A newer version of the Gradio SDK is available:
6.2.0
metadata
title: Automatic Machine Translation
emoji: 🌐
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.44.0
app_file: app.py
pinned: false
license: apache-2.0
short_description: Chinese ↔ English oral translation with Qwen3-0.6B + LoRA
🌐 口语化自动机器翻译 | Oral Machine Translation
基于 Qwen3-0.6B + LoRA 微调的中英双向口语翻译系统
A bilingual (Chinese ↔ English) oral translation system built with Qwen3-0.6B fine-tuned using LoRA.
✨ Features
- 🔄 Bidirectional Translation: Chinese → English and English → Chinese
- 💬 Oral Style: Optimized for conversational, natural language
- ⚡ Lightweight: Only 0.6B parameters, fast inference
- 🎯 LoRA Fine-tuned: Parameter-efficient training on OpenSubtitles dataset
📊 Performance
| Model | BLEU Score | Improvement |
|---|---|---|
| Baseline (Qwen3-0.6B) | 1.24 | - |
| LoRA Fine-tuned | 11.89 | +858% |
🛠️ Technical Details
- Base Model: Qwen/Qwen3-0.6B
- Fine-tuning Method: LoRA (rank=16, alpha=32, dropout=0.05)
- Training Data: OpenSubtitles en-zh (5,000 samples)
- Training Time: ~6 minutes (dual GPU)
- Framework: Transformers + PEFT
🚀 Usage
Simply select the translation direction and enter your text. The model will generate a natural, conversational translation.
Example Translations
Chinese → English:
- 输入: 你好呀
- 输出: Hi, how are you?
English → Chinese:
- 输入: See you later
- 输出: 回头见。
📝 Citation
If you use this model, please cite:
@misc{mt-qwen-oral,
title={MT-Qwen-Oral: Bilingual Oral Translation with Qwen and LoRA},
author={Hzzzzx0},
year={2025},
publisher={Hugging Face},
howpublished={\url{https://huggingface.co/spaces/Hzzzzx0/Automatic_machine_translation}}
}
📄 License
Apache 2.0