A minimalistic chatbot script to experiment with LLMs.
-
Be on a machine with an NVIDIA card with 12-24 GB of VRAM.
-
Get the environment ready
conda create -n llm-playground python=3.10
conda activate llm-playground
conda install -y cuda -c nvidia/label/cuda-11.7.0
conda install -y pytorch=2 pytorch-cuda=11.7 -c pytorch
- If on WSL, help bitsandbytes understand where to grab the dynamic libs from
export LD_LIBRARY_PATH=/usr/lib/wsl/lib
- Install the requirements
pip install -r requirements.txt
- Run the script
python app.py
usage: app.py [-h] [-s]
Chatbot Demo
options:
-h, --help show this help message and exit
-s, --share Enable sharing of the Gradio interface
MIT