Hugging Face Overview
Overview
This model is a fine-tuned version of Meta Llama-3-8B, specifically designed for binary gender classification of Indian names. Developed by Shi-07, it leverages Supervised Fine-Tuning (SFT) with Unsloth and PEFT/LoRA to achieve high performance in identifying gender based on linguistic and cultural patterns. The model processes names and outputs a classification (0 for Male, 1 for Female).
Key Capabilities
- Specialized Gender Classification: Accurately classifies the gender of Indian names.
- Llama-3-8B Architecture: Built on the robust Llama-3-8B foundation, fine-tuned for this specific task.
- High Accuracy: Achieves 94.00% accuracy and a 0.9398 weighted F1-Score on a balanced test set of Indian names.
- Optimized Fine-tuning: Utilizes Unsloth for 2x speed optimization during fine-tuning and 4-bit quantization for efficiency.
Limitations and Considerations
- Binary Output: Strictly trained for Male (0) and Female (1) classifications; neutral names will be forced into one of these categories, typically defaulting to male.
- Indian Names Only: Performance is guaranteed only for Indian names; it is not designed for Western or other non-Indian names.
Recommended Use Case
This model is ideal for applications requiring precise, automated gender identification from a list of Indian names, such as data analysis, demographic studies, or personalized content delivery within an Indian context. Users must adhere to the specified prompt template for optimal results.