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

Merge branch 'frontier' of github.com:binary-husky/chatgpt_academic into frontier

Browse Source
This commit is contained in:
binary-husky 2023-09-23 23:47:42 +08:00
commit f0a5c49a9c
15 changed files with 46 additions and 24 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
config.py
View File

@ -80,7 +80,7 @@ DEFAULT_FN_GROUPS = ['对话', '编程', '学术', '智能体']
# 模型选择是 (注意: LLM_MODEL是默认选中的模型, 它*必须*被包含在AVAIL_LLM_MODELS列表中 )
LLM_MODEL = "gpt-3.5-turbo" # 可选 ↓↓↓
AVAIL_LLM_MODELS = ["gpt-3.5-turbo-16k", "gpt-3.5-turbo", "azure-gpt-3.5", "api2d-gpt-3.5-turbo",
"gpt-4", "gpt-4-32k", "api2d-gpt-4", "chatglm", "moss", "newbing", "stack-claude"]
"gpt-4", "gpt-4-32k", "azure-gpt-4", "api2d-gpt-4", "chatglm", "moss", "newbing", "stack-claude"]
# P.S. 其他可用的模型还包括 ["qianfan", "llama2", "qwen", "gpt-3.5-turbo-0613", "gpt-3.5-turbo-16k-0613",
# "spark", "sparkv2", "chatglm_onnx", "claude-1-100k", "claude-2", "internlm", "jittorllms_pangualpha", "jittorllms_llama"]
@ -183,6 +183,9 @@ ALLOW_RESET_CONFIG = False
PATH_PRIVATE_UPLOAD = "private_upload"
# 日志文件夹的位置，请勿修改
PATH_LOGGING = "gpt_log"
# 除了连接OpenAI之外，还有哪些场合允许使用代理，请勿修改
WHEN_TO_USE_PROXY = ["Download_LLM", "Download_Gradio_Theme"]
"""
在线大模型配置关联关系示意图

4
crazy_functions/Langchain知识库.py
View File

@ -53,14 +53,14 @@ def 知识库问答(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_pro
yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
print('Checking Text2vec ...')
from langchain.embeddings.huggingface import HuggingFaceEmbeddings
with ProxyNetworkActivate(): # 临时地激活代理网络
with ProxyNetworkActivate('Download_LLM'): # 临时地激活代理网络
HuggingFaceEmbeddings(model_name="GanymedeNil/text2vec-large-chinese")
# < -------------------构建知识库--------------- >
chatbot.append(['<br/>'.join(file_manifest), "正在构建知识库..."])
yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
print('Establishing knowledge archive ...')
with ProxyNetworkActivate(): # 临时地激活代理网络
with ProxyNetworkActivate('Download_LLM'): # 临时地激活代理网络
kai = knowledge_archive_interface()
kai.feed_archive(file_manifest=file_manifest, id=kai_id)
kai_files = kai.get_loaded_file()

2
crazy_functions/crazy_utils.py
View File

@ -651,7 +651,7 @@ class knowledge_archive_interface():
from toolbox import ProxyNetworkActivate
print('Checking Text2vec ...')
from langchain.embeddings.huggingface import HuggingFaceEmbeddings
with ProxyNetworkActivate(): # 临时地激活代理网络
with ProxyNetworkActivate('Download_LLM'): # 临时地激活代理网络
self.text2vec_large_chinese = HuggingFaceEmbeddings(model_name="GanymedeNil/text2vec-large-chinese")
return self.text2vec_large_chinese

1
main.py
View File

@ -285,6 +285,7 @@ def main():
auto_opentab_delay()
demo.queue(concurrency_count=CONCURRENT_COUNT).launch(
quiet=True,
server_name="0.0.0.0",
server_port=PORT,
favicon_path="docs/logo.png",

9
request_llm/bridge_all.py
View File

@ -145,6 +145,15 @@ model_info = {
"token_cnt": get_token_num_gpt35,
},
"azure-gpt-4":{
"fn_with_ui": chatgpt_ui,
"fn_without_ui": chatgpt_noui,
"endpoint": azure_endpoint,
"max_token": 8192,
"tokenizer": tokenizer_gpt35,
"token_cnt": get_token_num_gpt35,
},
# api_2d
"api2d-gpt-3.5-turbo": {
"fn_with_ui": chatgpt_ui,

21
request_llm/bridge_chatglm.py
View File

@ -3,7 +3,7 @@ from transformers import AutoModel, AutoTokenizer
import time
import threading
import importlib
from toolbox import update_ui, get_conf
from toolbox import update_ui, get_conf, ProxyNetworkActivate
from multiprocessing import Process, Pipe
load_message = "ChatGLM尚未加载，加载需要一段时间。注意，取决于`config.py`的配置，ChatGLM消耗大量的内存（CPU）或显存（GPU），也许会导致低配计算机卡死 ……"
@ -48,16 +48,17 @@ class GetGLMHandle(Process):
while True:
try:
if self.chatglm_model is None:
self.chatglm_tokenizer = AutoTokenizer.from_pretrained(_model_name_, trust_remote_code=True)
if device=='cpu':
self.chatglm_model = AutoModel.from_pretrained(_model_name_, trust_remote_code=True).float()
with ProxyNetworkActivate('Download_LLM'):
if self.chatglm_model is None:
self.chatglm_tokenizer = AutoTokenizer.from_pretrained(_model_name_, trust_remote_code=True)
if device=='cpu':
self.chatglm_model = AutoModel.from_pretrained(_model_name_, trust_remote_code=True).float()
else:
self.chatglm_model = AutoModel.from_pretrained(_model_name_, trust_remote_code=True).half().cuda()
self.chatglm_model = self.chatglm_model.eval()
break
else:
self.chatglm_model = AutoModel.from_pretrained(_model_name_, trust_remote_code=True).half().cuda()
self.chatglm_model = self.chatglm_model.eval()
break
else:
break
break
except:
retry += 1
if retry > 3:

