nileagi-suk
Collection
NileAGI Sukuma dataset and small STT, TTS, and MT models. • 5 items • Updated
Request access
Describe your intended use below. Downloads unlock after approval.
Weights are released under CC BY-NC-SA 4.0 for non-commercial research and education. Credit NileAGI, keep adaptations under the same licence, and contact us before any commercial product or paid service. Do not use the model to harm people, break the law, or present outputs as human writing.
Log in or Sign Up to review the conditions and access this model content.
Sukuma MT (reverse)
Authors: Zephania and Isack Odero
Sukuma → Swahili machine translation for the macron orthography (ā ē ī ō ū). Reads Sukuma text and writes Swahili text.
Companion to the forward SKU nileagi/nileagi-suk-mt (Swahili → Sukuma).
| Forward (quality) | nileagi/nileagi-suk-mt |
| Forward (lite) | nileagi/nileagi-suk-mt-lite |
| Collection | nileagi/nileagi-suk |
Summary
| Task | Machine translation |
| Direction | Sukuma → Swahili |
| Language codes | suk_Latn → swh_Latn |
| Base | facebook/nllb-200-distilled-600M |
| Orthography | Latin with vowel macrons (input) |
| Primary metric | chrF2 |
| Test chrF2 | 39.5 · BLEU 11.7 |
Held-out document-group test. Prefer chrF2; BLEU is secondary.
How to use
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("nileagi/nileagi-suk-mt-rev", use_fast=False)
model = AutoModelForSeq2SeqLM.from_pretrained("nileagi/nileagi-suk-mt-rev")
tok.src_lang = "suk_Latn"
inputs = tok("Ūlī mhola?", return_tensors="pt")
bos = tok.convert_tokens_to_ids("swh_Latn")
out = model.generate(**inputs, forced_bos_token_id=bos, max_new_tokens=128)
print(tok.decode(out[0], skip_special_tokens=True))
Always set suk_Latn and force BOS swh_Latn.
Intended use
- Literary / read-aloud Sukuma → Swahili
- Research on low-resource Bantu MT
Out of scope
- Direct English ↔ Sukuma
- Chat, legal, medical, or news translation
- Conversational or social-media Sukuma without macrons (domain shift)
- Commercial products without a NileAGI licence
Limitations
- Literary macron orthography; macron-free input looks like domain shift
- Research-preview scores — not a substitute for a human translator
- Names and rare stems are often approximated
Access
Weights are gated under CC BY-NC-SA 4.0.
Attribute NileAGI. Adapted weights must stay CC BY-NC-SA 4.0. Commercial: hi@nileagi.com.
License
Weights: CC BY-NC-SA 4.0.
See LICENSE and NOTICE.md.
Citation
@misc{nileagi-suk-mt-rev-2026,
title = {Sukuma machine translation (Sukuma to Swahili)},
author = {Zephania and Isack Odero},
year = {2026},
howpublished = {Hugging Face},
url = {https://huggingface.co/nileagi/nileagi-suk-mt-rev},
note = {NileAGI}
}
- Downloads last month
- 1