restapi

app.py

@@ -175,25 +175,22 @@ class QwenReranker(RerankerModel):
     def _process_inputs(self, pairs: List[str]):
         """Process input pairs for Qwen model."""
         inputs = self.tokenizer(
-            pairs, 
-            padding=False, 
-            truncation='longest_first',
-            return_attention_mask=False, 
-            max_length=self.max_length - len(self.prefix_tokens) - len(self.suffix_tokens)
+            pairs,
+            padding="max_length",        # ✅ pad langsung ke max_length
+            truncation=True,             # ✅ aktifkan truncation
+            max_length=self.max_length - len(self.prefix_tokens) - len(self.suffix_tokens),
+            return_tensors="pt"
         )
-        
+
+        # Tambahkan prefix & suffix tokens
         for i, ele in enumerate(inputs['input_ids']):
-            inputs['input_ids'][i] = self.prefix_tokens + ele + self.suffix_tokens
-        
-        inputs = self.tokenizer.pad(
-            inputs, 
-            padding=True, 
-            return_tensors="pt", 
-            max_length=self.max_length
-        )
-        
-        for key in inputs:
-            inputs[key] = inputs[key].to(self.model.device)
+            inputs['input_ids'][i] = torch.tensor(
+                self.prefix_tokens + ele.tolist() + self.suffix_tokens, 
+                dtype=torch.long
+            )
+
+        inputs = {k: v.to(self.model.device) for k, v in inputs.items()}
+
         
         return inputs