‹ BackHN Continuity

Thread

Show HN: Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash

236 points · 93 comments · HenryNdubuaku

  1. sroussey · · focus · HN ↗
    Would love to see this implemented with @huggingface/kernels for shader compilation for Webgpu.
    1. HenryNdubuaku · · focus · HN ↗
      noted, we'd look into this, thanks
      1. sroussey · · focus · HN ↗
        For reference: <a href="https:&#x2F;&#x2F;huggingface.co&#x2F;blog&#x2F;webgpu-kernels" rel="nofollow">https:&#x2F;&#x2F;huggingface.co&#x2F;blog&#x2F;webgpu-kernels

        I think it will be the basis for a rewrite of transformers.js v5, but no need for you to wait as you would likely want direct access. It is also way better than loading WASM, and faster to boot!

Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.