COMET models are trained to predict quality scores for translations. This is a mock version which provides random scores.
Clone this repository:
gh repo clone capstanlqc/comet-api-mock
Create a virtual environment and activate it:
cd /path/to/comet-api-mock
python -m venv venv && source venv/bin/activate
Install dependencies:
pip install -r requirements.txt
Copy .env.example to .env:
cp .env-example .env
Update the environment variables you need to use in .env (none in the mock version).
Run the API (for testing and development):
fastapi dev code/main.py
The mandatory model parameter accepts values such as provider/model. The optional mode parameter accepts either mock (default value, if missing) or any other value (e.g. dev or prod, not yet defined).
The following call will get mock random scores (same as without mode parameter):
curl --location --request GET 'http://127.0.0.1:8000/api/scores?model=Unbabel/wmt22-cometkiwi-da&mode=mock' \
--header 'Content-Type: application/json' \
--data '[
{"src": "How to Demonstrate Your Strategic Thinking Skills", "mt": "Cómo demostrar su capacidad de pensamiento estratégico" },{ "src": "Why is Accuracy important in the workplace?", "mt": "¿Por qué es importante la precisión en el trabajo" }, { "src": "When faced with a large amount of analysis ask for support setting up a team to approach the issue in different ways.", "mt": "Cuando se enfrente a una gran cantidad de análisis, pida ayuda para crear un equipo que aborde la cuestión de diferentes maneras." }
]'