A lightweight neural network library for Deno with CPU, GPU, and WASM backends, designed for serverless and edge environments.
Netsaur is a machine learning library designed for the Deno runtime, providing tools to create and train neural networks with support for CPU, GPU, and WebAssembly backends. It solves the problem of deploying ML models in serverless and edge environments by eliminating dependency management and offering portable execution. The library includes pre-built examples for common tasks like XOR, linear regression, and image classification.
Deno developers looking to integrate machine learning into applications, especially those targeting serverless or edge deployments. It's suitable for both beginners learning neural networks and experienced practitioners needing a lightweight, dependency-free ML solution.
Developers choose Netsaur for its seamless Deno integration, zero-installation design, and multi-backend flexibility. Unlike heavier ML frameworks, it offers a streamlined API optimized for modern runtimes, making it ideal for edge computing and quick prototyping without sacrificing performance.
Powerful Powerful Machine Learning library with GPU, CPU and WASM backends
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Imported directly via JSR with no dependencies, ideal for serverless environments as highlighted in the README's 'Installation' section.
Supports CPU (Rust-native) and WASM backends with GPU planned, allowing deployment from edge to serverless, evidenced in the 'Backends' section.
Built specifically for Deno with TypeScript examples, enabling seamless use in modern JavaScript/TypeScript projects without runtime conflicts.
Simple sequential network setup with layers like Dense and Sigmoid, shown in the QuickStart example to reduce boilerplate code.
GPU backend is marked as 'TODO' or 'WIP' in the README, limiting high-performance training options for production use.
Only basic neural network layers are demonstrated (e.g., dense, sigmoid), lacking support for complex architectures like convolutional or recurrent layers.
As a niche library for Deno, it has fewer contributors and resources compared to established ML frameworks, which may affect long-term support and feature development.