chunzhi/chatgpt_academic
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

update

Browse Source
This commit is contained in:
Your Name 2023-03-24 13:12:25 +08:00
parent ecdeda8e92
commit 44b40ff726
5 changed files with 176 additions and 141 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

91
colorful.py Normal file
View File

@ -0,0 +1,91 @@
import platform
from sys import stdout
if platform.system()=="Linux":
pass
else:
from colorama import init
init()
# Do you like the elegance of Chinese characters?
def print红(*kw,**kargs):
print("\033[0;31m",*kw,"\033[0m",**kargs)
def print绿(*kw,**kargs):
print("\033[0;32m",*kw,"\033[0m",**kargs)
def print黄(*kw,**kargs):
print("\033[0;33m",*kw,"\033[0m",**kargs)
def print蓝(*kw,**kargs):
print("\033[0;34m",*kw,"\033[0m",**kargs)
def print紫(*kw,**kargs):
print("\033[0;35m",*kw,"\033[0m",**kargs)
def print靛(*kw,**kargs):
print("\033[0;36m",*kw,"\033[0m",**kargs)
def print亮红(*kw,**kargs):
print("\033[1;31m",*kw,"\033[0m",**kargs)
def print亮绿(*kw,**kargs):
print("\033[1;32m",*kw,"\033[0m",**kargs)
def print亮黄(*kw,**kargs):
print("\033[1;33m",*kw,"\033[0m",**kargs)
def print亮蓝(*kw,**kargs):
print("\033[1;34m",*kw,"\033[0m",**kargs)
def print亮紫(*kw,**kargs):
print("\033[1;35m",*kw,"\033[0m",**kargs)
def print亮靛(*kw,**kargs):
print("\033[1;36m",*kw,"\033[0m",**kargs)
def print亮红(*kw,**kargs):
print("\033[1;31m",*kw,"\033[0m",**kargs)
def print亮绿(*kw,**kargs):
print("\033[1;32m",*kw,"\033[0m",**kargs)
def print亮黄(*kw,**kargs):
print("\033[1;33m",*kw,"\033[0m",**kargs)
def print亮蓝(*kw,**kargs):
print("\033[1;34m",*kw,"\033[0m",**kargs)
def print亮紫(*kw,**kargs):
print("\033[1;35m",*kw,"\033[0m",**kargs)
def print亮靛(*kw,**kargs):
print("\033[1;36m",*kw,"\033[0m",**kargs)
print_red = print红
print_green = print绿
print_yellow = print黄
print_blue = print蓝
print_purple = print紫
print_indigo = print靛
print_bold_red = print亮红
print_bold_green = print亮绿
print_bold_yellow = print亮黄
print_bold_blue = print亮蓝
print_bold_purple = print亮紫
print_bold_indigo = print亮靛
if not stdout.isatty():
# redirection, avoid a fucked up log file
print红 = print
print绿 = print
print黄 = print
print蓝 = print
print紫 = print
print靛 = print
print亮红 = print
print亮绿 = print
print亮黄 = print
print亮蓝 = print
print亮紫 = print
print亮靛 = print
print_red = print
print_green = print
print_yellow = print
print_blue = print
print_purple = print
print_indigo = print
print_bold_red = print
print_bold_green = print
print_bold_yellow = print
print_bold_blue = print
print_bold_purple = print
print_bold_indigo = print

3
config.py
View File

@ -21,6 +21,9 @@ WEB_PORT = -1
# 如果OpenAI不响应（网络卡顿、代理失败、KEY失效），重试的次数限制
MAX_RETRY = 2
# 选择的OpenAI模型是（gpt4现在只对申请成功的人开放）
LLM_MODEL = "gpt-3.5-turbo"
# 检查一下是不是忘了改config
if API_KEY == "sk-此处填API秘钥":
assert False, "请在config文件中修改API密钥, 添加海外代理之后再运行"

