Skip to content

chore: bump llama.cpp to b10333 - #31

Merged
leehack merged 1 commit into
mainfrom
automation/bump-llama-cpp
Aug 10, 2026
Merged

leehack merged 1 commit into
mainfrom
automation/bump-llama-cpp

Conversation

@leehack

@leehack leehack commented Aug 6, 2026 •

Copy link
Copy Markdown
Owner

llama.cpp update

Upstream changelog

Release notes for b10333
Details

ggml-cpu : fix missing Q5_0 dispatch in SpaceMiT backend (#26792)

Website:

macOS/iOS:

Linux:

Android:

Windows:

openEuler:

  • DISABLED
  • openEuler x86 (310p)
  • openEuler x86 (910b, ACL Graph)
  • openEuler aarch64 (310p)
  • openEuler aarch64 (910b, ACL Graph)

UI:

Commit range

Commits from b10276 to b10333 (first 80)
  • ui: Update vulnerable packages + cleanup Storybook config (#26607) (e031d95)
  • build : remove GGML_METAL_USE_BF16 from all build scripts (#26604) (d52ec04)
  • ui: show generation statistics by default in chat settings (#26624) (3e3a7a4)
  • vendor : apply patches for subprocess.h (#26606) (61881b1)
  • convert: Add endianness conversion for Q1 and TQ2 quantizations (#26618) (020760a)
  • server: Adding spec-decode counters to /metrics endpoint (#26389) (a035a88)
  • security : clarify about AI-generated reports (#26579) (9303cdd)
  • fit: Fix memory allocation for MTP layers (#26605) (9a688e5)
  • mtmd: support multi-row batching for deepseek-ocr (#26154) (717dad5)
  • grammar : degrade max repetition >= 2000 to unbounded (#26613) (cd0fa60)
  • mtmd: Unlimited-OCR fix max_tiles, setting in converter (#25614) (b06aa77)
  • tests: re-enable MiniMax M3 in test-llama-archs (#26633) (360e134)
  • server: harden the file_glob_search directory walk (#26626) (f9e832c)
  • mtmd/ggml: add ggml_build_forward_order (#26649) (c8e03ce)
  • vulkan: fix submission batching size, add debug tools for diagnosing causes of DeviceLost drivers errors (#26371) (803b7fc)
  • model-conversion : add --model-name to conversion scripts (#26665) (9de0fcf)
  • ci : onboard AMD ROCm CI with gfx1151 fixes (#26544) (a1f96d4)
  • convert : accept "ExaoneMoeForCausalLM" arch spelling (#26660) (e700bfb)
  • model-loader : fix quantized reshaped tensor strides (#26672) (3db4ff8)
  • convert : fix DeepseekV4 rope parameters with transformers 5.x (#26673) (eb5667a)
  • server: fix empty response for /cors-proxy (#26656) (6a32c29)
  • mtmd: add chunk save/load function (#26645) (15586e2)
  • metal : avoid threadgroup matrix array instantiation in kernel_lightning_indexer (#26646) (e40bf88)
  • ci: abort if build requirements are missing (#26368) (d8d9887)
  • cuda: fix warnings for unused variable/function (#26688) (5b87ed3)
  • sycl : enhance OP set_rows to support all missed data types (#26515) (c074cb3)
  • sycl : fix error Error OP FLASH_ATTN_EXT on arc770 (#26441) (eef5f3e)
  • sycl : update guide Q&A and script for device setting (#26442) (c1f4109)
  • sycl : Support DSv4 OPs: LIGHTNING_INDEXER,DSV4_HC_COMB,DSV4_HC_POST,DSV4_HC_PRE (#26568) (31558db)
  • sycl: *glu flat path (#26354) (6b5c2ef)
  • sycl: fix UE4M3 parsing (#25608) (fc3f10b)
  • Mitigate crashing issue on Windows MSYS2 UCRT64 environment (GCC 16.1.0) (#26555) (42e9881)
  • ui: read model modalities from the router model list (#26709) (82bb485)
  • ggml : add aarch64 HWCAP fallbacks and fix fp16 variant detection (#25554) (cb26014)
  • mtmd: stop feeding the text stream again during Qwen3-TTS generation (#26706) (217df17)
  • server: (router) do not evict busy models (#26567) (e1470ee)
  • server: (router) add LRU scheduler (#26572) (dff15d4)
  • ui: set npm min-release-age to protect against supply-chain attacks (#26711) (34e9ee5)
  • webui: load the model selected via ?model= when ?load=true (#26707) (f4f7758)
  • server : clarify comment in eval_llama_cmpl_schema [no ci] [no release] (#26720) (9ba73fd)
  • ggml : bump version to 0.19.0 (ggml/1581) (933f46f)
  • sync : ggml (4cf5cab)
  • mtmd: fix longest_edge ignoring min/max pixels (#26638) (4cb22cd)
  • ui: Filesystem @mentions for Chat Form (#26715) (2363478)
  • metal : fix NORM/RMS_NORM for row lengths that leave a partial simdgroup (#26708) (a194a75)
  • sycl: coalesce the ssm_conv window loads (#26612) (f8e3026)
  • allozaur/feat/chat slash commands (#26716) (6de1b63)
  • tests : speed-up server test suite 3x (#26734) (1621a3d)
  • allozaur/feat/chat form contenteditable (#26717) (fc6545d)
  • tts: account for the vocoder pass in the timings line (#26733) (3653e6d)
  • CUDA: fix thread/block count in quantized cpy kernel launches (#26731) (69bf643)
  • server: add initial tool isolation support (via docker) (#26507) (dd2c7c4)
  • server, ui: only offer a working directory when a tool reads it (#26762) (18f7ad7)
  • CUDA: fuse rms_norm + mul + rope (+ view + set_rows) (#26767) (687e778)
  • server: report the isolate working directory from get_info (#26773) (7ba604f)
  • ci: rm GGML_HIP_ROCWMMA_FATTN (#26760) (61141f1)
  • ggml-cpu : fix missing Q5_0 dispatch in SpaceMiT backend (#26792) (0865990)

