chatglm-fitness-RLHF / request_llm /bridge_azure_test.py
fb700's picture
Upload 132 files
2ed7200
raw
history blame
10.5 kB
"""
该文件中主要包含三个函数
不具备多线程能力的函数:
1. predict: 正常对话时使用,具备完备的交互功能,不可多线程
具备多线程调用能力的函数
2. predict_no_ui:高级实验性功能模块调用,不会实时显示在界面上,参数简单,可以多线程并行,方便实现复杂的功能逻辑
3. predict_no_ui_long_connection:在实验过程中发现调用predict_no_ui处理长文档时,和openai的连接容易断掉,这个函数用stream的方式解决这个问题,同样支持多线程
"""
import logging
import traceback
import importlib
import openai
import time
# 读取config.py文件中关于AZURE OPENAI API的信息
from toolbox import get_conf, update_ui, clip_history, trimmed_format_exc
TIMEOUT_SECONDS, MAX_RETRY, AZURE_ENGINE, AZURE_ENDPOINT, AZURE_API_VERSION, AZURE_API_KEY = \
get_conf('TIMEOUT_SECONDS', 'MAX_RETRY',"AZURE_ENGINE","AZURE_ENDPOINT", "AZURE_API_VERSION", "AZURE_API_KEY")
def get_full_error(chunk, stream_response):
"""
获取完整的从Openai返回的报错
"""
while True:
try:
chunk += next(stream_response)
except:
break
return chunk
def predict(inputs, llm_kwargs, plugin_kwargs, chatbot, history=[], system_prompt='', stream = True, additional_fn=None):
"""
发送至azure openai api,流式获取输出。
用于基础的对话功能。
inputs 是本次问询的输入
top_p, temperature是chatGPT的内部调优参数
history 是之前的对话列表(注意无论是inputs还是history,内容太长了都会触发token数量溢出的错误)
chatbot 为WebUI中显示的对话列表,修改它,然后yeild出去,可以直接修改对话界面内容
additional_fn代表点击的哪个按钮,按钮见functional.py
"""
print(llm_kwargs["llm_model"])
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"]
raw_input = inputs
logging.info(f'[raw_input] {raw_input}')
chatbot.append((inputs, ""))
yield from update_ui(chatbot=chatbot, history=history, msg="等待响应") # 刷新界面
payload = generate_azure_payload(inputs, llm_kwargs, history, system_prompt, stream)
history.append(inputs); history.append("")
retry = 0
while True:
try:
openai.api_type = "azure"
openai.api_version = AZURE_API_VERSION
openai.api_base = AZURE_ENDPOINT
openai.api_key = AZURE_API_KEY
response = openai.ChatCompletion.create(timeout=TIMEOUT_SECONDS, **payload);break
except:
retry += 1
chatbot[-1] = ((chatbot[-1][0], "获取response失败,重试中。。。"))
retry_msg = f",正在重试 ({retry}/{MAX_RETRY}) ……" if MAX_RETRY > 0 else ""
yield from update_ui(chatbot=chatbot, history=history, msg="请求超时"+retry_msg) # 刷新界面
if retry > MAX_RETRY: raise TimeoutError
gpt_replying_buffer = ""
is_head_of_the_stream = True
if stream:
stream_response = response
while True:
try:
chunk = next(stream_response)
except StopIteration:
from toolbox import regular_txt_to_markdown; tb_str = '```\n' + trimmed_format_exc() + '```'
chatbot[-1] = (chatbot[-1][0], f"[Local Message] 远程返回错误: \n\n{tb_str} \n\n{regular_txt_to_markdown(chunk)}")
yield from update_ui(chatbot=chatbot, history=history, msg="远程返回错误:" + chunk) # 刷新界面
return
if is_head_of_the_stream and (r'"object":"error"' not in chunk):
# 数据流的第一帧不携带content
is_head_of_the_stream = False; continue
if chunk:
#print(chunk)
try:
if "delta" in chunk["choices"][0]:
if chunk["choices"][0]["finish_reason"] == "stop":
logging.info(f'[response] {gpt_replying_buffer}')
break
status_text = f"finish_reason: {chunk['choices'][0]['finish_reason']}"
gpt_replying_buffer = gpt_replying_buffer + chunk["choices"][0]["delta"]["content"]
history[-1] = gpt_replying_buffer
chatbot[-1] = (history[-2], history[-1])
yield from update_ui(chatbot=chatbot, history=history, msg=status_text) # 刷新界面
except Exception as e:
traceback.print_exc()
yield from update_ui(chatbot=chatbot, history=history, msg="Json解析不合常规") # 刷新界面