39
functional_crazy.py
View File

@ -1,26 +1,18 @@
# """
# 'primary' for main call-to-action,
# 'secondary' for a more subdued style,
# 'stop' for a stop button.
# """
from predict import predict_no_ui
fast_debug = False
def 自我程序解构简单案例(txt, top_p, temperature, chatbot, history, systemPromptTxt, WEB_PORT):
import time
from predict import predict_no_ui_no_history
def 高阶功能模板函数(txt, top_p, temperature, chatbot, history, systemPromptTxt, WEB_PORT):
for i in range(5):
i_say = f'我给出一个数字，你给出该数字的平方。我给出数字：{i}'
gpt_say = predict_no_ui_no_history(inputs=i_say, top_p=top_p, temperature=temperature)
gpt_say = predict_no_ui(inputs=i_say, top_p=top_p, temperature=temperature)
chatbot.append((i_say, gpt_say))
history.append(i_say)
history.append(gpt_say)
yield chatbot, history, '正常'
time.sleep(10)
def 解析项目本身(txt, top_p, temperature, chatbot, history, systemPromptTxt, WEB_PORT):
import time, glob, os
from predict import predict_no_ui
file_manifest = [f for f in glob.glob('*.py')]
for index, fp in enumerate(file_manifest):
@ -30,7 +22,7 @@ def 解析项目本身(txt, top_p, temperature, chatbot, history, systemPromptTx
前言 = "接下来请你分析自己的程序构成，别紧张，" if index==0 else ""
i_say = 前言 + f'请对下面的程序文件做一个概述文件名是{fp}，文件代码是 ```{file_content}```'
i_say_show_user = 前言 + f'[{index}/{len(file_manifest)}] 请对下面的程序文件做一个概述: {os.path.abspath(fp)}'
chatbot.append((i_say_show_user, "[local] waiting gpt response."))
chatbot.append((i_say_show_user, "[waiting gpt response]"))
yield chatbot, history, '正常'
if not fast_debug:
@ -43,7 +35,7 @@ def 解析项目本身(txt, top_p, temperature, chatbot, history, systemPromptTx
time.sleep(2)
i_say = f'根据以上你自己的分析，对程序的整体功能和构架做出概括。然后用一张markdown表格整理每个文件的功能（包括{file_manifest}）。'
chatbot.append((i_say, "[local] waiting gpt response."))
chatbot.append((i_say, "[waiting gpt response]"))
yield chatbot, history, '正常'
if not fast_debug:
@ -64,7 +56,6 @@ def report_execption(chatbot, history, a, b):
def 解析源代码(file_manifest, project_folder, top_p, temperature, chatbot, history, systemPromptTxt):
import time, glob, os
from predict import predict_no_ui
print('begin analysis on:', file_manifest)
for index, fp in enumerate(file_manifest):
with open(fp, 'r', encoding='utf-8') as f:
@ -73,7 +64,7 @@ def 解析源代码(file_manifest, project_folder, top_p, temperature, chatbot,
前言 = "接下来请你逐文件分析下面的工程" if index==0 else ""
i_say = 前言 + f'请对下面的程序文件做一个概述文件名是{os.path.relpath(fp, project_folder)}，文件代码是 ```{file_content}```'
i_say_show_user = 前言 + f'[{index}/{len(file_manifest)}] 请对下面的程序文件做一个概述: {os.path.abspath(fp)}'
chatbot.append((i_say_show_user, "[local] waiting gpt response."))
chatbot.append((i_say_show_user, "[waiting gpt response]"))
print('[1] yield chatbot, history')
yield chatbot, history, '正常'
@ -98,7 +89,7 @@ def 解析源代码(file_manifest, project_folder, top_p, temperature, chatbot,
all_file = ', '.join([os.path.relpath(fp, project_folder) for index, fp in enumerate(file_manifest)])
i_say = f'根据以上你自己的分析，对程序的整体功能和构架做出概括。然后用一张markdown表格整理每个文件的功能（包括{all_file}）。'
chatbot.append((i_say, "[local] waiting gpt response."))
chatbot.append((i_say, "[waiting gpt response]"))
yield chatbot, history, '正常'
if not fast_debug:
@ -159,22 +150,22 @@ def 解析一个C项目的头文件(txt, top_p, temperature, chatbot, history, s
def get_crazy_functionals():
return {
"程序解构简单案例": {
"Color": "stop", # 按钮颜色
"Function": 自我程序解构简单案例
},
"请解析并解构此项目本身": {
"[实验功能] 请解析并解构此项目本身": {
"Color": "stop", # 按钮颜色
"Function": 解析项目本身
},
"解析一整个Python项目（输入栏给定项目完整目录）": {
"[实验功能] 解析一整个Python项目（输入栏给定项目完整目录）": {
"Color": "stop", # 按钮颜色
"Function": 解析一个Python项目
},
"解析一整个C++项目的头文件（输入栏给定项目完整目录）": {
"[实验功能] 解析一整个C++项目的头文件（输入栏给定项目完整目录）": {
"Color": "stop", # 按钮颜色
"Function": 解析一个C项目的头文件
},
"[实验功能] 高阶功能模板函数": {
"Color": "stop", # 按钮颜色
"Function": 高阶功能模板函数
},
}

