Update qwen.py

qwen.py

@@ -8,6 +8,9 @@ model = AutoModelForCausalLM.from_pretrained(
     torch_dtype="auto",
     device_map="auto"
 )
+
+model = model.half()
+
 tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-7B-Instruct")
 
 def respond(