")
 # Recharger l'embedding model
 embedder = SentenceTransformer("intfloat/e5-small-v2")
 
@@ -145,8 +146,3 @@ model = IntentClassifierWithOOS(
     oos_threshold=best_threshold,
     device="cpu"
 )
-
-if __name__ == "__main__":
-    # Test with a sample query
-    result = model.predict("Can you play some jazz music?")
-    print(result)