---
language:
- ta
- en
license: apache-2.0
library_name: llama.cpp
tags:
- tamil
- bilingual
- nlp
- classification
- crossberryweb
- gguf
# Specific tag to enable Inference API for GGUF
extra_gated_heading: "Tamila Model Access"
extra_gated_button_content: "Acknowledge"
---

# 🚀 Tamila Master v0.3

**Created by [crossberryweb](https://huggingface.co/crossberryweb)**

Tamila is a high-performance bilingual model (Tamil/English) trained on a massive global corpus of over 2.2 million segments.

## 🔗 Project Links
- **Live Demo (HF Space):** [https://huggingface.co/spaces/Crossberry/tamila-test-app](https://huggingface.co/spaces/Crossberry/tamila-test-app)
- **Web Deployment:** [crossberry.vercel.app](https://crossberry.vercel.app)
- **Dataset Repository:** [Hugging Face Tamila](https://huggingface.co/datasets/crossberryweb/tamila)

## 📊 Model Benchmarks
| Task | Dataset | Accuracy | Loss |
| :--- | :--- | :--- | :--- |
| Global Corpus Tuning | 2.2M Segments | 1.0000 | 6.64e-10 |
| Literature (Thirukkural) | Kaggle NLP | 0.9868 | 0.0612 |
| Technical (Kimi K2) | PDF Extract | 1.0000 | 1.17e-06 |

## 🛠 Future Roadmap
- [ ] Integration with advanced Transformer architectures.
- [ ] Expanded support for regional Tamil dialects.
- [ ] Real-time API integration for mobile applications.

## 📖 More Info
This model utilizes a custom MLP architecture optimized for GGUF deployment. It categorizes text into four primary contexts: History/Literature, Technical/AI, Tanglish, and General Corpus.

---
*Developed for the open-source community by Crossberryweb.*