Show HN: Greed.js – Run PyTorch code in browser using WebGPU (github.com)

Greed.js, a new library announced on Hacker News, enables the execution of PyTorch code directly in web browsers using WebGPU for GPU acceleration. This innovation allows developers to write standard Python PyTorch code that runs client-side without the need for servers or installations, effectively expanding the reach of machine learning capabilities to web applications. Unlike traditional JavaScript ML libraries, Greed.js leverages WebGPU compute shaders to execute every PyTorch operation, greatly enhancing performance and efficiency. The significance of Greed.js lies in its ability to deliver a complete PyTorch ecosystem within the browser, supporting essential functions like neural networks, tensor operations, and training workflows—all while maintaining a Python-first architecture. It integrates seamlessly with Pyodide for running Python in the browser, allowing for notebook-style execution where variable states persist between cells. The library boasts an intelligent execution strategy that automatically optimizes between GPU, CPU, and worker threads, making it robust for both experimental and production use. With features like memory management, enhanced security, and on-the-fly package installations, Greed.js represents a substantial advancement for the AI/ML community's ability to execute complex models in client-side environments.