|
|
--- |
|
|
license: mit |
|
|
base_model: |
|
|
- deepseek-ai/DeepSeek-V3.1 |
|
|
tags: |
|
|
- dash |
|
|
- folium |
|
|
- deepseek-v3.1-4bit |
|
|
- DeepSeek |
|
|
- DeepSeek-V3.1-4bit |
|
|
- Dash |
|
|
- Folium |
|
|
--- |
|
|
# Model Card: World Explorer with DeepSeek AI |
|
|
|
|
|
## Model Details |
|
|
|
|
|
**Model Name:** World Explorer with DeepSeek AI |
|
|
**Model Type:** Geographic Information System with AI Integration |
|
|
**Architecture:** Dash Web Application with Folium Maps and DeepSeek Language Model |
|
|
**Integration:** mlx-community/DeepSeek-V3.1-4bit, Dash, Folium |
|
|
**Version:** 1.0.0 |
|
|
**Release Date:** September 5, 2025 |
|
|
**Developers:** Martin Rivera |
|
|
|
|
|
## Model Description |
|
|
|
|
|
The World Explorer with DeepSeek AI is an interactive web application that combines geographic visualization with advanced language model capabilities. It integrates: |
|
|
|
|
|
- **mlx-community/DeepSeek-V3.1-4bit**: A quantized version of the DeepSeek V3.1 language model optimized for MLX |
|
|
- **Dash**: A Python framework for building analytical web applications |
|
|
- **Folium**: A Python library for creating interactive maps and geographic visualizations |
|
|
|
|
|
This application provides an immersive experience for exploring world geography, country information, capitals, and famous landmarks through an intuitive interface powered by AI. |
|
|
|
|
|
## Intended Use |
|
|
|
|
|
This model is designed for: |
|
|
|
|
|
- Educational purposes for learning about world geography |
|
|
- Tourism planning and destination research |
|
|
- Geographic data visualization and exploration |
|
|
- Demonstrating the integration of language models with geographic information systems |
|
|
- Research in human-computer interaction with AI-powered geographic interfaces |
|
|
|
|
|
## Features |
|
|
|
|
|
### Core Capabilities |
|
|
- Interactive world map with precise geographic coordinates |
|
|
- Country information including capitals and coordinates |
|
|
- Detailed landmark data with exact geographic positions |
|
|
- AI-powered natural language queries about geographic information |
|
|
- Real-time responses to questions about countries, capitals, and landmarks |
|
|
|
|
|
### Technical Integration |
|
|
- MLX-optimized DeepSeek model for efficient inference |
|
|
- Responsive Dash web interface with Bootstrap styling |
|
|
- Folium-based interactive maps with marker clusters |
|
|
- Multi-threaded model loading for non-blocking user experience |
|
|
- Advanced query parsing for geographic context understanding |
|
|
|
|
|
## Data |
|
|
|
|
|
The application uses a comprehensive dataset of 10 countries with: |
|
|
|
|
|
- Precise coordinates for each country's capital |
|
|
- 30 famous landmarks with exact geographic coordinates |
|
|
- Curated information about each geographic feature |
|
|
|
|
|
### Countries Included: |
|
|
- France, United States, Japan, India, Brazil |
|
|
- Egypt, Australia, Italy, China, United Kingdom |
|
|
|
|
|
### Landmark Examples: |
|
|
- Eiffel Tower (48.858222, 2.2945) |
|
|
- Great Wall of China (40.68, 117.23) |
|
|
- Taj Mahal (27.175, 78.041944) |
|
|
- Statue of Liberty (40.689167, -74.044444) |
|
|
- Pyramids of Giza (29.9725, 31.128333) |
|
|
|
|
|
## Installation and Usage |
|
|
|
|
|
### Requirements |
|
|
```bash |
|
|
dash==2.14.2 |
|
|
dash-bootstrap-components==1.5.0 |
|
|
folium==0.15.1 |
|
|
pandas==2.1.4 |
|
|
numpy==1.26.4 |
|
|
mlx==0.0.6 |
|
|
transformers==4.37.2 |
|
|
``` |
|
|
|
|
|
### Installation |
|
|
```bash |
|
|
git clone <repository> |
|
|
cd world-explorer-deepseek |
|
|
pip install -r requirements.txt |
|
|
``` |
|
|
|
|
|
### Running the Application |
|
|
```bash |
|
|
python app.py |
|
|
``` |
|
|
The application will be available at `http://localhost:8050` |
|
|
|
|
|
### Model Setup |
|
|
The application requires the MLX-optimized DeepSeek model from: |
