search

OSU-NLP-Group / HippoRAG

[NeurIPS'24] HippoRAG is a novel RAG framework inspired by human long-term memory that enables LLMs to continuously integrate knowledge across external documents. RAG + Knowledge Graphs + Personalized PageRank.
https://arxiv.org/abs/2405.14831
MIT License
1.42k stars 118 forks source link

I can't use ollama's model #53

Open aJarOfCode opened 2 months ago

aJarOfCode commented 2 months ago

The documentation says you can use langchain's model options I changed the code in the example src/test_hipporag.py the hipporag = HippoRAG(args.dataset, 'openai', args.extraction_model, args.retrieval_model, doc_ensemble=args.doc_ensemble, dpr_only=args.dpr_only, qa_model=LangChainModel('openai', 'gpt-3.5-turbo')) I changed it to hipporag = HippoRAG(args.dataset, 'ollama', args.extraction_model, args.retrieval_model, doc_ensemble=args.doc_ensemble, dpr_only=args.dpr_only, qa_model=LangChainModel('ollama', 'llama3')) When I run python src/test_hipporag.py --dataset hotpotqa --extraction_model llama3 --retrieval_model colbertv2 in the terminal, it returns:

/ home/ajarofcode/anaconda3 / envs/hipporag/lib/python3.9 / site - packages/torch/cuda/set py: 497: UserWarning: Can't initialize NVML warnings.warn("Can't initialize NVML") 2024-09-07 10:00:14,440 - hipporag - CRITICAL - No extraction files found: output/openie_hotpotqa_results_nerllama3*.json ; please check if working directory is correct or if the extraction has been done. Traceback (most recent call last): File "/home/ajarofcode/HippoRAG/src/test_hipporag.py", line 26, in hipporag = HippoRAG(args.dataset, 'ollama', args.extraction_model, args.retrieval_model, doc_ensemble=args.doc_ensemble, dpr_only=args.dpr_only, File "/home/ajarofcode/HippoRAG/src/hipporag.py", line 107, in init self.build_graph() File "/home/ajarofcode/HippoRAG/src/hipporag.py", line 450, in build_graph for edge, weight in tqdm(self.graph_plus.items(), total=len(self.graph_plus), desc='Building Graph'): AttributeError: 'HippoRAG' object has no attribute 'graph_plus' What should I do, please?

yhshu commented 2 months ago

Hello, I think you need to check your cuda environment, because I see:

/ home/ajarofcode/anaconda3 / envs/hipporag/lib/python3.9 / site - packages/torch/cuda/set py: 497: UserWarning: Can't initialize NVML
warnings.warn("Can't initialize NVML")

I believe this is an Ollama issue rather than HippoRAG, and you may get help from Ollama community by checking the cuda environments. Subsequent execution is not possible because of indexing failures.

aJarOfCode commented 2 months ago

      Thank you very much for your reply. Since my computer is a windows operating system, I can only use a Linux virtual machine, so I had a small problem installing the drivers. I really want HippoRAG to be multi-platform.       I wish you a happy life.

发件人: Yiheng Shu @.> 发送时间: 2024年9月7日 20:47 收件人: OSU-NLP-Group/HippoRAG @.> 抄送: Ajarofcode @.>; Author @.> 主题: Re: [OSU-NLP-Group/HippoRAG] I can't use ollama's model (Issue #53)

Hello, I think you need to check your cuda environment, because I see:

/ home/ajarofcode/anaconda3 / envs/hipporag/lib/python3.9 / site - packages/torch/cuda/set py: 497: UserWarning: Can't initialize NVML warnings.warn("Can't initialize NVML")

This is an Ollama issue rather than HippoRAG. Subsequent execution is not possible because of indexing failures.

— Reply to this email directly, view it on GitHubhttps://github.com/OSU-NLP-Group/HippoRAG/issues/53#issuecomment-2335176937, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AYYCQVZZOG375GJPDAONGO3ZVLYX7AVCNFSM6AAAAABNZY2CEGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMZVGE3TMOJTG4. You are receiving this because you authored the thread.Message ID: @.***>