Merge branch 'master' into huggingface

README.md

@@ -1,3 +1,15 @@
+---
+title: ChatImprovement
+emoji: 😻
+colorFrom: blue
+colorTo: blue
+sdk: gradio
+sdk_version: 3.32.0
+app_file: app.py
+pinned: false
+---
+
+# ChatGPT 学术优化
 > **Note**
 >
 > 2023.5.27 对Gradio依赖进行了调整，Fork并解决了官方Gradio的若干Bugs。请及时**更新代码**并重新更新pip依赖。安装依赖时，请严格选择`requirements.txt`中**指定的版本**： 

app.py

@@ -1,8 +1,10 @@
 import os; os.environ['no_proxy'] = '*' # 避免代理网络产生意外污染
 
 def main():
+    import subprocess, sys
+    subprocess.check_call([sys.executable, '-m', 'pip', 'install', 'gradio-stable-fork'])
     import gradio as gr
-    if gr.__version__ not in ['3.28.3','3.32.2']: assert False, "需要特殊依赖，请务必用 pip install -r requirements.txt 指令安装依赖，详情信息见requirements.txt"
+    if gr.__version__ not in ['3.28.3','3.32.3']: assert False, "请用 pip install -r requirements.txt 安装依赖"
     from request_llm.bridge_all import predict
     from toolbox import format_io, find_free_port, on_file_uploaded, on_report_generated, get_conf, ArgsGeneralWrapper, DummyWith
     # 建议您复制一个config_private.py放自己的秘密, 如API和代理网址, 避免不小心传github被别人看到
@@ -54,6 +56,7 @@ def main():
     cancel_handles = []
     with gr.Blocks(title="ChatGPT 学术优化", theme=set_theme, analytics_enabled=False, css=advanced_css) as demo:
         gr.HTML(title_html)
+        gr.HTML('''<center><a href="https://huggingface.co/spaces/qingxu98/gpt-academic?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>请您打开此页面后务必点击上方的“复制空间”（Duplicate Space）按钮！<font color="#FF00FF">使用时，先在输入框填入API-KEY然后回车。</font><br/>切忌在“复制空间”（Duplicate Space）之前填入API_KEY或进行提问，否则您的API_KEY将极可能被空间所有者攫取！<br/>支持任意数量的OpenAI的密钥和API2D的密钥共存，例如输入"OpenAI密钥1,API2D密钥2"，然后提交，即可同时使用两种模型接口。</center>''')
         cookies = gr.State({'api_key': API_KEY, 'llm_model': LLM_MODEL})
         with gr_L1():
             with gr_L2(scale=2):
@@ -63,7 +66,7 @@ def main():
             with gr_L2(scale=1):
                 with gr.Accordion("输入区", open=True) as area_input_primary:
                     with gr.Row():
-                        txt = gr.Textbox(show_label=False, placeholder="Input question here.").style(container=False)
+                        txt = gr.Textbox(show_label=False, lines=2, placeholder="输入问题或API密钥，输入多个密钥时，用英文逗号间隔。支持OpenAI密钥和API2D密钥共存。").style(container=False)
                     with gr.Row():
                         submitBtn = gr.Button("提交", variant="primary")
                     with gr.Row():
@@ -197,10 +200,7 @@ def main():
         threading.Thread(target=warm_up_modules, name="warm-up", daemon=True).start()
 
     auto_opentab_delay()
-    demo.queue(concurrency_count=CONCURRENT_COUNT).launch(
-        server_name="0.0.0.0", server_port=PORT,
-        favicon_path="docs/logo.png", auth=AUTHENTICATION,
-        blocked_paths=["config.py","config_private.py","docker-compose.yml","Dockerfile"])
+    demo.queue(concurrency_count=CONCURRENT_COUNT).launch(server_name="0.0.0.0", share=False, favicon_path="docs/logo.png", blocked_paths=["config.py","config_private.py","docker-compose.yml","Dockerfile"])
 
