Skip to content

C-ML Documentation Index

Welcome to the C-ML documentation. This index provides an overview of all available documentation.

Getting Started

  • README - Project overview and quick start
  • Training Guide - Start here to learn how to train neural networks with automatic metrics tracking
  • Visualization UI - Interactive training dashboard and model visualization

Core Documentation

Automatic Differentiation

  • Autograd System - Complete guide to automatic differentiation

  • Overview of the autograd system

  • API reference for gradient computation
  • Usage examples and best practices

  • Technical implementation details

  • Autograd Implementation - Technical deep dive

  • Detailed implementation notes

  • Design decisions and architecture
  • Performance considerations
  • Advanced features

Neural Network Layers

  • Neural Network Layers - Complete layer reference

  • Available layers (Linear, Conv2d, BatchNorm2d, LayerNorm, Pooling, etc.)

  • Layer usage examples
  • API reference

  • Implementation status

  • Layers Implementation - Implementation details

  • Layer implementation status

  • Testing recommendations
  • Feature completeness

Training

  • Training Guide - Comprehensive training guide

  • Model definition

  • Parameter collection
  • Optimizer usage
  • Loss functions
  • Complete training loop examples
  • Training Metrics - Automatic tracking of training, validation, and test metrics
  • Visualization - Export metrics to JSON for interactive visualization
  • Best practices

Training Metrics

The C-ML library includes built-in training metrics tracking that works automatically:

  • Automatic Epoch Timing - Tracks time per epoch and total training time (no manual code needed)
  • Loss and Accuracy Tracking - Automatically records training, validation, and test metrics
  • Gradient Norm Monitoring - Tracks gradient health during training
  • Learning Rate Tracking - Monitors learning rate changes and scheduler information
  • Loss Reduction Rate - Computes percentage reduction in loss
  • Loss Stability - Calculates standard deviation of recent losses
  • Early Stopping Support - Tracks early stopping status (actual vs expected epochs)
  • LR Scheduler Visualization - Displays scheduler type and parameters in UI
  • Real-time JSON Export - Continuously exports metrics to training.json for visualization (when VIZ=1 or CML_VIZ=1)

All metrics are captured automatically when using cml_init() and cml_cleanup(). See Training Guide for detailed usage.

Visualization UI

C-ML includes an interactive web-based visualization UI:

  • Training Results Dashboard - Real-time visualization of training metrics
  • Computational Graph Visualization - Visual representation of ops topology
  • Model Architecture View - Interactive model structure visualization using Cytoscape
  • Bias-Variance Analysis - Plot training, validation, and test metrics together
  • Early Stopping Visualization - Visual indicators for early stopping with actual vs expected epochs
  • LR Scheduler Display - Show scheduler type and parameters in metrics panel
  • Automatic Launch - Set VIZ=1 to automatically launch before program runs

See README for setup and usage.

Integration

  • Integration Summary - Library integration overview
  • Build system integration
  • Dependency management
  • Component integration
  • Training metrics integration

Development

  • Implementation Status - Current implementation status
  • Completed features
  • Recently added features
  • Implementation notes
  • TODO Implementations - Planned features and improvements
  • High priority items
  • Medium priority items
  • Low priority items
  • Performance optimizations

Documentation Structure

docs/
├── INDEX.md                      # This file
├── AUTOGRAD.md                   # Autograd system guide
├── AUTOGRAD_IMPLEMENTATION.md   # Autograd technical details
├── NN_LAYERS.md                  # Neural network layers reference
├── LAYERS_COMPLETE.md            # Layer implementation details
├── TRAINING.md                   # Training guide (with metrics)
├── INTEGRATION_SUMMARY.md       # Integration overview
├── IMPLEMENTATION_STATUS.md     # Implementation status
└── TODO_IMPLEMENTATIONS.md       # Future work

For Beginners

  1. Start with README for an overview
  2. Read Training Guide for a complete example with metrics
  3. Explore Neural Network Layers for available layers
  4. Try the Visualization UI to see your training progress

For Advanced Users

  1. Review Autograd Implementation for technical details
  2. Check Layers Implementation for implementation status
  3. See Implementation Status for current features
  4. Use Training Metrics for comprehensive monitoring

For Contributors

  1. Review Integration Summary for build system details
  2. Check TODO Implementations for planned work
  3. See Implementation Status for current state
  4. Follow contribution guidelines in README

Examples

Example code is available in the examples/ directory and root:

  • main.c - Simple XOR classification example
  • examples/test.c - Comprehensive training with train/val/test splits and automatic metrics
  • examples/early_stopping_lr_scheduler.c - Early stopping and learning rate scheduling example
  • examples/autograd_example.c - Autograd system examples
  • examples/training_loop_example.c - Training loop pattern demonstration
  • examples/export_graph.c - Graph export for visualization

API Reference

For detailed API documentation, see:

  • include/cml.h - Main library header with inline documentation
  • include/tensor/ - Tensor operations
  • include/autograd/ - Automatic differentiation
  • include/nn/ - Neural network components
  • include/optim/ - Optimizers
  • include/Core/training_metrics.h - Training metrics API

All header files include comprehensive inline documentation using Doxygen-style comments.

Key Features

Core Library

  • Automatic differentiation (autograd)
  • Neural network layers
  • Optimizers (SGD, Adam)
  • Loss functions
  • Tensor operations

Training Utilities

  • Training Metrics - Built-in automatic metrics tracking (no manual code needed)
  • Automatic Timing - Epoch time calculation
  • Gradient Monitoring - Gradient norm tracking
  • Early Stopping - Track early stopping status (actual vs expected epochs)
  • LR Scheduling - Display scheduler type and parameters in UI
  • Real-time JSON Export - Continuously export metrics for visualization (when VIZ=1 or CML_VIZ=1)
  • Centralized Cleanup - CleanupContext for resource management
  • Dataset Splitting - dataset_split_three() for train/val/test splits
  • Automatic Evaluation - training_metrics_evaluate_dataset() for validation/test evaluation

Visualization

  • Interactive Dashboard - Real-time training visualization
  • Graph Visualization - Computational graph and model architecture
  • Bias-Variance Analysis - Training/validation/test metrics comparison

Support

For issues, questions, or contributions, please refer to the project repository.