{"id":31625,"name":"llama.cpp","ecosystem":"submodules","repository_url":null,"issues_count":41,"created_at":"2025-06-07T10:37:03.269Z","updated_at":"2025-06-07T10:37:03.269Z","purl":"pkg:submodules/llama.cpp","unique_repositories_count":5,"unique_repositories_count_past_30_days":1,"recent_issues":[{"uuid":"4505731598","node_id":"PR_kwDOSDCeKc7egGk7","number":29,"state":"closed","title":"Bump llama.cpp from `5d14e5d` to `1acee6b`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-05-26T03:08:35.000Z","author_association":null,"state_reason":null,"created_at":"2026-05-22T21:37:02.000Z","updated_at":"2026-05-26T03:08:37.000Z","time_to_close":279093,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`5d14e5d`","new_version":"`1acee6b`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `5d14e5d` to `1acee6b`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1acee6bf8939948f9bcbf4b14034e4b475f06069\"\u003e\u003ccode\u003e1acee6b\u003c/code\u003e\u003c/a\u003e server: only parse empty msg if continuing an assistant msg (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23506\"\u003e#23506\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ef570f63087b6a5a2930210a13f87990e8113927\"\u003e\u003ccode\u003eef570f6\u003c/code\u003e\u003c/a\u003e perplexity : fix integer overflow (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23496\"\u003e#23496\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/cc9e331213b6a9cb186aabe01a4ec6a61419dd80\"\u003e\u003ccode\u003ecc9e331\u003c/code\u003e\u003c/a\u003e SYCL: improve MoE prefill throughput (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23142\"\u003e#23142\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bcfd1989e9a90af74669d94057ff2468682c3f4a\"\u003e\u003ccode\u003ebcfd198\u003c/code\u003e\u003c/a\u003e sycl : Level Zero detection in ggml_sycl_init (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23097\"\u003e#23097\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/56f16f235c4a6ffd0cd316e1d4b5dcfbf2dcb7a4\"\u003e\u003ccode\u003e56f16f2\u003c/code\u003e\u003c/a\u003e SYCL : gated_delta_net K\u0026gt;1 (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23174\"\u003e#23174\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/8cc67efcd4834a46b18a0cf32c9b1c99762daeac\"\u003e\u003ccode\u003e8cc67ef\u003c/code\u003e\u003c/a\u003e SYCL: add BF16 to DMMV kernel path (~4x tg speedup on Intel Arc) (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21580\"\u003e#21580\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/95feeab52e41ceaf71e87b2dd01895f6d8815b60\"\u003e\u003ccode\u003e95feeab\u003c/code\u003e\u003c/a\u003e docs: Update documentation with Granite 4.0/4.1 (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23404\"\u003e#23404\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/99d4026b116605ed8e1f3ab179b3c63bc4637195\"\u003e\u003ccode\u003e99d4026\u003c/code\u003e\u003c/a\u003e ggml-zendnn : add Q8_0 quantization support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23414\"\u003e#23414\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9c92e96a64fe0f03f5f3e5ab720a151941da1de5\"\u003e\u003ccode\u003e9c92e96\u003c/code\u003e\u003c/a\u003e cmake : build router app only during standalone builds (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23521\"\u003e#23521\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/afcda09d154a285cd366135f98ffc1d357f7ddbd\"\u003e\u003ccode\u003eafcda09\u003c/code\u003e\u003c/a\u003e vocab : fix HybridDNA tokenizer (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23466\"\u003e#23466\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/5d14e5d19bd6af7fc38eb92d96aa185e5948a03d...1acee6bf8939948f9bcbf4b14034e4b475f06069\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/MiloDevs/go-llama.cpp/pull/29","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/MiloDevs%2Fgo-llama.cpp/issues/29","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/29/packages"},{"uuid":"4490664723","node_id":"PR_kwDOSDCeKc7dvMRO","number":27,"state":"closed","title":"Bump llama.cpp from `5d14e5d` to `6a257d4`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-05-21T21:37:37.000Z","author_association":null,"state_reason":null,"created_at":"2026-05-20T23:56:07.000Z","updated_at":"2026-05-21T21:37:39.000Z","time_to_close":78090,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`5d14e5d`","new_version":"`6a257d4`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `5d14e5d` to `6a257d4`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6a257d44633d4a752183ed778b88d2924d0a6b9d\"\u003e\u003ccode\u003e6a257d4\u003c/code\u003e\u003c/a\u003e mtmd, model : merge HunyuanOCR into HunyuanVL and fix OCR vision precision (#...