Metadata-Version: 2.1
Name: sacrecomet
Version: 0.0.1
Summary: Tool to guide you through reporting the use of COMET for machine translation evaluation.
Author-email: Vilém Zouhar <vzouhar@ethz.ch>, Tsz Kin Lam <tlam@ed.ac.uk>, Pinzhen Chen <pinzhen.chen@ed.ac.uk>, Nikita Moghe <nikita.moghe@ed.ac.uk>
Project-URL: Homepage, https://github.com/PinzhenChen/sacreCOMET
Keywords: machine-translation,evaluation,quality-estimaation
Requires-Python: >=3.7
Description-Content-Type: text/markdown

# SacreCOMET

The tool has two functionalities.
First, it creates a signature with your setup and COMET model:

```
pip install sacrecomet

# Without anything will try to detect the local environment and will
# ask you questions about which COMET model you used.
# Example output: Python3.11.8|Comet2.2.2|fp32|unite-mup

sacrecomet 

# Arguments can also be specified non-interactively:

sacrecomet --model unite-mup --prec fp32
```

The other functionality is to find specific citations for COMET models that you're using:

```
sacrecomet cite Unbabel/xcomet-xl

https://arxiv.org/abs/2310.10482
@misc{guerreiro2023xcomet,
    title={xCOMET: Transparent Machine Translation Evaluation through Fine-grained Error Detection}, 
    ...
```

Please open an issue/pull request in the [repository](https://github.com/PinzhenChen/sacreCOMET) if you wish to add models/functionality.


<!-- 
Notes for maintainers:

cd python-tool
# newer version might not work
pip install 'build<0.10.0' twine


python3 -m build

# first tets
twine check dist/*
twine upload -r testpypi dist/*

# live
twine upload dist/* -u __token__
# user __token__ as username and the API token generated online
-->
