Update app.py

app.py

@@ -10,11 +10,13 @@ adapter_model_name = "tiankuo1111/LLAMA3-TEST"
 # 加载 tokenizer
 tokenizer = AutoTokenizer.from_pretrained(base_model_name)
 
-# 加载基础模型
-base_model = AutoModelForCausalLM.from_pretrained(base_model_name, torch_dtype=torch.float16, device_map="auto")
+# 加载基础模型到 CPU
+base_model = AutoModelForCausalLM.from_pretrained(base_model_name, torch_dtype=torch.float32, device_map=None)
+base_model.to("cpu")  # 确保在 CPU 上运行
 
 # 加载 LoRA 适配器
 model = PeftModel.from_pretrained(base_model, adapter_model_name)
+model.to("cpu")  # 确保适配器也在 CPU 上运行
 
 # 运行测试
 def chat_with_model(prompt):