MTee translation model for general domain
A general domain translation model for the MTee machine translation platform. The platform was developed in 2021 as a collaboration between the TartuNLP, the NLP research group at the University of Tartu, and Tilde. More information about the project can be found here.
The model uses a modular architecture, where each language has its own encoder and decoder that is used for all translation directions. The model can be used with our custom version of FairSeq and it is compatible with the MTee platform and its NMT workers. Additionally, it is fully compatible with TartuNLP's translation API components (API and NMT workers).
Supported translation directions: et-en, en-et, et-de, de-et, et-ru, ru-et.
| Included files: | |
|---|---|
| Fairseq translation model | modular_model.pt |
| SentecePiece models | sp-model.{lang}.model |
| translation model vocabularies | dict.{lang}.txt |
from fairseq.checkpoint_utils import load_model_ensemble_and_task_from_hf_hub models, cfg, task = load_model_ensemble_and_task_from_hf_hub( "tartuNLP/mtee-general" )