复现

Step1：按照https://github.com/HKUDS/MiniRAG/blob/main/README_CN.md上的教程完成源码安装，这边建议安装稳定版本
Step2：main.py即可
如果直接使用的话，不横向对比的话，本质上调用main.py即可完成测试
注意解压数据文件
第一次跑出这个结果为零data是正常的，因为第一次跑还没构建向量数据库

INFO:nano-vectordb:Load (22, 384) data
INFO:nano-vectordb:Init {'embedding_dim': 384, 'metric': 'cosine', 'storage_file': './LiHua-World/vdb_entities.json'} 22 data
INFO:nano-vectordb:Load (22, 384) data
INFO:nano-vectordb:Init {'embedding_dim': 384, 'metric': 'cosine', 'storage_file': './LiHua-World/vdb_entities_name.json'} 22 data
INFO:nano-vectordb:Load (21, 384) data
INFO:nano-vectordb:Init {'embedding_dim': 384, 'metric': 'cosine', 'storage_file': './LiHua-World/vdb_relationships.json'} 21 data
INFO:nano-vectordb:Load (8, 384) data
INFO:nano-vectordb:Init {'embedding_dim': 384, 'metric': 'cosine', 'storage_file': './LiHua-World/vdb_chunks.json'} 8 data
INFO:minirag:Loaded document status storage with 0 records

复现时的一些小问题

问题1：

执行Step_0_index.py

Traceback (most recent call last):
  File "/MiniRAG/reproduce/Step_0_index.py", line 12, in <module>
    from minirag.llm import (
ImportError: cannot import name 'gpt_4o_mini_complete' from 'minirag.llm' (triton_project/MiniRAG/minirag/llm/__init__.py)

解决方法

# 在minirag.llm下的Init中去添加
# 以下两个是Step_0_index.py文件需要的
from .openai import gpt_4o_mini_complete
from .hf import hf_embed
# 以下是Step_1_QA.py文件需要的
from .hf import hf_model_complete

问题2：

直接稳定版本minirag0.0.2时候

Traceback (most recent call last):
  File "/triton_project/MiniRAG-0.0.2/main.py", line 6, in <module>
    from minirag import MiniRAG, QueryParam
  File "/triton_project/MiniRAG-0.0.2/minirag/__init__.py", line 1, in <module>
    from .minirag import MiniRAG as MiniRAG, QueryParam as QueryParam
  File "/triton_project/MiniRAG-0.0.2/minirag/minirag.py", line 9, in <module>
    from .operate import (
  File "/triton_project/MiniRAG-0.0.2/minirag/operate.py", line 26, in <module>
    from .base import (
  File "/triton_project/MiniRAG-0.0.2/minirag/base.py", line 104, in <module>
    class BaseGraphStorage(StorageNameSpace):
  File "/triton_project/MiniRAG-0.0.2/minirag/base.py", line 107, in BaseGraphStorage
    @abstractmethod
     ^^^^^^^^^^^^^^

解决方法

# 在文件MiniRAG-0.0.2/minirag/base.py开头添加导入语句
from abc import abstractmethod

问题3：

如果使用的是qwen，注意去hf上把qwen的权重下载下来，然后在py文件中修改路径

if args.model == "PHI":
    LLM_MODEL = "microsoft/Phi-3.5-mini-instruct"
elif args.model == "GLM":
    LLM_MODEL = "THUDM/glm-edge-1.5b-chat"
elif args.model == "MiniCPM":
    LLM_MODEL = "openbmb/MiniCPM3-4B"
elif args.model == "qwen":
    LLM_MODEL = "修改路径"
else:
    print("Invalid model name")
    exit(1)`

问题4：

同理，可以去hf上下载下来all-MiniLM-L6-v2，这本身也是一个小模型用来将文本转换为向量表示

OSError: We couldn't connect to 'https://huggingface.co' to load this file, couldn't find it in the cached files and it looks like sentence-transformers/all-MiniLM-L6-v2 is not the path to a directory containing a file named config.json.
Checkout your internet connection or see how to run the library in offline mode at 'https://huggingface.co/docs/transformers/installation#offline-mode'.

使用自己的数据

使用txt格式，并放入dataset下面即可