Nllb200 Francais Wolof

🚀 Model Card: NLLB-200 French-Wolof(🇫🇷↔️🇸🇳) Translation Model

This is a fine - tuned version of Meta's NLLB - 200 (600M distilled) model. It is specialized for French to Wolof translation, aiming to enhance the accessibility of content between these two languages.

🚀 Quick Start

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# Load model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("Lahad/nllb200-francais-wolof")
model = AutoModelForSeq2SeqLM.from_pretrained("Lahad/nllb200-francais-wolof")

# Translation function
def translate(text, max_length=128):
    inputs = tokenizer(
        text,
        max_length=max_length,
        padding="max_length",
        truncation=True,
        return_tensors="pt"
    )
    
    outputs = model.generate(
        input_ids=inputs["input_ids"],
        attention_mask=inputs["attention_mask"],
        forced_bos_token_id=tokenizer.convert_tokens_to_ids("wol_Latn"),
        max_length=max_length
    )
    
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

✨ Features

• Direct Use:
  • Facilitates text translation between French and Wolof.
  • Enables content localization.
  • Assists in language learning.
  • Supports cross - cultural communication.
• Out - of - Scope Use:
  • Commercial use without proper licensing is prohibited.
  • Not suitable for translating highly technical or specialized content.
  • Inadequate for legal or medical document translation where professional human translation is required.
  • Not designed for real - time speech translation.

📦 Installation

Since this is a Hugging Face model, you can use the transformers library to interact with it. You can install the transformers library using the following command:

pip install transformers

💻 Usage Examples

Basic Usage

# Example usage of the translation function
text = "This is a test sentence."
translation = translate(text)
print(translation)

Advanced Usage

# You can adjust the max_length parameter according to your needs
text = "This is a longer test sentence that requires more tokens to translate."
translation = translate(text, max_length = 256)
print(translation)

📚 Documentation

Model Details

Bias, Risks, and Limitations

1. Language Variety Limitations:
   • Has limited coverage of regional Wolof dialects.
   • May not handle cultural nuances effectively.
2. Technical Limitations:
   • Has a maximum context window of 128 tokens.
   • Shows reduced performance on technical/specialized content.
   • May struggle with informal language and slang.
3. Potential Biases:
   • Training data may reflect cultural biases.
   • May perform better on standard/formal language.

Recommendations

• Use the model for general communication and content translation.
• Verify translations for critical communications.
• Consider regional language variations.
• Implement human review for sensitive content.
• Test translations in the intended context before deployment.

🔧 Technical Details

Training Details

Training Data

• Dataset: galsenai/centralized_wolof_french_translation_data
• Split: 80% training, 20% testing
• Format: JSON pairs of French and Wolof translations

Training Procedure

Preprocessing

• Dynamic tokenization with padding.
• Maximum sequence length: 128 tokens.
• Source/target language tags: fr_Latn/wol_Latn.

Training Hyperparameters

• Learning rate: 2e - 5
• Batch size: 8 per device
• Training epochs: 3
• FP16 training: Enabled
• Evaluation strategy: Per epoch

Evaluation

• Testing Data: 20% of the dataset
• Metrics:
  • Cloud Provider: [Not Specified]
• Evaluation Factors:
  • Translation accuracy
  • Semantic preservation
  • Grammar correctness

Environmental Impact

• Hardware Type: NVIDIA T4 GPU
• Hours used: 5
• Cloud Provider: [Not Specified]
• Compute Region: [Not Specified]
• Carbon Emitted: [Not Calculated]

Model Architecture and Objective

• Architecture: NLLB - 200 (Distilled 600M version)
• Objective: Neural Machine Translation
• Parameters: 600M
• Context Window: 128 tokens

Compute Infrastructure

• Training Hardware: NVIDIA T4 GPU
• Training Time: 5 hours
• Software Framework: Hugging Face Transformers

📄 License

The model is licensed under CC - BY - NC - 4.0.

Model Card Contact

For questions about this model, please create an issue on the model's Hugging Face repository.