github-featured / ollama--ollama

发布

0 发布 616 Git 标签

RSS 订阅源
  • v0.20.8-rc0 2cba7756c5
    比较

    Gemma4 on MLX (#15244)

    frostbyte_neo 发布于 2026-04-13 23:36:51 +00:00

    • gemma4: implement Gemma 4 model for MLX (text-only runtime)

    • gemma4: two MoE + SWA prefill perf fixes

    Two performance optimizations in the gemma4 forward pass

    1. Memoize the sliding-window prefill mask across layers.
    2. Softmax only over the selected experts in Router.Forward.
    • review comments
    下载附件