index.md

Ember ML Documentation

Welcome to the Ember ML documentation. Ember ML is a hardware-optimized neural network library that supports multiple backends (PyTorch, MLX, NumPy) to run efficiently on different hardware platforms (CUDA, Apple Metal, and other platforms).

Documentation Sections

• API Reference: Detailed API documentation for all modules
  • Frontend Usage Guide: Comprehensive guide on using the Ember ML frontend
  • Tensor Architecture: Detailed explanation of the tensor operations architecture
• Architecture: System architecture and design principles
  • Ember ML Architecture: Comprehensive overview of the Ember ML architecture
  • Function-First Design: Detailed explanation of the function-first design pattern
• Tutorials: Step-by-step guides for common tasks
• Examples: Code examples and use cases
• Plans: Development plans and roadmaps

Quick Start

Installation

pip install ember-ml

Basic Usage

import ember_ml
from ember_ml.nn.tensor import EmberTensor
from ember_ml import ops

# Set the backend
ember_ml.backend.set_backend('mlx')  # or 'torch' or 'numpy'

# Create a tensor
tensor = EmberTensor([[1, 2, 3], [4, 5, 6]])

# Perform operations
result = ops.matmul(tensor, EmberTensor([[1], [2], [3]]))
print(result)  # EmberTensor([[14], [32]])

For more detailed instructions, see the Getting Started guide and the Frontend Usage Guide.

Key Features

• Hardware-Optimized Neural Networks: Implementation of cutting-edge neural network architectures optimized for different hardware platforms
• Multi-Backend Support: Backend-agnostic tensor operations that work with PyTorch, MLX, NumPy, and other computational backends
• Function-First Design: Efficient memory usage through separation of functions from class implementations
• Liquid Neural Networks: Design and implementation of liquid neural networks and other advanced architectures
• Neural Circuit Policies: Biologically-inspired neural architectures with custom wiring configurations

Core Components

Neural Network Architectures

The project implements various cutting-edge neural network architectures:

• Liquid Neural Networks (LNN): Dynamic networks with adaptive connectivity
• Neural Circuit Policies (NCP): Biologically-inspired neural architectures
• Stride-Aware Continuous-time Fully Connected (CfC) networks
• Specialized attention mechanisms and temporal processing units

For more details, see the Architecture Documentation.

Multi-Backend Support

The project implements backend-agnostic tensor operations that can use different computational backends:

• MLX (optimized for Apple Silicon)
• PyTorch (for CUDA and other GPU platforms)
• NumPy (for CPU computation)
• Future support for additional backends

Feature Extraction

The project includes tools for extracting features from large datasets, including:

• TerabyteFeatureExtractor: Extracts features from large datasets
• TemporalStrideProcessor: Processes temporal data with variable strides

Getting Help

If you encounter any issues or have questions:

1. Check the tutorials and examples in this documentation
2. Search for similar issues in the GitHub repository
3. Ask a question in the Discussion forum

License

Ember ML is released under the MIT License.