binary-husky / binary-husky/gpt_academic
[Bug]: 知识库插件不可用,多处报错
- Dominant language
- Python
- Stars
- 71.3k
- Forks
- 8.3k
- PR merge metrics
- No merged PRs in 30d
Description
### Installation Method | 安装方法与平台
Anaconda (I used latest requirements.txt)
### Version | 版本
Latest | 最新版
### OS | 操作系统
Windows
### Describe the bug | 简述
调用知识库插件时出现多处报错,且过久未更新。翻阅issues中有诸多相似问题,猜测是遇到了同样错误。我这里找到了一部分,尚未完全解决,还有待作者看这篇issue
知识库问答.py 64行from langchain.embeddings.huggingface import HuggingFaceEmbeddings 过时
知识库问答.py 65-66行 with ProxyNetworkActivate('Download_LLM'):
HuggingFaceEmbeddings(model_name="GanymedeNil/text2vec-large-chinese") 报错,No sentence-transformers model found with name GanymedeNil/text2vec-large-chinese. Creating a new one with mean pooling. 不知是不是huggingface调用时出问题。
vector_fns/vector_database.py 12行from langchain.vectorstores import FAISS 过时
vector_fns/general_file_loader.py 中from langchain.document_loaders import UnstructuredFileLoader 过时
以下更新后仍无法在普通anaconda命令行操作,必须在管理员权限下运行,且只能训练txt文本,训练markdown会卡住。
而且在训练txt文本后,使用“知识库文件注入”功能(这里名字应该也有错误,函数文件中原应该是“读取知识库作答”)仍有错误。
提示在vector_fns/vector_database.py 220行self.vector_store = FAISS.load_local(vs_path, text2vec)中添加参数allow_dangerous_deserialization=True。
在添加此参数后,重新运行程序,则训练文本又会卡住。
整个知识库函数均不可用,望作者mark并修复。
### Screen Shot | 有帮助的截图
暂无截图,由于报错点过多,没有在修改时逐个保存报错代码截图,作者可以自行运行知识库函数并查看。
### Terminal Traceback & Material to Help Reproduce Bugs | 终端traceback(如有) + 帮助我们复现的测试材料样本(如有)
见简述
Contributor guide
No contributing guide indexed for this repository
Research direction
Start by reproducing the knowledge-base plugin on Windows with the reported Anaconda setup, using both TXT and Markdown inputs. Read 知识库问答.py lines 64-66, vector_fns/vector_database.py lines 12 and 220, and vector_fns/general_file_loader.py; done means the plugin can train and answer from the knowledge base without the reported errors or hangs.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- huggingface, python
- Domain
- backend, machine-learning
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 15/100