RAG architecture based on the model family qwen2.5
#22
by
shevel
- opened
I use ollama and qwen2.5:32b in my naive RAG architecture with the set of documents.
The content of the document set is remained unchanged.
The parameneters of the RAG and model are unchanged either.
Is it possible that model improve the correctness of the answers with
encreasing the number of various quesries (questions)?
From my (may be stupit) consideration the model could change something to improve the consistency of the answers. Howevr such the changing
have to be written somehow in the files connected to the model (qwen2.5).
If it is true there are at least two consiquescies:
- the changed somehow model qwen2.5 became little bit different from the initial state of the model;
- the chagings must be writtent in files, however after several weeks to use the model I do not see any changing in the files connetycd to qwen2.5.
Any considerations or ideas are welcomed!