Downloads · 30 days
0
InterstellarCG/kikuyu-translator-final
kikuyu-translator-final is a translation model from InterstellarCG. Use it when you need text moved from one language to another. The card lists the license as cc-by-nc-4.0.
A fine-tuned translation model for English ↔ Kikuyu (Gĩkũyũ) language translation.
Downloads · 30 days
0
Access
Public
Updated Jan 22, 2026
Repo size
731 MB
Likes
2
Public
Click a slice to open those files.
.safetensors615 MB · 62%
From the Hugging Face model README
A fine-tuned translation model for English ↔ Kikuyu (Gĩkũyũ) language translation.
This model is based on Gemma 3 4B and fine-tuned on a diverse collection of parallel Kikuyu-English data. It supports bidirectional translation between English and Kikuyu.
This model was trained through 4 progressive stages over multiple days:
| Stage | Model | Data | Training Time | Description |
|---|---|---|---|---|
| 1 | kikuyu_translator | Bible + FLORES + Bloom | ~24 hours | Initial fine-tuning from Gemma 3 4B |
| 2 | kikuyu_translator_diverse | Diverse non-religious data | ~6 hours | Reduce religious overfitting |
| 3 | kikuyu_translator_mt560 | MT560 JW data (188k) | ~10 hours | Large-scale translation pairs |
| 4 | kikuyu_translator_final | Combined balanced data (141k) | ~15 hours | Final balanced training |
Total Training Time: ~55+ hours
| Source | Pairs | Type |
|---|---|---|
| FLORES-200 | 2,009 × 5 | News/diverse (5× upsampled) |
| Bloom Library | 632 × 5 | Children's stories/health |
| KevinKibe Agriculture | 420 × 5 | Agricultural terminology |
| Gemini Synthetic | 10 × 5 | Modern topics |
| Bible (50%) | 15,549 | Religious text |
| MT560 JW (30%) | 40,000 | Religious translations |
from unsloth import FastModel
from unsloth.chat_templates import get_chat_template
model, tokenizer = FastModel.from_pretrained(
"InterstellarCG/kikuyu-translator-final",
max_seq_length=2048,
load_in_4bit=True,
)
tokenizer = get_chat_template(tokenizer, chat_template="gemma-3")
# English to Kikuyu
messages = [{"role": "user", "content": "Translate the following English text to Kikuyu.\n\nThe government announced new policies to support farmers."}]
inputs = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt")
outputs = model.generate(input_ids=inputs, max_new_tokens=200, temperature=0.3)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
| Direction | Input | Output |
|---|---|---|
| KI→EN | "Thirikari ya Kenya nĩ ĩraheanire watho mwerũ wa gũteithia arĩmi." | "The Kenyan government has introduced a new policy to assist farmers." |
| KI→EN | "Ciana nĩ ciathomire mũno na ciakĩhota igeranio." | "The children read a lot and performed well in the tests." |
This model is released under CC-BY-NC-4.0. The training data includes multiple sources with varying licenses - please check the dataset repository for details.
If you use this model, please cite:
@misc{interstellarcg-kikuyu-translator-2025,
author = {InterstellarCG},
title = {Kikuyu Translator Final},
year = {2025},
publisher = {Hugging Face},
url = {https://huggingface.co/InterstellarCG/kikuyu-translator-final}
}