nnrt 26.4.1

Neural Network Runtime

NNRT

NNRT (Neural Network Runtime) is a high-performance numerical computation library designed for modern machine learning workloads. It provides a flexible and efficient multi-dimensional array system with strong GPU acceleration, making it suitable for both research and production environments.

Overview

NNRT is built as a lightweight yet powerful alternative to traditional numerical libraries. It focuses on:

Fast tensor-like computations (similar to NumPy) Native GPU acceleration for high throughput Clean and minimal API design Scalability from small experiments to large models

Features

• Multi-dimensional Array Engine: Efficient handling of N-dimensional data structures for numerical computing.
• GPU Acceleration: Optimized backend for leveraging GPU hardware to speed up computations.
• NumPy-like API: Familiar interface for easy adoption and quick development.
• High Performance Designed for low-latency and high-throughput operations.
• Extensible Core: Modular design for future expansion (autograd, neural networks, etc.)

Installation

pip install nnrt

Quick Example

import nnrt

x1 = nnrt.Tensor([1,2])
x2 = nnrt.Tensor([1,2])

x3 = x1*x2

print(x3)

GPU Example

import nnrt

x1 = nnrt.Tensor([1,2], device="cuda")
x2 = nnrt.Tensor([1,2], device="cuda")

x3 = x1*x2

print(x3)

If you find NNRT useful, consider giving it a star on GitHub!