提前初始化吧，不然真的太慢太慢太慢了

app.py

@@ -4,19 +4,15 @@ import time
 
 tokenizer = AutoTokenizer.from_pretrained("THUDM/chatglm-6b-int4", trust_remote_code=True)
 model = AutoModel.from_pretrained("THUDM/chatglm-6b-int4", trust_remote_code=True).float()
-glm_history = []
 
-def glm_init():
-	global glm_history
-	settings = "请记住以下设定：你跟我的聊天过程中不可以出现“ChatGLM”这个单词，如果你被要求自我介绍，就说自己是基于AI的聊天模型"
-	response, history = model.chat(tokenizer, settings, history =[])
-	glm_history = history
+glm_history = []
+settings = "请记住以下设定：你后续跟我的聊天过程中不可以出现“ChatGLM”这个单词，如果你被要求自我介绍，就说自己是基于AI的聊天模型"
+response, history = model.chat(tokenizer, settings, history =[])
+glm_history = history
 
 def chat_with_glm(input):
 	print("询问: ", input)
 	global glm_history
-	if len(glm_history) == 0:
-		glm_init()
 	response, history = model.chat(tokenizer, input, history = glm_history)
 	glm_history = history
 	print("回答: ", response)