from transformers import AutoModel, AutoTokenizer import time import threading import importlib from toolbox import update_ui, get_conf from multiprocessing import Process, Pipe load_message = "jittorllms尚未加载,加载需要一段时间。注意,请避免混用多种jittor模型,否则可能导致显存溢出而造成卡顿,取决于`config.py`的配置,jittorllms消耗大量的内存(CPU)或显存(GPU),也许会导致低配计算机卡死 ……" ################################################################################# class GetGLMHandle(Process): def __init__(self): super().__init__(daemon=True) self.parent, self.child = Pipe() self.jittorllms_model = None self.info = "" self.local_history = [] self.success = True self.check_dependency() self.start() self.threadLock = threading.Lock() def check_dependency(self): try: import pandas self.info = "依赖检测通过" self.success = True except: from toolbox import trimmed_format_exc self.info = r"缺少jittorllms的依赖,如果要使用jittorllms,除了基础的pip依赖以外,您还需要运行`pip install -r request_llm/requirements_jittorllms.txt -i https://pypi.jittor.org/simple -I`"+\ r"和`git clone https://gitlink.org.cn/jittor/JittorLLMs.git --depth 1 request_llm/jittorllms`两个指令来安装jittorllms的依赖(在项目根目录运行这两个指令)。" +\ r"警告:安装jittorllms依赖后将完全破坏现有的pytorch环境,建议使用docker环境!" + trimmed_format_exc() self.success = False def ready(self): return self.jittorllms_model is not None def run(self): # 子进程执行 # 第一次运行,加载参数 def validate_path(): import os, sys dir_name = os.path.dirname(__file__) env = os.environ.get("PATH", "") os.environ["PATH"] = env.replace('/cuda/bin', '/x/bin') root_dir_assume = os.path.abspath(os.path.dirname(__file__) + '/..') os.chdir(root_dir_assume + '/request_llm/jittorllms') sys.path.append(root_dir_assume + '/request_llm/jittorllms') validate_path() # validate path so you can run from base directory def load_model(): import types try: if self.jittorllms_model is None: device, = get_conf('LOCAL_MODEL_DEVICE') from .jittorllms.models import get_model # availabel_models = ["chatglm", "pangualpha", "llama", "chatrwkv"] args_dict = {'model': 'pangualpha'} print('self.jittorllms_model = get_model(types.SimpleNamespace(**args_dict))') self.jittorllms_model = get_model(types.SimpleNamespace(**args_dict)) print('done get model') except: self.child.send('[Local Message] Call jittorllms fail 不能正常加载jittorllms的参数。') raise RuntimeError("不能正常加载jittorllms的参数!") print('load_model') load_model() # 进入任务等待状态 print('进入任务等待状态') while True: # 进入任务等待状态 kwargs = self.child.recv() query = kwargs['query'] history = kwargs['history'] # 是否重置 if len(self.local_history) > 0 and len(history)==0: print('触发重置') self.jittorllms_model.reset() self.local_history.append(query) print('收到消息,开始请求') try: for response in self.jittorllms_model.stream_chat(query, history): print(response) self.child.send(response) except: from toolbox import trimmed_format_exc print(trimmed_format_exc()) self.child.send('[Local Message] Call jittorllms fail.') # 请求处理结束,开始下一个循环 self.child.send('[Finish]') def stream_chat(self, **kwargs): # 主进程执行 self.threadLock.acquire() self.parent.send(kwargs) while True: res = self.parent.recv() if res != '[Finish]': yield res else: break self.threadLock.release() global pangu_glm_handle pangu_glm_handle = None ################################################################################# def predict_no_ui_long_connection(inputs, llm_kwargs, history=[], sys_prompt="", observe_window=[], console_slience=False): """ 多线程方法 函数的说明请见 request_llm/bridge_all.py """ global pangu_glm_handle if pangu_glm_handle is None: pangu_glm_handle = GetGLMHandle() if len(observe_window) >= 1: observe_window[0] = load_message + "\n\n" + pangu_glm_handle.info if not pangu_glm_handle.success: error = pangu_glm_handle.info pangu_glm_handle = None raise RuntimeError(error) # jittorllms 没有 sys_prompt 接口,因此把prompt加入 history history_feedin = [] for i in range(len(history)//2): history_feedin.append([history[2*i], history[2*i+1]] ) watch_dog_patience = 5 # 看门狗 (watchdog) 的耐心, 设置5秒即可 response = "" for response in pangu_glm_handle.stream_chat(query=inputs, history=history_feedin, system_prompt=sys_prompt, max_length=llm_kwargs['max_length'], top_p=llm_kwargs['top_p'], temperature=llm_kwargs['temperature']): print(response) if len(observe_window) >= 1: observe_window[0] = response if len(observe_window) >= 2: if (time.time()-observe_window[1]) > watch_dog_patience: raise RuntimeError("程序终止。") return response def predict(inputs, llm_kwargs, plugin_kwargs, chatbot, history=[], system_prompt='', stream = True, additional_fn=None): """ 单线程方法 函数的说明请见 request_llm/bridge_all.py """ chatbot.append((inputs, "")) global pangu_glm_handle if pangu_glm_handle is None: pangu_glm_handle = GetGLMHandle() chatbot[-1] = (inputs, load_message + "\n\n" + pangu_glm_handle.info) yield from update_ui(chatbot=chatbot, history=[]) if not pangu_glm_handle.success: pangu_glm_handle = None return if additional_fn is not None: import core_functional importlib.reload(core_functional) # 热更新prompt core_functional = core_functional.get_core_functions() if "PreProcess" in core_functional[additional_fn]: inputs = core_functional[additional_fn]["PreProcess"](inputs) # 获取预处理函数(如果有的话) inputs = core_functional[additional_fn]["Prefix"] + inputs + core_functional[additional_fn]["Suffix"] # 处理历史信息 history_feedin = [] for i in range(len(history)//2): history_feedin.append([history[2*i], history[2*i+1]] ) # 开始接收jittorllms的回复 response = "[Local Message]: 等待jittorllms响应中 ..." for response in pangu_glm_handle.stream_chat(query=inputs, history=history_feedin, system_prompt=system_prompt, max_length=llm_kwargs['max_length'], top_p=llm_kwargs['top_p'], temperature=llm_kwargs['temperature']): chatbot[-1] = (inputs, response) yield from update_ui(chatbot=chatbot, history=history) # 总结输出 if response == "[Local Message]: 等待jittorllms响应中 ...": response = "[Local Message]: jittorllms响应异常 ..." history.extend([inputs, response]) yield from update_ui(chatbot=chatbot, history=history)