A minimalist model introspection laboratory
Illuminating your models with elegance and precision
Lucefact is a beautiful, futuristic web application designed for comprehensive machine learning model analysis. With a minimalist design philosophy and powerful TypeScript architecture, it provides deep insights into model architectures across multiple frameworks.
.h5,.keras- TensorFlow/Keras models.pt,.pth- PyTorch models.onnx- ONNX (Open Neural Network Exchange).pkl,.pickle- Scikit-learn and Python pickled models
# Clone or navigate to the project directory
cd Lucefact
# Install dependencies
npm install
# Start development server
npm run devThe application will open automatically at http://localhost:3000
npm run build
npm run preview- Vertical flow diagram of model layers
- Text-only nodes with glowing connections
- Interactive zoom and pan controls
- Hover tooltips with detailed information
-
Architecture Summary
- Total layers and parameters
- Average parameters per layer
- Most complex layer identification
-
Layer Distribution
- Layer type frequency analysis
- Activation function usage
- Architecture pattern detection
-
Parameter Density
- Trainable vs non-trainable ratio
- Layer size distribution
- Complexity metrics
-
Potential Issues
- Missing metadata detection
- Unbalanced architectures
- Resource requirement warnings
npm run type-check- Create a new parser in
src/parsers/ - Implement the
parse(file: File): Promise<ParseResult>method - Register in
src/parser.ts - Add file extension support
Example:
export class CustomParser {
async parse(file: File): Promise<ParseResult> {
// Your parsing logic
return { success: true, metadata };
}
}- Chrome/Edge (latest)
- Firefox (latest)
- Safari (latest)
Lucefact — illuminating your models
Version: Alpha v0.4