crunchdao-crunch-example 0.3.2

CrunchDAO example packages for machine learning competitions

CrunchDAO Crunch Example

This package provides example utilities and base classes for CrunchDAO machine learning competitions.

Installation

pip install crunchdao-crunch-example

Usage

Iris Classification

The package provides iris classification utilities with a dual structure:

# Import base class from namespace package
from crunchdao.crunch_example.iris import IrisModelBase
import pandas as pd

class MyIrisModel(IrisModelBase):
    def train(self, train_data: pd.DataFrame) -> None:
        # Implement your training logic here
        # train_data contains features and target labels
        pass
    
    def infer(self, dataframe: pd.DataFrame) -> pd.DataFrame:
        # Implement your inference logic here
        # dataframe contains features to predict on
        predictions = [0, 1, 2]  # Your model predictions
        
        return pd.DataFrame({
            'prediction': predictions
        })

# Use your model
model = MyIrisModel()

# Training data with features and target
train_data = pd.DataFrame({
    'sepal_length': [5.1, 4.9, 4.7],
    'sepal_width': [3.5, 3.0, 3.2],
    'petal_length': [1.4, 1.4, 1.3],
    'petal_width': [0.2, 0.2, 0.2],
    'species': [0, 0, 0]  # 0=setosa, 1=versicolor, 2=virginica
})

model.train(train_data)

# Test data with just features
test_data = pd.DataFrame({
    'sepal_length': [6.1, 5.9],
    'sepal_width': [2.9, 3.0],
    'petal_length': [4.7, 4.2],
    'petal_width': [1.4, 1.5]
})

predictions = model.infer(test_data)
print(predictions)

Package Structure

The package provides a dual structure:

Namespace Package (for base classes)

• crunchdao.crunch_example.iris - Namespace package containing IrisModelBase
  • Accessible via: from crunchdao.crunch_example.iris import IrisModelBase

Direct Package (for examples and utilities)

• crunchdao_crunch_example/ - Main package with examples and utilities
  • models/ - Example model implementations (not importable packages)
    • neural_network/ - Multi-layer perceptron with sklearn
    • random_forest/ - Ensemble tree-based classifier
    • svm/ - Support vector machine classifier
  • scripts/ - Utility scripts for training and model management (not importable packages)

Model Examples

The package includes complete model implementations:

• Neural Network (models/neural_network/) - Multi-layer perceptron with sklearn
• Random Forest (models/random_forest/) - Ensemble tree-based classifier
• SVM (models/svm/) - Support vector machine classifier

Each model includes:

• Pre-trained model files in resources/
• Implementation code in submissions/main.py
• Dependencies in submissions/requirements.txt

Utility Scripts

The scripts/ directory contains:

• pretrain_models.py - Train all models on iris dataset
• pretrain_neural_network.py - Specific neural network training
• sync_models.py - Sync models with storage systems
• cleanup_docker.py - Docker environment cleanup

Development

Requirements

• Python 3.11+
• uv (for dependency management)

Setup

# Clone the repository
git clone https://github.com/crunchdao/coordinator-setup
cd coordinator-setup/examples/crunch_examples

# Install dependencies
uv sync

# Install in development mode
uv pip install -e .

Testing

uv run pytest

Building

# Build the package
uv build

# Upload to PyPI (requires authentication)
uv publish

License

MIT License - see LICENSE file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.