     # 如果需要在二级路径下运行
     # CUSTOM_PATH, = get_conf('CUSTOM_PATH')

config.py

@@ -45,10 +45,9 @@ 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
 
-# 模型选择是 (注意: LLM_MODEL是默认选中的模型, 同时它必须被包含在AVAIL_LLM_MODELS切换列表中 )
-LLM_MODEL = "gpt-3.5-turbo" # 可选 ↓↓↓
-AVAIL_LLM_MODELS = ["gpt-3.5-turbo-16k", "gpt-3.5-turbo", "azure-gpt35", "api2d-gpt-3.5-turbo", "gpt-4", "api2d-gpt-4", "chatglm", "moss", "newbing", "newbing-free", "stack-claude"]
-# P.S. 其他可用的模型还包括 ["gpt-3.5-turbo-0613", "gpt-3.5-turbo-16k-0613", "newbing-free", "jittorllms_rwkv", "jittorllms_pangualpha", "jittorllms_llama"]
+# OpenAI模型选择是（gpt4现在只对申请成功的人开放）
+LLM_MODEL = "gpt-3.5-turbo" # 可选 "chatglm"
+AVAIL_LLM_MODELS = ["newbing-free", "gpt-3.5-turbo", "gpt-4", "api2d-gpt-4", "api2d-gpt-3.5-turbo"]
 
 # 本地LLM模型如ChatGLM的执行方式 CPU/GPU
 LOCAL_MODEL_DEVICE = "cpu" # 可选 "cuda"

crazy_functions/latex_utils.py

@@ -27,6 +27,24 @@ def set_forbidden_text(text, mask, pattern, flags=0):
         mask[res.span()[0]:res.span()[1]] = PRESERVE
     return text, mask
 
+def reverse_forbidden_text(text, mask, pattern, flags=0, forbid_wrapper=True):
+    """
+    Move area out of preserve area (make text editable for GPT)
+    count the number of the braces so as to catch compelete text area. 
+    e.g.
+    \begin{abstract} blablablablablabla. \end{abstract} 
+    """
+    if isinstance(pattern, list): pattern = '|'.join(pattern)
+    pattern_compile = re.compile(pattern, flags)
+    for res in pattern_compile.finditer(text):
+        if not forbid_wrapper:
+            mask[res.span()[0]:res.span()[1]] = TRANSFORM
+        else:
+            mask[res.regs[0][0]: res.regs[1][0]] = PRESERVE   # '\\begin{abstract}'
+            mask[res.regs[1][0]: res.regs[1][1]] = TRANSFORM   # abstract
+            mask[res.regs[1][1]: res.regs[0][1]] = PRESERVE   # abstract
+    return text, mask
+
 def set_forbidden_text_careful_brace(text, mask, pattern, flags=0):
     """
     Add a preserve text area in this paper (text become untouchable for GPT).
@@ -326,6 +344,7 @@ def split_subprocess(txt, project_folder, return_dict, opts):
     # reverse 操作必须放在最后
     text, mask = reverse_forbidden_text_careful_brace(text, mask, r"\\caption\{(.*?)\}", re.DOTALL, forbid_wrapper=True)
     text, mask = reverse_forbidden_text_careful_brace(text, mask, r"\\abstract\{(.*?)\}", re.DOTALL, forbid_wrapper=True)
+    text, mask = reverse_forbidden_text(text, mask, r"\\begin\{abstract\}(.*?)\\end\{abstract\}", re.DOTALL, forbid_wrapper=True)
     root = convert_to_linklist(text, mask)
 
     # 修复括号
@@ -672,10 +691,9 @@ def remove_buggy_lines(file_path, log_path, tex_name, tex_name_pure, n_fix, work
         print("Fatal error occurred, but we cannot identify error, please download zip, read latex log, and compile manually.")
         return False, -1, [-1]
     
-
-def compile_latex_with_timeout(command, timeout=60):
+def compile_latex_with_timeout(command, cwd, timeout=60):
     import subprocess
-    process = subprocess.Popen(command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    process = subprocess.Popen(command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, cwd=cwd)
     try:
         stdout, stderr = process.communicate(timeout=timeout)
     except subprocess.TimeoutExpired:
@@ -699,24 +717,24 @@ def 编译Latex(chatbot, history, main_file_original, main_file_modified, work_f
 
         # https://stackoverflow.com/questions/738755/dont-make-me-manually-abort-a-latex-compile-when-theres-an-error
         yield from update_ui_lastest_msg(f'尝试第 {n_fix}/{max_try} 次编译, 编译原始PDF ...', chatbot, history)   # 刷新Gradio前端界面
-        os.chdir(work_folder_original); ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_original}.tex'); os.chdir(current_dir)
+        ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_original}.tex', work_folder_original)
 
         yield from update_ui_lastest_msg(f'尝试第 {n_fix}/{max_try} 次编译, 编译转化后的PDF ...', chatbot, history)   # 刷新Gradio前端界面
-        os.chdir(work_folder_modified); ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_modified}.tex'); os.chdir(current_dir)
+        ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_modified}.tex', work_folder_modified)
         
