Sirius-1: Multilingual Translation Model

This is a fine-tuned version of google/gemma-3-12b-it for multilingual translation tasks. The model has been trained on 35 languages using LoRA (Low-Rank Adaptation) technique.

Model Details

• Base Model: google/gemma-3-12b-it
• Fine-tuning Method: LoRA (Low-Rank Adaptation)
• Task: English-to-X translation
• Languages Supported: 35 languages
• Training Data: Custom multilingual translation datasets

Supported Languages

The model supports translation from English to the following 35 languages:

• Chinese (Simplified)
• French
• German
• Japanese
• Korean
• Russian
• Spanish
• Gujarati
• Bengali
• Kazakh
• Persian
• Italian
• Portuguese
• Dutch
• Swedish
• Danish
• Finnish
• Greek
• Czech
• Hungarian
• Romanian
• Bulgarian
• Ukrainian
• Thai
• Vietnamese
• Indonesian
• Malay
• Turkish
• Polish
• Swahili
• Tamil
• Telugu
• Urdu
• Arabic
• Hindi

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# Load the model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("hectordiazgomez/sirius-1")
model = AutoModelForCausalLM.from_pretrained("hectordiazgomez/sirius-1")

# Example translation
input_text = "Translate to French: Hello, how are you?"
inputs = tokenizer(input_text, return_tensors="pt")

with torch.no_grad():
    outputs = model.generate(**inputs, max_length=100, temperature=0.7)
    
translation = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(translation)

Citation

If you use this model, please cite:

@misc{sirius-1-2024,
  title={Sirius-1: Multilingual Translation Model},
  author={hectordiazgomez},
  year={2024},
  howpublished={\url{https://huggingface.co/hectordiazgomez/sirius-1}}
}