app.py

from transformers import AutoTokenizer, AutoModel
import gradio as gr
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 reset_glm():
    global 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
	response, history = model.chat(tokenizer, input, history = glm_history)
	glm_history = history
	print("回答: ", response)
	print('----------------------')
	return response

def on_submit(input, history):
	global glm_history
	respond = "[" + time.strftime("%Y-%m-%d %H:%M:%S", time.localtime()) + "]\n\n"
	if input == '--history':
		if len(history) > 0:
			history_str = ""
			for h in history:
				if len(history_str) > 0:
					history_str += "\n"
				history_str += '\n'.join(h)
			history.append((input, respond + history_str))
	elif input == '--clear':
		history = []
		reset_glm()
	else:
		history.append((input, respond + chat_with_glm(input)))
	return "", history

with gr.Blocks() as demo:
	gr.Markdown("<center><h1>中文聊天机器人</h1></center>")
	chatbot = gr.Chatbot()
	tv_input = gr.Textbox()
	tv_input.submit(fn = on_submit, inputs = [tv_input, chatbot], outputs = [tv_input, chatbot])
	gr.Markdown("<b>说明：</b><br/>输入\"--clear\"清空历史，输入\"--history\"查看历史")
reset_glm()
demo.launch()