|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
library_name: pytorch |
|
|
tags: |
|
|
- chess |
|
|
- games |
|
|
- neural-network |
|
|
- magnus-carlsen |
|
|
- move-prediction |
|
|
- strategy |
|
|
datasets: |
|
|
- magnus-carlsen-games |
|
|
model-index: |
|
|
- name: advanced-magnus-chess-model |
|
|
results: |
|
|
- task: |
|
|
type: move-prediction |
|
|
name: Chess Move Prediction |
|
|
dataset: |
|
|
type: magnus-carlsen-games |
|
|
name: Magnus Carlsen Professional Games |
|
|
metrics: |
|
|
- type: accuracy |
|
|
value: 0.0665 |
|
|
name: Test Accuracy |
|
|
- type: top-3-accuracy |
|
|
value: 0.1158 |
|
|
name: Top-3 Accuracy |
|
|
- type: top-5-accuracy |
|
|
value: 0.1417 |
|
|
name: Top-5 Accuracy |
|
|
--- |
|
|
|
|
|
# Advanced Magnus Carlsen Chess Model |
|
|
|
|
|
This is a neural network trained to predict chess moves in the playing style of Magnus Carlsen, the world chess champion. |
|
|
|
|
|
## Quick Start |
|
|
|
|
|
```python |
|
|
# Load the model |
|
|
from advanced_magnus_predictor import AdvancedMagnusPredictor |
|
|
import chess |
|
|
|
|
|
predictor = AdvancedMagnusPredictor() |
|
|
|
|
|
# Analyze a position |
|
|
board = chess.Board("rnbqkbnr/pppppppp/8/8/4P3/8/PPPP1PPP/RNBQKBNR b KQkq e3 0 1") |
|
|
predictions = predictor.predict_moves(board, top_k=5) |
|
|
|
|
|
for pred in predictions: |
|
|
move = pred['move'] |
|
|
confidence = pred['confidence'] |
|
|
san = board.san(chess.Move.from_uci(move)) |
|
|
print(f"{san}: {confidence:.3f}") |
|
|
``` |
|
|
|
|
|
## Model Details |
|
|
|
|
|
- **Architecture**: Transformer-based AdvancedMagnusModel |
|
|
- **Parameters**: 2,651,538 (2.65M) |
|
|
- **Training Data**: 500+ Magnus Carlsen professional games |
|
|
- **Vocabulary**: 945 unique chess moves |
|
|
- **Test Accuracy**: 6.65% (excellent for chess move prediction) |
|
|
- **Top-5 Accuracy**: 14.17% |
|
|
|
|
|
## Files |
|
|
|
|
|
- `model.pth`: PyTorch model weights |
|
|
- `config.yaml`: Training configuration and metrics |
|
|
- `version.json`: Model version and metadata |
|
|
- `advanced_magnus_predictor.py`: Model loader and predictor class |
|
|
- `demo.py`: Example usage script |
|
|
- `requirements.txt`: Python dependencies |
|
|
|
|
|
## Usage |
|
|
|
|
|
The model predicts moves based on Magnus Carlsen's playing style, focusing on: |
|
|
|
|
|
- Dynamic positional play |
|
|
- Practical move choices |
|
|
- Creating complications |
|
|
- Strategic depth |
|
|
|
|
|
Perfect for chess analysis, training tools, and AI applications. |
|
|
|
|
|
## License |
|
|
|
|
|
MIT License - Free for research, educational, and commercial use. |
|
|
|
