CTranslate2-converted OPUS-MT model
This repository contains a CTranslate2-converted copy of an OPUS-MT model for the language pair indicated by the repository name. I did not train or create the model — I only converted an existing OPUS-MT checkpoint so it can be used easily with CTranslate2 for local inference.
Included (typical)
- CTranslate2 model files (e.g. model binaries and generation config)
- SentencePiece tokenizers:
- Tokenizer / metadata files:
tokenizer_config.jsonvocab.jsonshared_vocabulary.json (when present)
config.json (model metadata)converted.ok (conversion sentinel)
Provenance & License
- Original model: an OPUS-MT model (Helsinki-NLP / OPUS-MT). This repository only contains a converted copy.
- License and dataset / citation requirements are determined by the original model and its authors.
Before redistributing or using this model in a commercial setting, check the original model page on Hugging Face (Helsinki-NLP/OPUS-MT or the specific original repo) for the exact license and citation text.
Conversion Details
- Converted to CTranslate2 format using the conversion tooling (see the original project’s conversion scripts if you need to reproduce or tweak the conversion).
- Presence of
converted.ok indicates the conversion completed successfully.
Notes & Troubleshooting
- If
Translator() fails to load, ensure all model files (binaries + tokenizer files) were downloaded and that you have a compatible ctranslate2 version installed.
- Large models may require Git LFS on download or the Hugging Face snapshot API.
- For performance, run on a machine with a suitable CPU/GPU and a recent
ctranslate2 build.
Attribution / Contact
If you have questions about the conversion, please check the original OPUS-MT model page for details and issues. This repository is only a converted copy and does not replace the original project or its documentation.
