Update app.py

app.py

@@ -4,7 +4,12 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 # 加载本地模型和tokenizer
 model_name = "ganchengguang/OIELLM-8B-Instruction"  # 替换为你的模型名称
 tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForCausalLM.from_pretrained(model_name, load_in_8bit=True, device_map='auto')
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    load_in_8bit=True,
+    device_map='auto',
+    low_cpu_mem_usage=True
+)
 
 # 定义语言和选项的映射
 options = {