Web bridge review focus

Please pay extra attention to upstream changes touching:

  • WebGPU, WASM, Emscripten, pthreads, or memory64 build behavior
  • ggml backend APIs used by the bridge
  • model loading, tokenizer, chat template, context/state persistence, or cache semantics
  • CMake/build flags that can affect the generated JS/WASM artifacts

Validation

  • Emscripten build passed
  • Browser WebGPU/state-persistence smoke passed
  • Generated bridge artifacts include wasm32 and memory64 outputs
  • No stale hard-coded llama.cpp tag remains in CI/publish defaults

Automation behavior

This PR is managed from the stable branch automation/bump-llama-cpp. If another llama.cpp release appears before merge, the scheduled workflow updates this same PR instead of opening a duplicate. The workflow skips if a non-automation PR already changes llama_cpp.version.

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Updates the repository’s pinned llama.cpp source tag so CI/build scripts that resolve llama_cpp.version will fetch and build against upstream b10290 instead of b10276.

Changes:

  • Bump llama.cpp pin from b10276 to b10290.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

@leehack
leehack force-pushed the automation/bump-llama-cpp branch from bfef268 to 9111954 Compare August 7, 2026 05:09
@leehack leehack changed the title chore: bump llama.cpp to b10290 chore: bump llama.cpp to b10298 Aug 7, 2026
@leehack
leehack force-pushed the automation/bump-llama-cpp branch from 9111954 to cc50ab9 Compare August 8, 2026 04:33
@leehack leehack changed the title chore: bump llama.cpp to b10298 chore: bump llama.cpp to b10326 Aug 8, 2026
@leehack leehack changed the title chore: bump llama.cpp to b10326 chore: bump llama.cpp to b10331 Aug 9, 2026
@leehack
leehack force-pushed the automation/bump-llama-cpp branch from cc50ab9 to 6dfa794 Compare August 9, 2026 04:42
@leehack
leehack force-pushed the automation/bump-llama-cpp branch from 6dfa794 to e7c31cf Compare August 10, 2026 05:00
@leehack leehack changed the title chore: bump llama.cpp to b10331 chore: bump llama.cpp to b10333 Aug 10, 2026
@leehack
leehack merged commit afccd54 into main Aug 10, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants