Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,84 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
+
- fr
|
| 5 |
+
library_name: transformers
|
| 6 |
+
pipeline_tag: translation
|
| 7 |
+
license: cc-by-nc-4.0
|
| 8 |
+
base_model: facebook/nllb-200-distilled-600M
|
| 9 |
+
tags:
|
| 10 |
+
- translation
|
| 11 |
+
- gaming
|
| 12 |
+
- renpy
|
| 13 |
+
- visual-novel
|
| 14 |
+
- french
|
| 15 |
+
datasets:
|
| 16 |
+
- custom
|
| 17 |
+
metrics:
|
| 18 |
+
- bleu
|
| 19 |
+
- sacrebleu
|
| 20 |
+
widget:
|
| 21 |
+
- text: "Hello! Welcome to our game."
|
| 22 |
+
example_title: "Gaming Interface"
|
| 23 |
+
- text: "I love you more than anything."
|
| 24 |
+
example_title: "Romance Dialogue"
|
| 25 |
+
- text: "What do you choose?"
|
| 26 |
+
example_title: "Choice Menu"
|
| 27 |
+
---
|
| 28 |
+
|
| 29 |
+
# 🎮 NLLB-RenPy: Specialized French Gaming Translator
|
| 30 |
+
|
| 31 |
+
## 🌟 Model Description
|
| 32 |
+
|
| 33 |
+
This model is a fine-tuned version of **facebook/nllb-200-distilled-600M** specifically trained for **English-to-French translation** in gaming contexts, particularly **RenPy visual novels**.
|
| 34 |
+
|
| 35 |
+
### 🎯 Specialized For:
|
| 36 |
+
- 🎮 Gaming interfaces and menus
|
| 37 |
+
- 💬 Character dialogues and narratives
|
| 38 |
+
- 💕 Romance and emotional expressions
|
| 39 |
+
- 🔄 Interactive choices and options
|
| 40 |
+
- 📱 UI elements and notifications
|
| 41 |
+
|
| 42 |
+
### 🏆 Performance Highlights:
|
| 43 |
+
- **Superior quality** vs Google Translate/DeepL for gaming
|
| 44 |
+
- **Context-aware** translations maintaining gaming tone
|
| 45 |
+
- **Optimized** for visual novel terminology
|
| 46 |
+
- **Consistent** character voice preservation
|
| 47 |
+
|
| 48 |
+
## 🚀 Quick Start
|
| 49 |
+
|
| 50 |
+
```python
|
| 51 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
| 52 |
+
|
| 53 |
+
# Load model
|
| 54 |
+
model = AutoModelForSeq2SeqLM.from_pretrained("your-username/nllb-renpy")
|
| 55 |
+
tokenizer = AutoTokenizer.from_pretrained("your-username/nllb-renpy")
|
| 56 |
+
|
| 57 |
+
# Translate
|
| 58 |
+
text = "Hello! Welcome to our game."
|
| 59 |
+
inputs = tokenizer(text, return_tensors="pt")
|
| 60 |
+
outputs = model.generate(**inputs, forced_bos_token_id=tokenizer.convert_tokens_to_ids("fra_Latn"))
|
| 61 |
+
result = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 62 |
+
print(result) # "Bonjour! Bienvenue dans notre jeu."
|
| 63 |
+
```
|
| 64 |
+
|
| 65 |
+
📊 Training Details
|
| 66 |
+
|
| 67 |
+
Base Model: facebook/nllb-200-distilled-600M
|
| 68 |
+
Training Data: 15,000+ specialized gaming translations
|
| 69 |
+
Languages: English → French
|
| 70 |
+
Epochs: 2.0
|
| 71 |
+
Final Loss: 0.4441
|
| 72 |
+
Training Time: 2h41min
|
| 73 |
+
|
| 74 |
+
🎯 Use Cases
|
| 75 |
+
|
| 76 |
+
Perfect for translating:
|
| 77 |
+
|
| 78 |
+
✅ RenPy/Ren'Py visual novels
|
| 79 |
+
✅ Gaming interfaces and menus
|
| 80 |
+
✅ Character dialogues and stories
|
| 81 |
+
✅ Interactive fiction content
|
| 82 |
+
✅ Dating simulation games
|
| 83 |
+
|
| 84 |
+
|