search

OpenSPG / KAG

KAG is a logical form-guided reasoning and retrieval framework based on OpenSPG engine and LLMs. It is used to build logical reasoning and factual Q&A solutions for professional domain knowledge bases. It can effectively overcome the shortcomings of the traditional RAG vector similarity calculation model.
https://spg.openkg.cn/en-US
Apache License 2.0
625 stars 46 forks source link

invalid vectorizer config: model not found #34

Open xsg5886779 opened 2 weeks ago

xsg5886779 commented 2 weeks ago

<class 'RuntimeError'>: invalid vectorizer config: model not found at '/var/www/models/bge-m3', nor model url specified

配置文件： namespace = KagDemo host_addr = http://localhost:8887 id = 1

[vectorizer] vectorizer = kag.common.vectorizer.LocalBGEVectorizer model = /var/www/models/bge-m3 vector_dimensions = 768

[llm] client_type = vllm base_url = http://192.168.100.86:8000/v1/chat/completions api_key = put your deepseek api key her model = qwen25-7B-chat

[prompt] biz_scene = default language = zh

model = /var/www/models/bge-m3 这个是存在的 model = qwen25-7B-chat 也是存在的，但是就是创建不了知识库

xionghuaidong commented 2 weeks ago

model not found at '/var/www/models/bge-m3'

'/var/www/models/bge-m3' is not a directory in the release-openspg-server container.

Only '~/.cache/vectorizer/BAAI/bge-base-zh-v1.5' exists in the release-openspg-server container.

By the way, the key model is incorrect, it should be path. And the bge-m3 model has vector_dimensions = 1024.

If you want to use the bge-m3 model like the built-in bge-base-zh-v1.5 model. You should copy the bge-m3 models files into the release-openspg-server container to a location like ~/.cache/vectorizer/BAAI/bge-m3, then use the following config.

[vectorizer]
vectorizer = kag.common.vectorizer.LocalBGEM3Vectorizer
path = ~/.cache/vectorizer/BAAI/bge-m3
vector_dimensions = 1024