search

deepset-ai / haystack

AI orchestration framework to build customizable, production-ready LLM applications. Connect components (models, vector DBs, file converters) to pipelines or agents that can interact with your data. With advanced retrieval methods, it's best suited for building RAG, question answering, semantic search or conversational agent chatbots.
https://haystack.deepset.ai
Apache License 2.0
17.69k stars 1.92k forks source link

unable to run tutorial 10 Knowledge graph #1447

Closed aqsa27 closed 3 years ago

aqsa27 commented 3 years ago

Hi,

I am trying to run the tutorial_10 Knowledge graph and i am facing the following error:

Also can someone help me to understand how can i run the tutorial 10 on my own dataset or my own KG?

or any link that can help me to understand the tutorial 10 better. I am trying to build a Question Answering System using KG

Screenshot 2021-09-13 at 12-52-19 Untitled85 - Jupyter Notebook

julian-risch commented 3 years ago

Hi @aqsa27 Here is the link to our documentation page about question answering on knowledge graphs: https://haystack.deepset.ai/usage/knowledge-graph

Regarding your error message: the problem might be related to an unsuccessful installation of faiss. Could you please try to run conda install faiss-cpu -c pytorch instead of pip install faiss? It's one line in the requirements.txt: https://github.com/deepset-ai/haystack/blob/9a1e3fec8607975c5b7359ce41cf0869b64695ec/requirements.txt#L24 if it is already installed, try to uninstall it first and then re-install again. pip uninstall faiss-cpu

What operating system are you using? Am I assuming correctly that you are working with Windows?

aqsa27 commented 3 years ago

yes I am using windows. I will try your solution and update here. Thank you

aqsa27 commented 3 years ago

@julian-risch is it possible to run the tutorial 10 on my local on my own data set, The use case is I have a data set with question Answer and i have a KG built for it. And i have a set of new similar questions that i need to find the answers from the original dataset. 1) Can i use the pre-trained model to translate text queries(in my case new similar questions) to find answers from my data set? 2) Or do i need to train the model as mentioned in the documentation with my Ques Answer dataset?

If this is not possible in haystack, can you tell me any other source that can be used?

julian-risch commented 3 years ago

We have a section "Trying Question Answering on Knowledge Graphs with Custom Data" on this documentation page here: https://haystack.deepset.ai/components/knowledge-graph with all the information you are looking for, including a link to a seq2seq example for summarization with BART in transformers You can't use the pre-trained model for text2sparql translation on your own data. Instead you need to train your own model. I will close this issue for now. Feel free to reopen if you're still facing problems running the tutorial 10.