A context-aware spell-check ML System that provides appropriate feedback to the user in real-time after they make an error.
Before you begin, ensure you have the following installed:
- Python 3.11 or newer
Follow these steps to install and set up the project:
- Clone the repository
- Use of the precomputed data requires no additional installations, but if complete replication is desired the data should be downloaded from https://userinterfaces.aalto.fi/136Mkeystrokes/ and unpacked to the
data/folder: thescripts/folder then has the processing. This takes a long time. - Install llama-cpp-python with these instructions: https://llama-cpp-python.readthedocs.io/en/latest/#installation-configuration
- Open a terminal in where you cloned the repository to
- Install JAX -
pip install jax - Install the following -
pip install scipy tensorflow tensorflow-probability keras pandas - To install the CRITIC module -
pip install --editable . - Download the Llama 3.2 model to the
models/directory - https://huggingface.co/jxtngx/Meta-Llama-3.2-1B-Q4_K_M-GGUF/resolve/main/llama-3.2-1b-q4_k_m.gguf - To test the installation, a sample correction can be run using
python src/critic/llama.py - Evaluation of the model is done in
notebooks/evaluation.ipynb - Training of the keyboard model and other processing is done in
notebooks/