README.md

---
language:
- ain
pipeline_tag: translation
license: cc-by-nc-4.0
---

# Disclaimer
This model is only a preliminary experimental result. This model's capability is at best limited and unreliable.

# Acknowledgements
I am indebted to [Michal Ptaszynski](https://huggingface.co/ptaszynski) for his guidance and encouragement, [Karol Nowakowski](https://huggingface.co/karolnowakowski) for his work to compile an expansive parallel corpus, [David Dale](https://huggingface.co/cointegrated) for his [Medium article](https://cointegrated.medium.com/how-to-fine-tune-a-nllb-200-model-for-translating-a-new-language-a37fc706b865) that helped me to quickly and smoothly take my first steps.

# How to use this model
The following is adapted from [slone/nllb-rus-tyv-v1](https://huggingface.co/slone/nllb-rus-tyv-v1).

```Python
# the version of transformers is important!
!pip install sentencepiece transformers==4.33 > /dev/null
import torch
from transformers import NllbTokenizer, AutoModelForSeq2SeqLM

def fix_tokenizer(tokenizer, new_lang):
    """ Add a new language token to the tokenizer vocabulary (this should be done each time after its initialization) """
    old_len = len(tokenizer) - int(new_lang in tokenizer.added_tokens_encoder)
    tokenizer.lang_code_to_id[new_lang] = old_len-1
    tokenizer.id_to_lang_code[old_len-1] = new_lang
    # always move "mask" to the last position
    tokenizer.fairseq_tokens_to_ids["<mask>"] = len(tokenizer.sp_model) + len(tokenizer.lang_code_to_id) + tokenizer.fairseq_offset

    tokenizer.fairseq_tokens_to_ids.update(tokenizer.lang_code_to_id)
    tokenizer.fairseq_ids_to_tokens = {v: k for k, v in tokenizer.fairseq_tokens_to_ids.items()}
    if new_lang not in tokenizer._additional_special_tokens:
        tokenizer._additional_special_tokens.append(new_lang)
    # clear the added token encoder; otherwise a new token may end up there by mistake
    tokenizer.added_tokens_encoder = {}
    tokenizer.added_tokens_decoder = {}

MODEL_URL = "TwentyNine/nllb-ain-kana-latin-converter-v1"
model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_URL)
tokenizer = NllbTokenizer.from_pretrained(MODEL_URL)
fix_tokenizer(tokenizer, 'ain_Japn')
fix_tokenizer(tokenizer, 'ain_Latn')

def convert(
    text,
    model=model,
    tokenizer=tokenizer,
    src_lang='ain_Japn',
    tgt_lang='ain_Latn',
    max_length='auto',
    num_beams=4,
    n_out=None,
    **kwargs
):
    tokenizer.src_lang = src_lang
    encoded = tokenizer(text, return_tensors="pt", truncation=True, max_length=512)
    if max_length == 'auto':
        max_length = int(32 + 2.0 * encoded.input_ids.shape[1])
    model.eval()
    generated_tokens = model.generate(
        **encoded.to(model.device),
        forced_bos_token_id=tokenizer.lang_code_to_id[tgt_lang],
        max_length=max_length,
        num_beams=num_beams,
        num_return_sequences=n_out or 1,
        **kwargs
    )
    out = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
    if isinstance(text, str) and n_out is None:
        return out[0]
    return

convert("ポイ　セタ　クコン　ルスイ")
# GOOD: 'pon seta ku=kor rusuy'

convert("タント がっこう　オルン　パイェ")
# OK:    'tanto がっこう or un paye'
# IDEAL: 'tanto GAKKO or un paye' or  'tanto GAKKOU or un paye'

convert("セコﾛ　ハウェアン　コﾛ　イシレニネ")
# WRONG: 'sekor hawean korsiren hine'
# IDEAL: 'sekor hawean kor i=siren hine'
```