---
license: cc-by-nc-4.0
language:
- es
- zap
pipeline_tag: translation
tags:
- translation
- nllb
- zapotec
- diidxaza
- low-resource
- zai
- Juchitan
- translator
- oaxaca
- Machine
- lora
- facebook
- finetuned

base_model: facebook/nllb-200-distilled-600M
library_name: transformers
---

# Spanish → Isthmus Zapotec (Diidxazá) Experimental translator

Neural machine translation from Spanish (`spa_Latn`) into Isthmus Zapotec
(Diidxazá), built by LoRA fine-tuning,
[`facebook/nllb-200-distilled-600M`](https://huggingface.co/facebook/nllb-200-distilled-600M)
and merging the adapter back into the base for a standalone model.

Isthmus Zapotec is not covered by any pretrained MT model, so an existing NLLB
language slot (`ayr_Latn`) is reused as the Zapotec output tag. Every training
target is Zapotec, so the model fully repurposes that tag. The forced target tag
is stored in `generation_config.json` (`forced_bos_token_id`), so generation
needs no special setup.

This is a side result of the main project called Didxazapp.

## Usage

```python
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

model_id = "simpleverso/espzap-nllb-600m"
model = AutoModelForSeq2SeqLM.from_pretrained(model_id)
tok = AutoTokenizer.from_pretrained(model_id, src_lang="spa_Latn")

def generate_text(text: str) -> str:
    inputs = tok(text, return_tensors="pt").to(model.device)
    out = model.generate(**inputs, num_beams=4, max_length=96)
    return tok.batch_decode(out, skip_special_tokens=True)[0]

generate_text("quiero ver tu risa todo el dia")
```


You can use the onnx version of the model directly in your browser here:
https://huggingface.co/spaces/simpleverso/espzap-translator

## Results

Trained on 18,452 unique parallel pairs.

| Metric | Dev (best epoch) | Test (held-out) |
|---|---|---|
| chrF | 42.1 | 37.7 |
| BLEU | 9.7  | 6.4  |


## Models output:
- Safetensor
- Onnx

## Limitations

- Trained on ~18.5k pairs; usable but imperfect. There is no standarized zapotec information, Im working to build one.
- Part of the training data is unreviewed, so some outputs may
  reflect noise in that source.
- Handles everyday sentences best; long or highly literary input degrades.

## License

The base model NLLB-200 is released under CC-BY-NC-4.0; this derivative inherits
that non-commercial license.

## Author
Gonzalo Santiago Martinez

## Disclaimer:
Any translation is provided for convenience only. The author makes no warranty of any kind, express or implied, as to the accuracy, reliability, or correctness of any translations. Any reliance on the translated text is at the user's own risk. The author are not liable for any harm, loss, or damage resulting from the use of or reliance on this translation tool.