Spaces:
Runtime error
Runtime error
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,13 +1,77 @@
|
|
| 1 |
---
|
| 2 |
title: Automatic Machine Translation
|
| 3 |
-
emoji:
|
| 4 |
-
colorFrom:
|
| 5 |
-
colorTo:
|
| 6 |
sdk: gradio
|
| 7 |
-
sdk_version:
|
| 8 |
app_file: app.py
|
| 9 |
pinned: false
|
| 10 |
-
|
|
|
|
| 11 |
---
|
| 12 |
|
| 13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
title: Automatic Machine Translation
|
| 3 |
+
emoji: 🌐
|
| 4 |
+
colorFrom: blue
|
| 5 |
+
colorTo: purple
|
| 6 |
sdk: gradio
|
| 7 |
+
sdk_version: 4.44.0
|
| 8 |
app_file: app.py
|
| 9 |
pinned: false
|
| 10 |
+
license: apache-2.0
|
| 11 |
+
short_description: Chinese ↔ English oral translation with Qwen3-0.6B + LoRA
|
| 12 |
---
|
| 13 |
|
| 14 |
+
# 🌐 口语化自动机器翻译 | Oral Machine Translation
|
| 15 |
+
|
| 16 |
+
基于 **Qwen3-0.6B + LoRA** 微调的中英双向口语翻译系统
|
| 17 |
+
|
| 18 |
+
A bilingual (Chinese ↔ English) oral translation system built with **Qwen3-0.6B** fine-tuned using **LoRA**.
|
| 19 |
+
|
| 20 |
+
## ✨ Features
|
| 21 |
+
|
| 22 |
+
- 🔄 **Bidirectional Translation**: Chinese → English and English → Chinese
|
| 23 |
+
- 💬 **Oral Style**: Optimized for conversational, natural language
|
| 24 |
+
- ⚡ **Lightweight**: Only 0.6B parameters, fast inference
|
| 25 |
+
- 🎯 **LoRA Fine-tuned**: Parameter-efficient training on OpenSubtitles dataset
|
| 26 |
+
|
| 27 |
+
## 📊 Performance
|
| 28 |
+
|
| 29 |
+
| Model | BLEU Score | Improvement |
|
| 30 |
+
|-------|------------|-------------|
|
| 31 |
+
| Baseline (Qwen3-0.6B) | 1.24 | - |
|
| 32 |
+
| **LoRA Fine-tuned** | **11.89** | **+858%** |
|
| 33 |
+
|
| 34 |
+
## 🛠️ Technical Details
|
| 35 |
+
|
| 36 |
+
- **Base Model**: [Qwen/Qwen3-0.6B](https://huggingface.co/Qwen/Qwen3-0.6B)
|
| 37 |
+
- **Fine-tuning Method**: LoRA (rank=16, alpha=32, dropout=0.05)
|
| 38 |
+
- **Training Data**: OpenSubtitles en-zh (5,000 samples)
|
| 39 |
+
- **Training Time**: ~6 minutes (dual GPU)
|
| 40 |
+
- **Framework**: Transformers + PEFT
|
| 41 |
+
|
| 42 |
+
## 🚀 Usage
|
| 43 |
+
|
| 44 |
+
Simply select the translation direction and enter your text. The model will generate a natural, conversational translation.
|
| 45 |
+
|
| 46 |
+
### Example Translations
|
| 47 |
+
|
| 48 |
+
**Chinese → English:**
|
| 49 |
+
- 输入: 你好呀
|
| 50 |
+
- 输出: Hi, how are you?
|
| 51 |
+
|
| 52 |
+
**English → Chinese:**
|
| 53 |
+
- 输入: See you later
|
| 54 |
+
- 输出: 回头见。
|
| 55 |
+
|
| 56 |
+
## 📝 Citation
|
| 57 |
+
|
| 58 |
+
If you use this model, please cite:
|
| 59 |
+
|
| 60 |
+
```bibtex
|
| 61 |
+
@misc{mt-qwen-oral,
|
| 62 |
+
title={MT-Qwen-Oral: Bilingual Oral Translation with Qwen and LoRA},
|
| 63 |
+
author={Hzzzzx0},
|
| 64 |
+
year={2025},
|
| 65 |
+
publisher={Hugging Face},
|
| 66 |
+
howpublished={\url{https://huggingface.co/spaces/Hzzzzx0/Automatic_machine_translation}}
|
| 67 |
+
}
|
| 68 |
+
```
|
| 69 |
+
|
| 70 |
+
## 📄 License
|
| 71 |
+
|
| 72 |
+
Apache 2.0
|
| 73 |
+
|
| 74 |
+
## 🔗 Links
|
| 75 |
+
|
| 76 |
+
- [GitHub Repository](https://github.com/yourusername/mt-qwen-oral)
|
| 77 |
+
- [Model Weights](https://huggingface.co/Hzzzzx0/qwen3-0.6b-oral-lora)
|