metadata

language:
  - fr
  - de
tags:
  - feature-extraction
  - embeddings
  - sentence-similarity

LaBSE for French and German

This is a shortened version of sentence-transformers/LaBSE. The model was prepaired with the direct help of cointegrated, the author of the LaBSE-en-ru model.

The current model includes only French and German tokens, and the vocabulary is thus 10% of the original while number of parameters in the whole model is 27% of the original.

To get the sentence embeddings, you can use the following code:

import torch
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("EIStakovskii/LaBSE-fr-de")
model = AutoModel.from_pretrained("EIStakovskii/LaBSE-fr-de")
sentences = ["Wie geht es dir?", "Comment vas-tu?"]
encoded_input = tokenizer(sentences, padding=True, truncation=True, max_length=64, return_tensors='pt')
with torch.no_grad():
    model_output = model(**encoded_input)
embeddings = model_output.pooler_output
embeddings = torch.nn.functional.normalize(embeddings)
print(embeddings)

Reference:

Fangxiaoyu Feng, Yinfei Yang, Daniel Cer, Narveen Ari, Wei Wang. Language-agnostic BERT Sentence Embedding. July 2020

License: https://tfhub.dev/google/LaBSE/1