fix truncate

app/ner/services/ner.py

@@ -22,9 +22,23 @@ class NER:
     async def predict(self, text: str, entity_tag: str = None):
         if self.pipeline is None:
             raise ValueError("Model not loaded. Please call load_model() first.")
-        pred = self.pipeline(text, 
-                             truncation=settings.TRUNCATE,
-                             max_length=settings.MAX_LENGTH)
+        
+        inputs = self.tokenizer(
+            text,
+            truncation=settings.TRUNCATE,  # Enable truncation
+            max_length=settings.MAX_LENGTH,  # Set maximum length
+            return_tensors="pt"  # Make sure the tokenized inputs are returned as PyTorch tensors
+        )
+
+        # Get the prediction from the model
+        pred = self.model(**inputs)
+        
+        # Convert model output to pipeline format
+        pred = self.pipeline.decode(pred.logits)
+        
+        # pred = self.pipeline(text, 
+        #                      truncation=settings.TRUNCATE,
+        #                      max_length=settings.MAX_LENGTH)
         if entity_tag:
             return self.extract_entities(pred, entity_tag)
         return pred