2
request_llm/bridge_llama2.py
View File

@ -30,7 +30,7 @@ class GetONNXGLMHandle(LocalLLMHandle):
with open(os.path.expanduser('~/.cache/huggingface/token'), 'w') as f:
f.write(huggingface_token)
model_id = 'meta-llama/Llama-2-7b-chat-hf'
with ProxyNetworkActivate():
with ProxyNetworkActivate('Download_LLM'):
self._tokenizer = AutoTokenizer.from_pretrained(model_id, use_auth_token=huggingface_token)
# use fp16
model = AutoModelForCausalLM.from_pretrained(model_id, use_auth_token=huggingface_token).eval()

1
request_llm/requirements_chatglm.txt
View File

@ -1,5 +1,4 @@
protobuf
transformers>=4.27.1
cpm_kernels
torch>=1.10
mdtex2html

1
request_llm/requirements_chatglm_onnx.txt
View File

@ -1,5 +1,4 @@
protobuf
transformers>=4.27.1
cpm_kernels
torch>=1.10
mdtex2html

1
request_llm/requirements_jittorllms.txt
View File

@ -2,6 +2,5 @@ jittor >= 1.3.7.9
jtorch >= 0.1.3
torch
torchvision
transformers==4.26.1
pandas
jieba

1
request_llm/requirements_moss.txt
View File

@ -1,5 +1,4 @@
torch
transformers==4.25.1
sentencepiece
datasets
accelerate

2
requirements.txt
View File

@ -2,7 +2,7 @@
pydantic==1.10.11
tiktoken>=0.3.3
requests[socks]
transformers
transformers>=4.27.1
python-markdown-math
beautifulsoup4
prompt_toolkit

4
themes/gradios.py
View File

@ -5,7 +5,7 @@ CODE_HIGHLIGHT, ADD_WAIFU, LAYOUT = get_conf('CODE_HIGHLIGHT', 'ADD_WAIFU', 'LAY
def dynamic_set_theme(THEME):
set_theme = gr.themes.ThemeClass()
with ProxyNetworkActivate():
with ProxyNetworkActivate('Download_Gradio_Theme'):
logging.info('正在下载Gradio主题，请稍等。')
if THEME.startswith('Huggingface-'): THEME = THEME.lstrip('Huggingface-')
if THEME.startswith('huggingface-'): THEME = THEME.lstrip('huggingface-')
@ -16,7 +16,7 @@ def adjust_theme():
try:
set_theme = gr.themes.ThemeClass()
with ProxyNetworkActivate():
with ProxyNetworkActivate('Download_Gradio_Theme'):
logging.info('正在下载Gradio主题，请稍等。')
THEME, = get_conf('THEME')
if THEME.startswith('Huggingface-'): THEME = THEME.lstrip('Huggingface-')

12
toolbox.py
View File

@ -956,7 +956,19 @@ class ProxyNetworkActivate():
"""
这段代码定义了一个名为TempProxy的空上下文管理器, 用于给一小段代码上代理
"""
def __init__(self, task=None) -> None:
self.task = task
if not task:
# 不给定task, 那么我们默认代理生效
self.valid = True
else:
# 给定了task, 我们检查一下
from toolbox import get_conf
WHEN_TO_USE_PROXY, = get_conf('WHEN_TO_USE_PROXY')
self.valid = (task in WHEN_TO_USE_PROXY)
def __enter__(self):
if not self.valid: return self
from toolbox import get_conf
proxies, = get_conf('proxies')
if 'no_proxy' in os.environ: os.environ.pop('no_proxy')

4
version
View File

@ -1,5 +1,5 @@
{
"version": 3.53,
"version": 3.54,
"show_feature": true,
"new_feature": "支持动态选择不同界面主题 <-> 提高稳定性&解决多用户冲突问题 <-> 支持插件分类和更多UI皮肤外观 <-> 支持用户使用自然语言调度各个插件(虚空终端) ！ <-> 改进UI，设计新主题 <-> 支持借助GROBID实现PDF高精度翻译 <-> 接入百度千帆平台和文心一言 <-> 接入阿里通义千问、讯飞星火、上海AI-Lab书生 <-> 优化一键升级 <-> 提高arxiv翻译速度和成功率"
"new_feature": "新增动态代码解释器（CodeInterpreter） <-> 增加文本回答复制按钮 <-> 细分代理场合 <-> 支持动态选择不同界面主题 <-> 提高稳定性&解决多用户冲突问题 <-> 支持插件分类和更多UI皮肤外观 <-> 支持用户使用自然语言调度各个插件(虚空终端) ！ <-> 改进UI，设计新主题 <-> 支持借助GROBID实现PDF高精度翻译 <-> 接入百度千帆平台和文心一言 <-> 接入阿里通义千问、讯飞星火、上海AI-Lab书生 <-> 优化一键升级 <-> 提高arxiv翻译速度和成功率"
}