2
main.py
View File

@ -106,7 +106,7 @@ with gr.Blocks() as demo:
# submitBtn.click(reset_textbox, [], [txt])
for k in functional:
functional[k]["Button"].click(predict,
[txt, top_p, temperature, chatbot, history, systemPromptTxt, FALSE, TRUE, gr.State(k)], [chatbot, history, statusDisplay], show_progress=True)
[txt, top_p, temperature, chatbot, history, systemPromptTxt, TRUE, gr.State(k)], [chatbot, history, statusDisplay], show_progress=True)
for k in crazy_functional:
crazy_functional[k]["Button"].click(crazy_functional[k]["Function"],
[txt, top_p, temperature, chatbot, history, systemPromptTxt, gr.State(PORT)], [chatbot, history, statusDisplay])

182
predict.py
View File

@ -6,11 +6,12 @@ import logging
import traceback
import requests
import importlib
from colorful import *
# config_private.py放自己的秘密如API和代理网址
# 读取时首先看是否存在私密的config_private配置文件（不受git管控），如果有，则覆盖原config文件
try: from config_private import proxies, API_URL, API_KEY, TIMEOUT_SECONDS, MAX_RETRY
except: from config import proxies, API_URL, API_KEY, TIMEOUT_SECONDS, MAX_RETRY
try: from config_private import proxies, API_URL, API_KEY, TIMEOUT_SECONDS, MAX_RETRY, LLM_MODEL
except: from config import proxies, API_URL, API_KEY, TIMEOUT_SECONDS, MAX_RETRY, LLM_MODEL
timeout_bot_msg = '[local] Request timeout, network error. please check proxy settings in config.py.'
@ -23,51 +24,12 @@ def get_full_error(chunk, stream_response):
return chunk
def predict_no_ui(inputs, top_p, temperature, history=[]):
messages = [{"role": "system", "content": ""}]
#
chat_counter = len(history) // 2
if chat_counter > 0:
for index in range(0, 2*chat_counter, 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"] != "":
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)
# messages
payload = {
"model": "gpt-3.5-turbo",
# "model": "gpt-4",
"messages": messages,
"temperature": temperature, # 1.0,
"top_p": top_p, # 1.0,
"n": 1,
"stream": False,
"presence_penalty": 0,
"frequency_penalty": 0,
}
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {API_KEY}"
}
headers, payload = generate_payload(inputs, top_p, temperature, history, system_prompt="", stream=False)
retry = 0
while True:
try:
# make a POST request to the API endpoint using the requests.post method, passing in stream=True
# make a POST request to the API endpoint, stream=False
response = requests.post(API_URL, headers=headers, proxies=proxies,
json=payload, stream=False, timeout=TIMEOUT_SECONDS*2); break
except TimeoutError as e:
@ -84,9 +46,7 @@ def predict_no_ui(inputs, top_p, temperature, history=[]):
raise ConnectionAbortedError("Json解析不合常规，可能是文本过长" + response.text)
def predict(inputs, top_p, temperature, chatbot=[], history=[], system_prompt='', retry=False,
def predict(inputs, top_p, temperature, chatbot=[], history=[], system_prompt='',
stream = True, additional_fn=None):
if additional_fn is not None:
@ -101,60 +61,13 @@ def predict(inputs, top_p, temperature, chatbot=[], history=[], system_prompt=''
chatbot.append((inputs, ""))
yield chatbot, history, "等待响应"
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {API_KEY}"
}
chat_counter = len(history) // 2
print(f"chat_counter - {chat_counter}")
messages = [{"role": "system", "content": system_prompt}]
if chat_counter:
for index in range(0, 2*chat_counter, 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 not (what_gpt_answer["content"] != "" or retry): continue
