v0.34.4-rc0:mlx:加速 Qwen 3.8 的提示處理 (#18550)
v0.34.4-rc0: mlx: speed up Qwen 3.8 prompt processing (#18550)
重點導讀
summary_zh: Ollama 版本 v0.34.4-rc0 加入 MLX 的 gated‑delta kernel,提升 Qwen 3.8 的提示處理速度。
AI 伴隨問答
針對這篇報導有疑問?可直接點選快捷問題或向 AI 追問背景脈絡與技術細節:
正文
- v0.34.4-rc0
-
c0f8da3 -
Verified
This commit was created on GitHub.com and signed with GitHub’s verified signature.
-
Choose a tag to compare
- v0.34.4-rc0
-
c0f8da3 -
Choose a tag to compare
-
Verified
This commit was created on GitHub.com and signed with GitHub’s verified signature.
* mlx: speed up Qwen 3.8 prompt processing Use MLX's gated-delta kernel for long scans and fold dense MLP global scales into SwiGLU. * address comments
Assets 2
來源: ollama release (github.com)
開啟原文