The model training can be found in the file commonlitreadabilityprize.ipynb
The training process stores some artifacts: model weights and dictionary for the vectorization layer.
Code is in file main.py.
It's a web server that runs the prediction.
To run it:
python main.py
To test it, run from the command line:
curl -H "Content-Type: application/json" -d '{"instances": [ "Hello world" ]}' -X POST http://localhost:5000