if what_gpt_answer["content"] == timeout_bot_msg: continue
messages.append(what_i_have_asked)
messages.append(what_gpt_answer)
else:
messages[-1]['content'] = what_gpt_answer['content']
if retry and chat_counter:
messages.pop()
else:
what_i_ask_now = {}
what_i_ask_now["role"] = "user"
what_i_ask_now["content"] = inputs
messages.append(what_i_ask_now)
chat_counter += 1
# messages
payload = {
"model": "gpt-3.5-turbo",
# "model": "gpt-4",
"messages": messages,
"temperature": temperature, # 1.0,
"top_p": top_p, # 1.0,
"n": 1,
"stream": stream,
"presence_penalty": 0,
"frequency_penalty": 0,
}
history.append(inputs)
headers, payload = generate_payload(inputs, top_p, temperature, history, system_prompt, stream)
history.append(inputs); history.append(" ")
retry = 0
while True:
try:
# make a POST request to the API endpoint using the requests.post method, passing in stream=True
# make a POST request to the API endpoint, stream=True
response = requests.post(API_URL, headers=headers, proxies=proxies,
json=payload, stream=True, timeout=TIMEOUT_SECONDS);break
except:
@ -164,37 +77,30 @@ def predict(inputs, top_p, temperature, chatbot=[], history=[], system_prompt=''
yield chatbot, history, "请求超时"+retry_msg
if retry > MAX_RETRY: raise TimeoutError
token_counter = 0
partial_words = ""
counter = 0
gpt_replying_buffer = ""
is_head_of_the_stream = True
if stream:
stream_response = response.iter_lines()
while True:
chunk = next(stream_response)
if chunk == b'data: [DONE]':
break
if counter == 0:
counter += 1
continue
counter += 1
# check whether each line is non-empty
# print(chunk.decode()[6:])
if is_head_of_the_stream:
is_head_of_the_stream = False; continue
if chunk:
# decode each line as response data is in bytes
try:
if len(json.loads(chunk.decode()[6:])['choices'][0]["delta"]) == 0:
logging.info(f'[response] {chatbot[-1][-1]}')
# 判定为数据流的结束，gpt_replying_buffer也写完了
logging.info(f'[response] {gpt_replying_buffer}')
break
# 处理数据流的主体
chunkjson = json.loads(chunk.decode()[6:])
status_text = f"finish_reason: {chunkjson['choices'][0]['finish_reason']}"
partial_words = partial_words + json.loads(chunk.decode()[6:])['choices'][0]["delta"]["content"]
if token_counter == 0:
history.append(" " + partial_words)
else:
history[-1] = partial_words
# 如果这里抛出异常，一般是文本过长，详情见get_full_error的输出
gpt_replying_buffer = gpt_replying_buffer + json.loads(chunk.decode()[6:])['choices'][0]["delta"]["content"]
history[-1] = gpt_replying_buffer
chatbot[-1] = (history[-2], history[-1])
token_counter += 1
yield chatbot, history, status_text
except Exception as e:
@ -207,4 +113,48 @@ def predict(inputs, top_p, temperature, chatbot=[], history=[], system_prompt=''
yield chatbot, history, "Json解析不合常规，很可能是文本过长" + error_msg
return
def generate_payload(inputs, top_p, temperature, history, system_prompt, stream):
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {API_KEY}"
}
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
if what_gpt_answer["content"] == timeout_bot_msg: 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_MODEL,
"messages": messages,
"temperature": temperature, # 1.0,
"top_p": top_p, # 1.0,
"n": 1,
"stream": stream,
"presence_penalty": 0,
"frequency_penalty": 0,
}
print(f" {LLM_MODEL} : {conversation_cnt} : {inputs}")
return headers,payload