Running Llama 2 and other Open-Source LLMs on CPU Inference Locally for Document Q&A

Context

Third-party commercial large language model (LLM) providers like OpenAI's GPT4 have democratized LLM use via simple API calls.
However, there are instances where teams would require self-managed or private model deployment for reasons like data privacy and residency rules.
The proliferation of open-source LLMs has opened up a vast range of options for us, thus reducing our reliance on these third-party providers.
When we host open-source LLMs locally on-premise or in the cloud, the dedicated compute capacity becomes a key issue. While GPU instances may seem the obvious choice, the costs can easily skyrocket beyond budget.
In this project, we will discover how to run quantized versions of open-source LLMs on local CPU inference for document question-and-answer (Q&A).

Ensure you have downloaded the GGML binary file from https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML and placed it into the models/ folder
To start parsing user queries into the application, launch the terminal from the project directory and run the following command: poetry run python main.py "<user query>"
For example, poetry run python main.py "What is the minimum guarantee payable by Adidas?"
Note: Omit the prepended poetry run if you are NOT using Poetry

Tools
LangChain: Framework for developing applications powered by language models
C Transformers: Python bindings for the Transformer models implemented in C/C++ using GGML library
FAISS: Open-source library for efficient similarity search and clustering of dense vectors.
Sentence-Transformers (all-MiniLM-L6-v2): Open-source pre-trained transformer model for embedding text to a 384-dimensional dense vector space for tasks like clustering or semantic search.
Llama-2-7B-Chat: Open-source fine-tuned Llama 2 model designed for chat dialogue. Leverages publicly available instruction datasets and over 1 million human annotations.
Poetry: Tool for dependency management and Python packaging

/assets: Images relevant to the project
/config: Configuration files for LLM application
/data: Dataset used for this project (i.e., Manchester United FC 2022 Annual Report - 177-page PDF document)
/models: Binary file of GGML quantized LLM model (i.e., Llama-2-7B-Chat)
/src: Python codes of key components of LLM application, namely llm.py, utils.py, and prompts.py
/vectorstore: FAISS vector store for documents
db_build.py: Python script to ingest dataset and generate FAISS vector store
main.py: Main Python script to launch the application and to pass user query via command line
pyproject.toml: TOML file to specify which versions of the dependencies used (Poetry)
requirements.txt: List of Python dependencies (and version)