llama.cpp 发布包含服务器改进和性能优化 · 跟踪 8 个来源

作者 PulseAugur 编辑部 · [13 个来源] · 2026-08-07 10:03

llama.cpp 项目发布了多个更新，包括 b10331 版本，该版本通过正确报告隔离工作目录来改进服务器功能。其他近期版本，如 b10330 及更早版本，则侧重于 CUDA 操作的性能优化、SYCL 的错误修复以及跨 macOS、Linux、Android 和 Windows 等各种平台的通用系统兼容性。这些更新反映了 llama.cpp 库在高效本地 LLM 部署方面的持续开发和完善。 AI

影响 llama.cpp 的持续更新提高了本地 LLM 推理的性能和兼容性。

排序理由该集群包含 llama.cpp 项目的多个次要版本说明，详细说明了错误修复和次要性能改进，而不是重大的新模型或功能发布。

在 llama.cpp — Releases 阅读 →

AI 生成摘要 · Google Gemini · 来自 13 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 0 / 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 项目的多个次要版本说明，详细说明了错误修复和次要性能改进，而不是重大的新模型或功能发布。
Source corroboration: 13 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: 53 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.
Coverage growth since scoring: +5 source(s) since last score; New sources have picked up this story since our last re-score. Score will update on the next scoring pass.

完整方法见我们的编辑标准。

报道来源 [13]

llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-10 10:53

b10338

<details open=""> model-saver : fix expert shared/chunk FFN length key clobber (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26693">#26693</a>) The saver called add_kv with LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH twice, the …
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-10 08:19

b10336

<details open=""> ggml-webgpu : refactor several wgsl files and simplify flash_attn wgsl. (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26134">#26134</a>) </details> Website: <ul> <li><a href="https://llama.ap…
llama.cpp — Releases TIER_1 (SO) · ServeurpersoCom · 2026-08-09 19:20

b10335

ui: degrade the working directory picker when file search is off (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26">#26</a>…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-09 11:23

b10333

<details open=""> ggml-cpu : fix missing Q5_0 dispatch in SpaceMiT backend (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26792">#26792</a>) </details> Website: <ul> <li><a href="https://llama.app" rel="nofollo…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-09 10:50

b10332

<details open=""> ci: rm <code>GGML_HIP_ROCWMMA_FATTN</code> (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26760">#26760</a>) Signed-off-by: Aaron Teo <a href="mailto:[email protected]">[email protected]</a> </details> <s…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-08 23:28

b10331

<details open=""> server: report the isolate working directory from get_info (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26773">#26773</a>) <ul> <li>server: report the isolate working directory from get_info</li> </ul> Without a…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-08 17:22

b10330

<details open=""> CUDA: fuse rms_norm + mul + rope (+ view + set_rows) (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26767">#26767</a>) <ul> <li> CUDA: fuse rms_norm + mul + rope (+ view + set_rows) </li> <li> tests: add br…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-07 14:07

b10308

<details open=""> Mitigate crashing issue on Windows MSYS2 UCRT64 environment (GCC 16.1.0) (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26555">#26555</a>) </details> Website: <ul> <li><a href="https://llama.a…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-07 13:22

b10307

<details open=""> sycl: fix UE4M3 parsing (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/25608">#25608</a>) The NVFP4 quantization format stores a scaling factor for every group of 16 weights, packed into a single UE4M3 byte.…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-07 12:35

b10306

<details open=""> sycl: *glu flat path (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26354">#26354</a>) <ul> <li>tests: add SWIGLU perf cases</li> </ul> perf mode had no GLU coverage. Adds SWIGLU at 17408 columns, 512 and 20…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-07 12:05

b10305

<details open=""> sycl : Support DSv4 OPs: LIGHTNING_INDEXER,DSV4_HC_COMB,DSV4_HC_POST,DSV4_HC_PRE (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26568">#26568</a>) <ul> <li> support DSv4 OPs: LIGHTNING_INDEXER,DSV4_HC_COMB,DSV4_HC…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-07 11:17

b10303

<details open=""> sycl : fix error Error OP FLASH_ATTN_EXT on arc770 (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26441">#26441</a>) </details> Website: <ul> <li><a href="https://llama.app" rel="nofollow">htt…
llama.cpp — Releases TIER_1 (SO) · github-actions[bot] · 2026-08-07 10:03

b10301

<details open=""> cuda: fix warnings for unused variable/function (<a class="issue-link js-issue-link" href="https://github.com/ggml-org/llama.cpp/pull/26688">#26688</a>) </details> Website: <ul> <li><a href="https://llama.app" rel="nofollow">https:…

本文如何被排名

报道来源 [13]

相关实体

相关话题