Update app.py

app.py

@@ -66,16 +66,16 @@ def respond_stream(message, history, system_message, max_tokens, temperature, to
         kwargs={**inputs, **{k: v for k, v in gen_kwargs.items() if v is not None}, "streamer": streamer}
     )
 
-    partial_text = ""  # Biriken metin
+    input_len = inputs["input_ids"].shape[1]
+    partial_text = ""
     token_count = 0
     start_time = time.time()
 
     with torch.inference_mode():
         thread.start()
-        for token_text in streamer:
-            token_count += 1
-            partial_text += token_text  # Yeni token'ı ekle
-            yield partial_text  # Tamamını gönder (append efekti)
+        for new_text in streamer:
+            partial_text += new_text
+            yield partial_text
         thread.join()
 
     end_time = time.time()