dealing with empty outputs

model.py

@@ -62,7 +62,7 @@ def run(message: str,
         top_p: float = 0.95,
         top_k: int = 50) -> Iterator[str]:
     prompt = get_prompt(message, chat_history, system_prompt)
-    inputs = tokenizer([prompt], return_tensors='pt', add_special_tokens=False).to(torch.device) #.to('cuda')
+    inputs = tokenizer([prompt], return_tensors='pt', add_special_tokens=False) #.to(torch.device) #.to('cuda')
 
     streamer = TextIteratorStreamer(tokenizer,
                                     timeout=10.,
@@ -83,5 +83,6 @@ def run(message: str,
 
     outputs = []
     for text in streamer:
+        print(f"output text ->{text}<- end of text")
         outputs.append(text)
         yield ''.join(outputs)