         if ok and os.path.exists(pj(work_folder_modified, f'{main_file_modified}.pdf')):
             # 只有第二步成功，才能继续下面的步骤
             yield from update_ui_lastest_msg(f'尝试第 {n_fix}/{max_try} 次编译, 编译BibTex ...', chatbot, history)    # 刷新Gradio前端界面
             if not os.path.exists(pj(work_folder_original, f'{main_file_original}.bbl')):
-                os.chdir(work_folder_original); ok = compile_latex_with_timeout(f'bibtex  {main_file_original}.aux'); os.chdir(current_dir)
+                ok = compile_latex_with_timeout(f'bibtex  {main_file_original}.aux', work_folder_original)
             if not os.path.exists(pj(work_folder_modified, f'{main_file_modified}.bbl')):
-                os.chdir(work_folder_modified); ok = compile_latex_with_timeout(f'bibtex  {main_file_modified}.aux'); os.chdir(current_dir)
+                ok = compile_latex_with_timeout(f'bibtex  {main_file_modified}.aux', work_folder_modified)
 
             yield from update_ui_lastest_msg(f'尝试第 {n_fix}/{max_try} 次编译, 编译文献交叉引用 ...', chatbot, history)  # 刷新Gradio前端界面
-            os.chdir(work_folder_original); ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_original}.tex'); os.chdir(current_dir)
-            os.chdir(work_folder_modified); ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_modified}.tex'); os.chdir(current_dir)
-            os.chdir(work_folder_original); ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_original}.tex'); os.chdir(current_dir)
-            os.chdir(work_folder_modified); ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_modified}.tex'); os.chdir(current_dir)
+            ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_original}.tex', work_folder_original)
+            ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_modified}.tex', work_folder_modified)
+            ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_original}.tex', work_folder_original)
+            ok = compile_latex_with_timeout(f'pdflatex -interaction=batchmode -file-line-error {main_file_modified}.tex', work_folder_modified)
 
             if mode!='translate_zh':
                 yield from update_ui_lastest_msg(f'尝试第 {n_fix}/{max_try} 次编译, 使用latexdiff生成论文转化前后对比 ...', chatbot, history) # 刷新Gradio前端界面
@@ -724,13 +742,11 @@ def 编译Latex(chatbot, history, main_file_original, main_file_modified, work_f
                 ok = compile_latex_with_timeout(f'latexdiff --encoding=utf8 --append-safecmd=subfile {work_folder_original}/{main_file_original}.tex  {work_folder_modified}/{main_file_modified}.tex --flatten > {work_folder}/merge_diff.tex')
 
                 yield from update_ui_lastest_msg(f'尝试第 {n_fix}/{max_try} 次编译, 正在编译对比PDF ...', chatbot, history)   # 刷新Gradio前端界面
-                os.chdir(work_folder); ok = compile_latex_with_timeout(f'pdflatex  -interaction=batchmode -file-line-error merge_diff.tex'); os.chdir(current_dir)
-                os.chdir(work_folder); ok = compile_latex_with_timeout(f'bibtex    merge_diff.aux'); os.chdir(current_dir)
-                os.chdir(work_folder); ok = compile_latex_with_timeout(f'pdflatex  -interaction=batchmode -file-line-error merge_diff.tex'); os.chdir(current_dir)
-                os.chdir(work_folder); ok = compile_latex_with_timeout(f'pdflatex  -interaction=batchmode -file-line-error merge_diff.tex'); os.chdir(current_dir)
+                ok = compile_latex_with_timeout(f'pdflatex  -interaction=batchmode -file-line-error merge_diff.tex', work_folder)
+                ok = compile_latex_with_timeout(f'bibtex    merge_diff.aux', work_folder)
+                ok = compile_latex_with_timeout(f'pdflatex  -interaction=batchmode -file-line-error merge_diff.tex', work_folder)
+                ok = compile_latex_with_timeout(f'pdflatex  -interaction=batchmode -file-line-error merge_diff.tex', work_folder)
 
-        # <--------------------->
-        os.chdir(current_dir)
 
         # <---------- 检查结果 ----------->
         results_ = ""
@@ -766,7 +782,6 @@ def 编译Latex(chatbot, history, main_file_original, main_file_modified, work_f
             yield from update_ui_lastest_msg(f'由于最为关键的转化PDF编译失败, 将根据报错信息修正tex源文件并重试, 当前报错的latex代码处于第{buggy_lines}行 ...', chatbot, history)   # 刷新Gradio前端界面
             if not can_retry: break
 
