Open allsey87 opened 2 weeks ago
I believe you can use Candle with WASM to do inference in the browser.
Sure, but there is a big difference in performance between using just WASM/SIMD to using it with either WebNN or WebGPU which provide hardware acceleration options.
On Thu, 14 Nov 2024, 22:23 Akshay Ballal, @.***> wrote:
I believe you can use Candle with WASM to do inference in the browser.
— Reply to this email directly, view it on GitHub https://github.com/huggingface/candle/issues/2604#issuecomment-2477437150, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA4626JCFVM4WBYYLYFBBYL2AUIF3AVCNFSM6AAAAABRNIEDZSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINZXGQZTOMJVGA . You are receiving this because you authored the thread.Message ID: @.***>
Maybe WGPU could be an option to handle different hardware, but since CUDA and Metal are already implemented, WebGPU might be a better option for improved performance and robustness.
I was looking through issues in this repository and noticed that no one has mentioned WebNN before. Is this on any of Candle's roadmaps for doing inference in the browser?