Translation
LiteRT
Safetensors
Spanish
Basque
marian
odegiber's picture
Update README.md
fc25d84 verified
metadata
license: apache-2.0
datasets:
  - Helsinki-NLP/tatoeba
  - openlanguagedata/flores_plus
language:
  - es
  - eu
metrics:
  - bleu
  - comet
  - chrf
pipeline_tag: translation

OPUS-MT-tiny-spa-eus

Distilled model from a Tatoeba-MT Teacher: Tatoeba-MT-models/itc-eus/opusTCv20210807_transformer-big_2022-07-23, which has been trained on the Tatoeba dataset.

We used the OpusDistillery to train new a new student with the tiny architecture, with a regular transformer decoder. For training data, we used Tatoeba. The configuration file fed into OpusDistillery can be found here.

How to run

from transformers import MarianMTModel, MarianTokenizer
model_name = "Helsinki-NLP/opus-mt_tiny_spa-eus"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)
tok = tokenizer("La gastronomía de Mayorca, como la de otras regiones similares del Mediterráneo, se basa en el pan, los vegetales y la carne (especialmente la porcina), y utiliza aceite de oliva en todas sus recetas.", return_tensors="pt").input_ids
output = model.generate(tok)[0]
tokenizer.decode(output, skip_special_tokens=True)

Benchmarks

Teacher

testset BLEU chr-F COMET
Flores+ 13.3 52.5 0.8407

Student

testset BLEU chr-F COMET
Flores+ 11.7 51.6 0.824

Marian models

We also provide Marian-compatible versions of this model. To use them, compile Marian and run decoding with marian-decoder, for example:

marian-decoder \
  -i input.txt \
  -c final.model.npz.best-perplexity.npz.decoder.yml \
  -m final.model.npz.best-perplexity.npz \
  -v vocab.spm vocab.spm