XLSR

Cross-lingual sentence representations project on GitHub: German-NLP-Group/xlsr

We have created a new project with two gloals:

• provide clean code to train models for cross-Lingual sentence representations

• provide models that further increase the performance of T-Systems-onsite/cross-en-de-roberta-sentence-transformer.

The model training consists of two parts steps:

1. automated hyperparameter oprimization with optuna and a 10-fold crossvalidation

2. training and saving the models with the found hyperparameterset

More details can be found on the README.md page of the project.