Llamaindex RAG实践

原始模型

回答存在严重幻觉

使用llama index构建RAG系统

必须在30%的服务器权限下才可以正常执行，否则显存会爆
代码

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Settings


from llama_index.embeddings.huggingface import HuggingFaceEmbedding


from llama_index.llms.huggingface import HuggingFaceLLM


 
embed_model = HuggingFaceEmbedding(
    model_name="/root/sentence-transformer"
)
 
Settings.embed_model = embed_model
 
llm = HuggingFaceLLM(
    model_name="/root/internlm2-chat-1_8b",
    tokenizer_name="/root/internlm2-chat-1_8b",
    model_kwargs={"trust_remote_code":True},
    tokenizer_kwargs={"trust_remote_code":True}
)
Settings.llm = llm
 
documents = SimpleDirectoryReader("/root/demo/data").load_data()
index = VectorStoreIndex.from_documents(documents)
query_engine = index.as_query_engine()
response = query_engine.query("在clickhouse中，alter命令支持什么表引擎？")
 
print(response)

可正常回答