|
|
``` |
|
|
MODEL_PATH = "/Users/martinrivera/deepseek_v3_1_4bit_mlx/deepseek_v3_4bit" (Modify with your own MODEL_PATH = "path/to DeepSeek") |
|
|
``` |
|
|
|
|
|
## Performance |
|
|
|
|
|
### Response Quality |
|
|
The model provides accurate geographic information with: |
|
|
- Precise coordinate data for all locations |
|
|
- Context-aware responses based on user queries |
|
|
- Natural language understanding of geographic concepts |
|
|
- Multi-level information retrieval (country → capital → landmarks) |
|
|
|
|
|
### Efficiency |
|
|
- 4-bit quantization for reduced memory footprint |
|
|
- MLX optimization for Apple Silicon performance |
|
|
- Async model loading for responsive user experience |
|
|
- Efficient geographic data structures for quick retrieval |
|
|
|
|
|
## Limitations |
|
|
|
|
|
### Current Limitations |
|
|
- Dataset limited to 10 countries and 30 landmarks |
|
|
- Requires specific model path configuration |
|
|
- Optimized for macOS with Apple Silicon |
|
|
- English language queries only |
|
|
|
|
|
### Geographic Coverage |
|
|
While the application includes major countries and landmarks, it does not represent a complete global dataset. Users should note that: |
|
|
|
|
|
- Some countries and regions are not included |
|
|
- Landmark selection focuses on well-known tourist destinations |
|
|
- Coordinate precision varies based on source data |
|
|
|
|
|
## Ethical Considerations |
|
|
|
|
|
### Data Accuracy |
|
|
All geographic data has been verified from multiple sources, but users should: |
|
|
- Verify critical information from official sources |
|
|
- Understand that coordinates may have slight variations between sources |
|
|
- Recognize that some landmarks span large areas (e.g., Great Barrier Reef) |
|
|
|
|
|
### Privacy |
|
|
- The application does not collect user data |
|
|
- All processing happens locally when possible |
|
|
- No personal information is stored or transmitted |
|
|
|
|
|
## Future Improvements |
|
|
|
|
|
Planned enhancements include: |
|
|
- Expanded geographic coverage to more countries |
|
|
- Additional landmark categories (natural wonders, historical sites) |
|
|
- Multi-language support |
|
|
- User-contributed data features |
|
|
- Advanced visualization options (3D maps, time-based data) |
|
|
- Integration with real-time geographic data sources |
|
|
|
|
|
## Citation |
|
|
|
|
|
If you use this model in your research or projects, please cite: |
|
|
|
|
|
```bibtex |
|
|
@software{world_explorer_deepseek_2025, |
|
|
title = {World Explorer with DeepSeek AI}, |
|
|
author = {Rivera, Martin}, |
|
|
year = {2025}, |
|
|
url = {https://huggingface.co/mlx-community/DeepSeek-V3.1-4bit}, |
|
|
note = {Integration of MLX-optimized DeepSeek with Dash and Folium} |
|
|
} |
|
|
``` |
|
|
|
|
|
## License |
|
|
|
|
|
This project integrates multiple components with different licenses: |
|
|
|
|
|
- **DeepSeek Model**: Subject to DeepSeek's model license terms |
|
|
- **Dash**: MIT License |
|
|
- **Folium**: MIT License |
|
|
- **Application Code**: MIT License |
|
|
|
|
|
Users should review and comply with all respective license terms. |
|
|
|
|
|
## Contact |
|
|
|
|
|
|
|
|
For questions about the DeepSeek model: |
|
|
- Hugging Face: [mlx-community/DeepSeek-V3.1-4bit](https://huggingface.co/mlx-community/DeepSeek-V3.1-4bit) |
|
|
|
|
|
## Acknowledgements |
|
|
|
|
|
This project builds upon: |
|
|
- DeepSeek AI for the language model capabilities |
|
|
- MLX community for Apple Silicon optimization |
|
|
- Plotly for the Dash framework |
|
|
- Python Folium library for geographic visualization |
|
|
- OpenStreetMap for base map data |
|
|
|
|
|
--- |
|
|
|
|
|
*This model card was generated as part of the World Explorer with DeepSeek AI project to provide transparency about the capabilities, limitations, and appropriate use of this geographic AI application.* |
