Tools for working with large language models
The bin/...
commands are set up to automatically use a Python virtual environment in venv
if it exist.
Hence, if you do not have all prerequisites already installed and available in your environment, you can set up an environment to use with llmtools like this:
python3 -m venv venv
venv/bin/pip3 install pymupdf python-Levenshtein nltk torch torchvision torchaudio openai sydney-py
venv/bin/pip3 install git+https://github.com/huggingface/transformers.git
To use the locallama backend, you also need access to llama.cpp and a suitable model file. Download and compile llama.cpp for your platform. Then setup a subdirectory as follows:
mkdir llama.cpp
cd llama.cpp
ln -s <path to your llama-cli> llama-cli
mkdir models
cd models
ln -s <path to your 8B llama model> lama-8B.gguf
(Or, you can also just clone the llama.cpp repo here and organize it accordinglt)