Book Reader

Build a book reader (Q&A LLM chain)

Books

Time Machine by H. G. Wells

Project Organization

├── AUTHORS.md
├── LICENSE
├── README.md
├── config
│   ├── `requirements.txt`: packages
│   └── `token_access.py`: the file is used to store tokens (remember to add the filename in `.gitignore` file to protect your token) 
├── data
├── notebooks
│   ├── `00-book-time-machine_flan_t5_large.ipynb`: using `google/flan-t5-large`
│   ├── `01-book-time-machine-llama2_7B.ipynb`: using `llama2_7B`
│   └── `02-book-time-machine-mistral_7B.ipynb`: using `mistral_7B`
└── model
    └──cache: storing models

How to use this repo?

1. Download this repo

git clone https://github.com/dujm/book-reader.git

2. Create a conda environment

# create an env (here I name it "llm") with a stable Python version (e.g. Python 3.8) 
conda create -n llm python=3.8

# activate env
conda activate llm

3. Install Python packages

pip install -r ./config/requirements.txt

4. Install Jupyterlab

• if you want to use jupyterlab (alternatively, you can use jupyter notebook)

conda install -c conda-forge jupyterlab

# add conda environment to jupyter lab
conda install ipykernel
ipython kernel install --user --name=llm

# open jupyter lab
jupyter lab

5. Set up Ollama (optional)

• Ollama is used in
  • notebooks/01-book-time-machine-llama2_7B.ipynb
  • notebooks/02-book-time-machine-mistral_7B.ipynb
• Below is for MacOS. Find more instructions on Ollama if you use other operating systems.

5.1 First-time using Ollma (for Mac users)

• Download file from Ollama website
• Open Ollama app
• Select a model from Model library.
• I selectedllama2 and mistral model.
• Download them in the terminal

# pull llama2 model
ollama pull llama2

# pull mistral model
ollama pull mistral

5.2 For future use of Ollama

• Open Ollama app

6. Run notebooks

Go to notebooks/

Tools that I used

• Chroma vector database
• HuggingFacePipeline
• Langchain
• Ollama
• SentenceTransformers

Thanks to

• Project Gutenberg
• Databricks

---

Go to Top