chunk = get_full_error(chunk, stream_response)
error_msg = chunk
yield from update_ui(chatbot=chatbot, history=history, msg="Json异常" + error_msg) # 刷新界面
return
def predict_no_ui_long_connection(inputs, llm_kwargs, history=[], sys_prompt="", observe_window=None, console_slience=False):
"""
发送至AZURE OPENAI API,等待回复,一次性完成,不显示中间过程。但内部用stream的方法避免中途网线被掐。
inputs:
是本次问询的输入
sys_prompt:
系统静默prompt
llm_kwargs:
chatGPT的内部调优参数
history:
是之前的对话列表
observe_window = None:
用于负责跨越线程传递已经输出的部分,大部分时候仅仅为了fancy的视觉效果,留空即可。observe_window[0]:观测窗。observe_window[1]:看门狗
"""
watch_dog_patience = 5 # 看门狗的耐心, 设置5秒即可
payload = generate_azure_payload(inputs, llm_kwargs, history, system_prompt=sys_prompt, stream=True)
retry = 0
while True:
try:
openai.api_type = "azure"
openai.api_version = AZURE_API_VERSION
openai.api_base = AZURE_ENDPOINT
openai.api_key = AZURE_API_KEY
response = openai.ChatCompletion.create(timeout=TIMEOUT_SECONDS, **payload);break
except:
retry += 1
traceback.print_exc()
if retry > MAX_RETRY: raise TimeoutError
if MAX_RETRY!=0: print(f'请求超时,正在重试 ({retry}/{MAX_RETRY}) ……')
stream_response = response
result = ''
while True:
try: chunk = next(stream_response)
except StopIteration:
break
except:
chunk = next(stream_response) # 失败了,重试一次?再失败就没办法了。
if len(chunk)==0: continue
if not chunk.startswith('data:'):
error_msg = get_full_error(chunk, stream_response)
if "reduce the length" in error_msg:
raise ConnectionAbortedError("AZURE OPENAI API拒绝了请求:" + error_msg)
else:
raise RuntimeError("AZURE OPENAI API拒绝了请求:" + error_msg)
if ('data: [DONE]' in chunk): break
delta = chunk["delta"]
if len(delta) == 0: break
if "role" in delta: continue
if "content" in delta:
result += delta["content"]
if not console_slience: print(delta["content"], end='')
if observe_window is not None:
# 观测窗,把已经获取的数据显示出去
if len(observe_window) >= 1: observe_window[0] += delta["content"]
# 看门狗,如果超过期限没有喂狗,则终止
if len(observe_window) >= 2:
if (time.time()-observe_window[1]) > watch_dog_patience:
raise RuntimeError("用户取消了程序。")
else: raise RuntimeError("意外Json结构:"+delta)
if chunk['finish_reason'] == 'length':
raise ConnectionAbortedError("正常结束,但显示Token不足,导致输出不完整,请削减单次输入的文本量。")
return result
def generate_azure_payload(inputs, llm_kwargs, history, system_prompt, stream):
"""
整合所有信息,选择LLM模型,生成 azure openai api请求,为发送请求做准备
"""
conversation_cnt = len(history) // 2
messages = [{"role": "system", "content": system_prompt}]
if conversation_cnt:
for index in range(0, 2*conversation_cnt, 2):
what_i_have_asked = {}
what_i_have_asked["role"] = "user"
what_i_have_asked["content"] = history[index]
what_gpt_answer = {}
what_gpt_answer["role"] = "assistant"
what_gpt_answer["content"] = history[index+1]
if what_i_have_asked["content"] != "":
if what_gpt_answer["content"] == "": continue
messages.append(what_i_have_asked)
messages.append(what_gpt_answer)
else:
messages[-1]['content'] = what_gpt_answer['content']
what_i_ask_now = {}
what_i_ask_now["role"] = "user"
what_i_ask_now["content"] = inputs
messages.append(what_i_ask_now)
payload = {
"model": llm_kwargs['llm_model'],
"messages": messages,
"temperature": llm_kwargs['temperature'], # 1.0,
"top_p": llm_kwargs['top_p'], # 1.0,
"n": 1,
"stream": stream,
"presence_penalty": 0,
"frequency_penalty": 0,
"engine": AZURE_ENGINE
}
try:
print(f" {llm_kwargs['llm_model']} : {conversation_cnt} : {inputs[:100]} ..........")
except:
print('输入中可能存在乱码。')
return payload