code

app.py

@@ -57,7 +57,8 @@ def respond(
     thread.start()
     # Print the generated text in real-time
     for new_text in streamer:
-        response += tokenizer.decode(tokenizer.encode(new_text), skip_special_tokens=True)
+        response += new_text
+        response = tokenizer.decode(tokenizer.encode(response), skip_special_tokens=True)
     yield response
     
 with gr.Blocks() as demo: