Respond with Special Tokens

main.py

@@ -80,7 +80,7 @@ async def generate_text(item: Item):
         pad_token_id=tokenizer.eos_token_id  # Set this to suppress warning
     )
 
-    resp = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    resp = tokenizer.decode(outputs[0], skip_special_tokens=False)
 
 
     return {"response": resp}