numrs — core library

NumRs is an experimental, high-performance numerical & Deep Learning Engine framework for Rust, inspired by NumPy and PyTorch.

This project is structured as a collection of specialized modules, each with its own specific responsibility. This README serves as a high-level index to the detailed documentation for each component.

Architecture Overview

NumRs is built on two main layers:

1. Numerical Engine: Handles arrays, types, and raw execution (SIMD/BLAS).
2. Machine Learning Framework: Built on top of the engine, providing autograd and neural networks.

graph TD
    NumRs[NumRs Core]
    
    subgraph Engine["Numerical Engine"]
        Array[Link: src/array]
        Ops[Link: src/ops]
        Backend[Link: src/backend]
    end
    
    subgraph ML["ML Framework"]
        Autograd[Link: src/autograd]
        NN[NN Modules]
    end
    
    NumRs --> Engine
    NumRs --> ML
    Autograd --> Ops
    Ops --> Backend
    Ops --> Array
    Backend --> Array

📚 Ecosystem Documentation

Select your preferred language to view the specific documentation:

Module Documentation (Internal internals)

Detailed architecture documentation for numrs-core developers:

📦 src/array

The Data Layer. Defines Array<T>, DynArray (dynamic typing), and the Type Promotion system.

🧮 src/ops

The User API. Contains the definitions for all mathematical operations (add, matmul, sum, etc.).

⚙️ src/backend

The Execution Engine. Manages the Zero-Cost Dispatch System and interfaces with hardware accelerators (MKL, BLIS, Accelerate, SIMD).

🧠 src/autograd

The ML Engine. Implements Tensor for Reverse Mode Automatic Differentiation, Neural Network layers, and Optimizers.

Quick Start

# Build with auto-detected optimizations (ASICS/BLAS)
cargo build --release

For detailed examples, see the examples/ directory.

License

AGPL-3.0-only