---
language_details: spa_Latn, quy_Latn, nso_Latn, fra_Latn, bos_Latn
pipeline_tag: translation
tags:
- mt5
license: apache-2.0
inference: false
---

# Description
Finetuned [google/mt5-large](https://huggingface.co/google/mt5-large) model to translate between Spanish ("spa_Latn") and Mapuzungún.
We support the following languages/graphemaries:
- Spanish (**spa_Latn**)
- Azümchefe (**quy_Latn**)
- Ragileo (**nso_Latn**)
- Unificado (**fra_Latn**)

# Example
```python
from transformers import T5TokenizerFast, AutoModelForSeq2SeqLM

tokenizer = T5TokenizerFast.from_pretrained("CenIA/mt5-large-spa-rap")
model = AutoModelForSeq2SeqLM.from_pretrained("CenIA/mt5-large-spa-rap")

def translate(sentence: str, translate_from="spa_Latn", translate_to="mri_Latn") -> str:
    inputs = tokenizer(translate_from + sentence, return_tensors="pt")
    result = model.generate(**inputs, forced_bos_token_id=tokenizer.convert_tokens_to_ids(translate_to))
    decoded = tokenizer.batch_decode(result, skip_special_tokens=True)[0]

    return decoded

traduction = translate("Hola, ¿cómo estás?")

print(traduction)
```