|
# asmllm1.8-v1-instruct
|
|
|
|
**A specialized language model for comprehensive understanding of Assam, India**
|
|
*Optimized for efficiency on constrained hardware*
|
|
|
|
[](https://opensource.org/licenses/MIT)
|
|
|
|
## Overview
|
|
|
|
asmllm1.8-v1-instruct is a 1.8 billion parameter language model specifically designed to provide accurate, culturally relevant information about the Indian state of Assam. Developed for practical deployment on consumer-grade hardware (tested on RTX 2050), this model offers:
|
|
|
|
- Culturally-aware responses about Assam
|
|
- Efficient resource utilization
|
|
- Clear, concise answer formulation
|
|
- Broad coverage of Assam-related topics
|
|
|
|
## Key Features
|
|
|
|
- **Hardware Efficiency**: Optimized for deployment on laptops and mid-range GPUs
|
|
- **Cultural Context**: Trained on Assam-specific data corpus
|
|
- **Practical Focus**: Prioritizes clarity over unnecessary complexity
|
|
- **Diverse Knowledge Base**: Covers geography, history, economy, and culture
|
|
|
|
## 🗃️ Training Data
|
|
|
|
Curated dataset of 2,000 records on the state of Assam,India covering:
|
|
|
|
🌄 **Geography**
|
|
- Physical features
|
|
- Climate patterns
|
|
- Natural resources
|
|
|
|
🎨 **Cultural Heritage**
|
|
- Traditional festivals (Bihu, Ambubachi Mela)
|
|
- Indigenous arts and crafts
|
|
- Local dialects and literature
|
|
|
|
💼 **Economic Landscape**
|
|
- Tea production
|
|
- Oil and natural gas reserves
|
|
- Agricultural practices
|
|
|
|
🏰 **Historical Context**
|
|
- Ancient kingdoms (Kamarupa, Chutiya)
|
|
- Colonial history (British era)
|
|
- Architectural landmarks (Kamakhya Temple, Rang Ghar)
|
|
|
|
🎓 **Educational Framework**
|
|
- Major institutions
|
|
- Educational initiatives
|
|
|
|
## ⚙️ Model Specifications
|
|
|
|
| Parameter | Value |
|
|
|----------------------|--------------------|
|
|
| 🧠 **Architecture** | Transformer-based |
|
|
| 🔢 **Parameters** | 1.8B |
|
|
| 📜 **Context Window**| 1000 tokens |
|
|
| 🎯 **Precision** | bfloat16 |
|
|
| 💻 **Hardware** | RTX 2050 (4GB VRAM)|
|
|
|
|
## 👨💻 Developed By
|
|
|
|
**Name:** Anand Dey
|
|
**Role:** Developer
|
|
**Contact:** [📧 [email protected]](mailto:[email protected]) |