-    os.chdir(current_dir)
     return False # 失败啦
 
 

crazy_functions/询问多个大语言模型.py

@@ -13,11 +13,11 @@ def 同时问询(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt
     web_port        当前软件运行的端口号
     """
     history = []    # 清空历史，以免输入溢出
-    chatbot.append((txt, "正在同时咨询ChatGPT和ChatGLM……"))
+    chatbot.append((txt, "正在同时咨询gpt-3.5和gpt-4……"))
     yield from update_ui(chatbot=chatbot, history=history) # 刷新界面 # 由于请求gpt需要一段时间，我们先及时地做一次界面更新
 
     # llm_kwargs['llm_model'] = 'chatglm&gpt-3.5-turbo&api2d-gpt-3.5-turbo' # 支持任意数量的llm接口，用&符号分隔
-    llm_kwargs['llm_model'] = 'chatglm&gpt-3.5-turbo' # 支持任意数量的llm接口，用&符号分隔
+    llm_kwargs['llm_model'] = 'gpt-3.5-turbo&gpt-4' # 支持任意数量的llm接口，用&符号分隔
     gpt_say = yield from request_gpt_model_in_new_thread_with_ui_alive(
         inputs=txt, inputs_show_user=txt, 
         llm_kwargs=llm_kwargs, chatbot=chatbot, history=history, 

crazy_functions/谷歌检索小助手.py

@@ -104,7 +104,7 @@ def 谷歌检索小助手(txt, llm_kwargs, plugin_kwargs, chatbot, history, syst
             meta_paper_info_list = meta_paper_info_list[batchsize:]
 
     chatbot.append(["状态？", 
-        "已经全部完成，您可以试试让AI写一个Related Works，例如您可以继续输入Write a \"Related Works\" section about \"你搜索的研究领域\" for me."])
+        "已经全部完成，您可以试试让AI写一个Related Works，例如您可以继续输入Write an academic \"Related Works\" section about \"你搜索的研究领域\" for me."])
     msg = '正常'
     yield from update_ui(chatbot=chatbot, history=history, msg=msg) # 刷新界面
     res = write_results_to_file(history)

request_llm/bridge_all.py

@@ -152,7 +152,7 @@ model_info = {
         "token_cnt": get_token_num_gpt4,
     },
 
-    # chatglm
+    # 将 chatglm 直接对齐到 chatglm2
     "chatglm": {
         "fn_with_ui": chatglm_ui,
         "fn_without_ui": chatglm_noui,
@@ -161,6 +161,15 @@ model_info = {
         "tokenizer": tokenizer_gpt35,
         "token_cnt": get_token_num_gpt35,
     },
+    "chatglm2": {
+        "fn_with_ui": chatglm_ui,
+        "fn_without_ui": chatglm_noui,
+        "endpoint": None,
+        "max_token": 1024,
+        "tokenizer": tokenizer_gpt35,
+        "token_cnt": get_token_num_gpt35,
+    },
+    
     # newbing
     "newbing": {
         "fn_with_ui": newbing_ui,

request_llm/bridge_chatglm.py

@@ -40,12 +40,12 @@ class GetGLMHandle(Process):
         while True:
             try:
                 if self.chatglm_model is None:
-                    self.chatglm_tokenizer = AutoTokenizer.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True)
+                    self.chatglm_tokenizer = AutoTokenizer.from_pretrained("THUDM/chatglm2-6b", trust_remote_code=True)
                     device, = get_conf('LOCAL_MODEL_DEVICE')
                     if device=='cpu':
-                        self.chatglm_model = AutoModel.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True).float()
+                        self.chatglm_model = AutoModel.from_pretrained("THUDM/chatglm2-6b", trust_remote_code=True).float()
                     else:
-                        self.chatglm_model = AutoModel.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True).half().cuda()
+                        self.chatglm_model = AutoModel.from_pretrained("THUDM/chatglm2-6b", trust_remote_code=True).half().cuda()
                     self.chatglm_model = self.chatglm_model.eval()
                     break
                 else:

requirements.txt

@@ -1,4 +1,3 @@
-./docs/gradio-3.32.2-py3-none-any.whl
 tiktoken>=0.3.3
 requests[socks]
 transformers
@@ -15,4 +14,4 @@ pymupdf
 openai
 numpy
 arxiv
-rich
+rich

toolbox.py

@@ -842,4 +842,4 @@ def objload(file='objdump.tmp'):
         return
     with open(file, 'rb') as f:
         return pickle.load(f)
-    
+