Hugging Face has released @huggingface/kernels, a library enabling the use of over 200 optimized WebGPU kernels for local AI applications directly from the Hugging Face Hub. These kernels cover essential machine learning operations and are published as individual, versioned packages with explicit contracts, correctness tests, and benchmark cases. To facilitate community involvement, Hugging Face also launched Fleet, a browser-based benchmarking tool that crowdsources performance and correctness data from diverse hardware, aiding in kernel improvement and optimization. AI
IMPACT Enables more efficient and accessible local AI inference within web browsers.
RANK_REASON Product launch of a library and associated tooling for running AI models in the browser.
Read on Mastodon — mastodon.social →
- Apache Software License 2.0
- Fleet
- Hugging Face Hub
- @huggingface/kernels
- WebGPU
- webgpu-kernels
- WebGPU Shading Language
- Hugging Face
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →