Update app.py

app.py

@@ -65,13 +65,24 @@ def generate(req: UserInputRequest):
             {"role": "user", "content": req.user_input}
         ]
 
-        chat_input = tokenizer.apply_chat_template(
+        # === Önce chat template string'ini al
+        chat_template_str = tokenizer.apply_chat_template(
             messages,
             add_generation_prompt=True,
-            return_tensors="pt"
+            return_tensors=None
+        )
+
+        # === Sonra tokenizer() ile input_ids + attention_mask hazırla
+        tokenized_inputs = tokenizer(
+            chat_template_str,
+            return_tensors="pt",
+            padding=True
         ).to(model.device)
 
-        input_len = chat_input.shape[-1]
+        input_ids = tokenized_inputs['input_ids']
+        attention_mask = tokenized_inputs['attention_mask']
+
+        input_len = input_ids.shape[-1]
         total_ctx = model.config.max_position_embeddings if hasattr(model.config, 'max_position_embeddings') else 4096
         max_new_tokens = max(1, total_ctx - input_len)
 
@@ -83,7 +94,8 @@ def generate(req: UserInputRequest):
         ]
 
         outputs = model.generate(
-            input_ids=chat_input,
+            input_ids=input_ids,
+            attention_mask=attention_mask,
             max_new_tokens=max_new_tokens,
             eos_token_id=terminators
         )