はてなブックマーク - GitHub - mizorewww/laya-mlx: Native MLX runtime for Laya typed decision models — 7–14 ms short decisions on M3 Max. No text generation, PyTorch, or cloud API. https://b.hatena.ne.jp/entry/s/github.com/mizorewww/laya-mlx Open-weight typed decisions, running natively on Apple Silicon. 13.4 ms median end-to-end for a short English typed decision. 7.4 ms with the multilingual checkpoint. 0 output tokens. Local MLX inf... misshiki https://b.hatena.ne.jp/misshiki/20260927#bookmark-4793303579403156674 Apple Silicon Macでpip install laya-mlx。laya.loadでモデルを読み、predictに文とchoice・score・noulの質問を渡す。分類・採点・真偽判定をローカル処理。 2026-09-27T08:43:57Z Apple 自然言語処理 機械学習 supermomonga https://b.hatena.ne.jp/supermomonga/20260922#bookmark-4793303579403156674 2026-09-21T16:07:56Z ZAORIKU https://b.hatena.ne.jp/ZAORIKU/20260921#bookmark-4793303579403156674 2026-09-21T12:39:32Z yug1224 https://b.hatena.ne.jp/yug1224/20260921#bookmark-4793303579403156674 2026-09-21T04:51:53Z mathtonica https://b.hatena.ne.jp/mathtonica/20260921#bookmark-4793303579403156674 2026-09-21T01:35:26Z あとで読む nguyen-oi https://b.hatena.ne.jp/nguyen-oi/20260920#bookmark-4793303579403156674 Apple Siliconローカルで爆速意思決定モデル動かせるの夢がある 2026-09-20T14:09:58Z fellfield https://b.hatena.ne.jp/fellfield/20260920#bookmark-4793303579403156674 2026-09-20T14:02:30Z yuiseki https://b.hatena.ne.jp/yuiseki/20260920#bookmark-4793303579403156674 2026-09-20T09:57:13Z toaruR https://b.hatena.ne.jp/toaruR/20260920#bookmark-4793303579403156674 2026-09-20T09:11:08Z AI