llama.cpp项目发布预发布更新，包含多项修复和优化

作者 PulseAugur 编辑部 · [12 个来源] · 2026-10-08 07:21

llama.cpp项目已发布数个预发布更新，其中包括b11514，该更新解决了Musa FWHT修复和针对多个操作系统的平台证明问题。其他更新如b11512和b11511分别侧重于模型修复和CUDA优化。值得注意的是，b11509包含对CUDA CCCL版本保护的修复，而b11505解决了Vulkan TOP_K在处理无限和NaN输入时的问题。这些更新还包括了多位开发者的贡献以及与cpp-httplib等库的集成。 AI

影响这些更新提高了在各种平台上运行LLM的性能和稳定性。

排序理由该集群包含llama.cpp项目的多个预发布更新，llama.cpp是一个用于运行大型语言模型的软件工具。

在 llama.cpp — Releases 阅读 →

AI 生成摘要 · Google Gemini · 来自 12 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 1 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含llama.cpp项目的多个预发布更新，llama.cpp是一个用于运行大型语言模型的软件工具。
Source corroboration: 12 independent sources; Strong cross-source corroboration — multiple independent publishers covered this within the clustering window.
Topics: product, infra; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Same-day; Cluster formed today. Ranking reflects the current source set at time of score.

完整方法见我们的编辑标准。

报道来源 [12]

llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-10-08 20:17

b11514

<details open=""> Musa FWHT fix (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/30167">#30167</a>) </details> Website: <ul> <li><a href="https://llama.app" rel="nofollow">https://llama.app</a></li> </ul> <str…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-10-08 17:42

b11512

<details open=""> model : fix DFlash output head sharing (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/30111">#30111</a>) <ul> <li>llama : fix DFlash output head sharing</li> </ul> Assisted-by: Codex <ul> <li>dflash : read tie…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-10-08 17:27

b11511

<details open=""> CUDA: fix MMQ out-of-bounds reads (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/29953">#29953</a>) </details> Website: <ul> <li><a href="https://llama.app" rel="nofollow">https://llama.app</a…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-10-08 16:20

b11510

<details open=""> CUDA : looped PAD kernel for more than 65535 rows or slices (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/30147">#30147</a>) </details> Website: <ul> <li><a href="https://llama.app" rel="nofo…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-10-08 15:53

b11509

<details open=""> CUDA: fix CCCL version guard breaking on major version rollover (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/29453">#29453</a>) <ul> <li>CUDA: fix CCCL version guard breaking on major version rollover</li> </ul> <p…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-10-08 15:31

b11507

<details open=""> llama: support MoE cache over multiple GPUs (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/30112">#30112</a>) </details> Website: <ul> <li><a href="https://llama.app" rel="nofollow">https://ll…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-10-08 14:41

b11505

<details open=""> vulkan : fix TOP_K for +inf/NaN inputs and k = 1 on negative values (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/30107">#30107</a>) The bucket search in topk_nary_search.comp started from the range [0, 0xF…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-10-08 13:32

b11503

<details open=""> vendor : update cpp-httplib to 0.60.1 (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/30134">#30134</a>) Signed-off-by: Adrien Gallouët <a href="mailto:[email protected]">[email protected]</a> </details> <p…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-10-08 13:00

b11501

<details open=""> sycl: FWHT optimizations (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/29605">#29605</a>) </details> Website: <ul> <li><a href="https://llama.app" rel="nofollow">https://llama.app</a></li> </…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-10-08 12:41

b11500

<details open=""> hex-mmadd: do not assume aligned read/write when bias-add is fused (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/30133">#30133</a>) </details> Website: <ul> <li><a href="https://llama.app" re…
llama.cpp — Releases TIER_1 (SO) · jeffbolznv · 2026-10-08 09:36

b11504

vulkan: extend sparse FA support to coopmat2 (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/30003">#30003</a>)
llama.cpp — Releases TIER_1 (SO) · saady789 · 2026-10-08 07:21

b11502

cuda : support arbitrary striding for unary ops on f16, f32, and bf16…

本文如何被排名

报道来源 [12]

相关实体

相关话题