nllb200-en-formosan-spm8k

Direction: English to Formosan
Base model: facebook/nllb-200-distilled-600M
Recipe: nllb200-spm8k-directional-v5
Release: 20260816-193900, validation-selected step 250,000

This is a directional model for 15 Formosan languages. It uses the private_no_bible leakage-controlled corpus, Formosan-aware 8k SentencePiece extension, balanced language/source sampling, direction/language controls, fixed coarse domain tags, and dialect tags. Domain and dialect metadata use independent training dropout so unknown and default are learned inference conditions. The model weights are public. The training corpus is distributed separately to authorized FormosanBank members through the access-controlled FormosanBank/formosan-mt-private dataset and is not included with the weights.

Model details

Item	Value
Base revision	f8d333a098d19b4fd9a8b18f94170487ad3f821d
Training rows	344,701
Effective batch size	64
Maximum sequence length	384
Learning rate	2e-05
Precision	bf16
Checkpoint selection	Validation chrF2
Formosan text	kindOf=standard, formosan-mt-standard-v3
Corpus SHA-256	c84174ee46027e50d24aedc3cf4016412f7a4cbdeb14ceed2133ee76a4724be1
Training profile SHA-256	4253b1b136fedebc589da942b80df71945212d1f4b92a8e98ec6fb160712f14b

Usage

import torch
from transformers import AutoModelForSeq2SeqLM, NllbTokenizer


model_id = "FormosanBank/nllb200-en-formosan-spm8k"
tokenizer = NllbTokenizer.from_pretrained(model_id, use_fast=False)
model = AutoModelForSeq2SeqLM.from_pretrained(model_id)
model.to("cuda" if torch.cuda.is_available() else "cpu")
NLLB_LIDS = {'ami': 'ami_Latn', 'bnn': 'bnn_Latn', 'ckv': 'ckv_Latn', 'dru': 'dru_Latn', 'pwn': 'pwn_Latn', 'pyu': 'pyu_Latn', 'ssf': 'ssf_Latn', 'sxr': 'sxr_Latn', 'szy': 'szy_Latn', 'tao': 'tao_Latn', 'tay': 'tay_Latn', 'trv': 'trv_Latn', 'tsu': 'tsu_Latn', 'xnb': 'xnb_Latn', 'xsy': 'xsy_Latn'}

def translate(text, lang_code, source_bucket="unknown", dialect="default"):

    tokenizer.src_lang = 'eng_Latn'
    source_bucket = source_bucket if source_bucket in ('dictionary', 'classroom', 'narrative', 'linguistic', 'education', 'media', 'culture', 'religious', 'unknown') else "unknown"
    domain_tag = f"<dom_{source_bucket}>"
    if tokenizer.convert_tokens_to_ids(domain_tag) == tokenizer.unk_token_id:
        domain_tag = "<dom_unknown>"
    dialect_tag = f"<dialect_{dialect}>"
    if tokenizer.convert_tokens_to_ids(dialect_tag) == tokenizer.unk_token_id:
        dialect_tag = "<dialect_default>"
    prompt = (
        f"<to_{lang_code}> <src_eng> {domain_tag} {dialect_tag} {text}"
    )
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    output = model.generate(
        **inputs,
        decoder_start_token_id=tokenizer.eos_token_id,
        forced_bos_token_id=tokenizer.convert_tokens_to_ids(NLLB_LIDS[lang_code]),
        max_new_tokens=256,
        num_beams=4,
    )
    return tokenizer.batch_decode(output, skip_special_tokens=True)[0]

print(translate('He went home.', "ami"))

The control tags are part of the training contract. Use unknown and default when source bucket or dialect metadata is unavailable.

Evaluation

The best checkpoint was selected on validation chrF2. Test and validation contain only eligible, human-translated sentence pairs. Synthetic pivots and lexical entries are train-only. The headline result uses default metadata controls, so it does not assume access to test-set domain or dialect labels.

Split	Rows
Train	344,701
Test	40,561
Validate	20,286

Scope	BLEU	chrF2	TER
Hard test	7.84	32.66	82.37
Selection validation	7.59	33.90	80.05

Test empty-output rate: 0.0000%.

Language	Samples	BLEU	chrF2	TER
ami	6,761	6.73	29.76	82.47
bnn	4,815	4.04	31.91	85.88
ckv	1,915	11.42	38.26	77.55
dru	3,782	3.31	25.52	99.47
pwn	5,237	9.81	37.14	74.82
pyu	3,158	8.51	35.37	80.83
ssf	1,077	17.52	45.12	62.67
sxr	937	7.34	37.34	83.29
szy	857	13.94	39.94	65.34
tao	868	10.89	40.36	76.80
tay	4,763	3.93	23.79	93.18
trv	2,888	8.07	29.90	79.66
tsu	1,111	11.17	37.77	72.10
xnb	1,503	9.02	40.83	77.28
xsy	889	18.90	44.55	66.22

The corpus gate enforces standard-tier Formosan text, 5/10 validation/test proportions from all deduplicated pairs, capacity-aware source balance, sentence-only evaluation, and zero exact, skeleton, one-edit, or configured high character n-gram train/evaluation conflicts. Document overlap is diagnostic. This release passed all leakage gates: exact 0, skeleton 0, one-edit 0, character n-gram 0. Document overlap: 924.

See eval/metrics.json for sacreBLEU signatures, per-language, source, dialect, and length diagnostics. publication.json records the corpus, profile, run, and checkpoint hashes used for this release.

Intended use

This model supports research, corpus development, and assisted translation for the 15 included Formosan languages. It is designed for the exact prompt and generation contract shown above.

Limitations

Outputs require knowledgeable speaker review. Aggregate metrics hide large differences among languages and domains. This model is not suitable for authoritative, medical, legal, or safety-critical translation.