Update app.py

app.py

@@ -58,7 +58,7 @@ def predict(message, history):
     #     if '</s>' in partial_message:  # Breaking the loop if the stop token is generated.
     #         break
     #     yield partial_message
-    inputs = tokenizer(message, return_tensors="pt")["input_ids"].to("cuda:0")
+    inputs = tokenizer(message, return_tensors="pt")["input_ids"].to(device)
     outputs = model.generate(inputs, do_sample=False, max_length=100)
     return(tokenizer.decode(outputs[0]))
 
@@ -67,5 +67,5 @@ def predict(message, history):
 gr.ChatInterface(predict,
                  title="Yuan2_2b_chatBot",
                  description="请提问",
-                 examples=['请问目前最先进的机器学习算法有哪些？']
+                 examples=['请问目前最先进的机器学习算法有哪些？','作一首关于春节的诗','北京烤鸭怎么做？']
                  ).launch()  # Launching the web interface.