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3a479c9132072815cb70a443b4efa45bb66b3f59\"\u003e\u003ccode\u003e3a479c9\u003c/code\u003e\u003c/a\u003e ui: Add max image size option (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22849\"\u003e#22849\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ad277572619fcfb6ddd38f4c6437283a4b2b8636\"\u003e\u003ccode\u003ead27757\u003c/code\u003e\u003c/a\u003e Move to backend sampling for MTP draft path (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23287\"\u003e#23287\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3a6db741a8189a45260536581f4ebb0a7f051f3c\"\u003e\u003ccode\u003e3a6db74\u003c/code\u003e\u003c/a\u003e opencl: refactor backend initilization (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23318\"\u003e#23318\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/510b5c2a35652390c71327ecb29c2fb14bfe0e8c\"\u003e\u003ccode\u003e510b5c2\u003c/code\u003e\u003c/a\u003e common/speculative : fix nullptr crash in get_devices_str (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23386\"\u003e#23386\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a8681a0ed2e3f2a7452e642639671bcec20b865c\"\u003e\u003ccode\u003ea8681a0\u003c/code\u003e\u003c/a\u003e mtmd : DeepSeek-OCR image processing fixes, img_tool::resize padding refactor...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/acd604fb277044e07c2bff01f4c169167b45f478\"\u003e\u003ccode\u003eacd604f\u003c/code\u003e\u003c/a\u003e vulkan: optimize operations in the IM2COL shader (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22685\"\u003e#22685\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6ce96713de33e3e4c1599025c0bacf3c3e524c6a\"\u003e\u003ccode\u003e6ce9671\u003c/code\u003e\u003c/a\u003e feat: Add WAV MIME type variants and improve audio format detection (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23396\"\u003e#23396\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c9872a2575acc65834deb15a1f5155f6dbc75229\"\u003e\u003ccode\u003ec9872a2\u003c/code\u003e\u003c/a\u003e hexagon: HMX quantized matmul rework (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23368\"\u003e#23368\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e947228222147356bc7e64154d3439e142481632\"\u003e\u003ccode\u003ee947228\u003c/code\u003e\u003c/a\u003e Programmatic Dependent Launch (PDL) for more performance on newer NVIDIA GPUs...\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/5d14e5d19bd6af7fc38eb92d96aa185e5948a03d...6a257d44633d4a752183ed778b88d2924d0a6b9d\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/MiloDevs/go-llama.cpp/pull/27","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/MiloDevs%2Fgo-llama.cpp/issues/27","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/27/packages"},{"uuid":"4299204512","node_id":"PR_kwDOSDCeKc7UIHD1","number":5,"state":"closed","title":"Bump llama.cpp from `5d14e5d` to `9789512`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-21T21:37:19.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-20T23:07:23.000Z","updated_at":"2026-04-21T21:37:21.000Z","time_to_close":80996,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`5d14e5d`","new_version":"`9789512`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `5d14e5d` to `9789512`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/97895129e5f2bde94d13dc01ca41ee79e9b629f2\"\u003e\u003ccode\u003e9789512\u003c/code\u003e\u003c/a\u003e ggml-cuda: flush legacy pool on OOM and retry (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22155\"\u003e#22155\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/86f8daacfe5e202b99ed396aa574a8b41a982048\"\u003e\u003ccode\u003e86f8daa\u003c/code\u003e\u003c/a\u003e mtmd: correct get_n_pos / get_decoder_pos (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22175\"\u003e#22175\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/cf8b0dbda9ac0eac30ee33f87bc6702ead1c4664\"\u003e\u003ccode\u003ecf8b0db\u003c/code\u003e\u003c/a\u003e server : remove /api endpoints (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22165\"\u003e#22165\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/fd6ae4ca1cd5446442f6c2e5e73a2a4c9bc44993\"\u003e\u003ccode\u003efd6ae4c\u003c/code\u003e\u003c/a\u003e Tensor-parallel: Fix delayed AllReduce on Gemma-4 MoE (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22129\"\u003e#22129\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/fb19f94c715c466230c72d2a32822f8a9e113708\"\u003e\u003ccode\u003efb19f94\u003c/code\u003e\u003c/a\u003e TP: fix 0-sized tensor slices, AllReduce fallback (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21808\"\u003e#21808\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7f251fdbce614a50141005dc70ce3787b7777a8e\"\u003e\u003ccode\u003e7f251fd\u003c/code\u003e\u003c/a\u003e ggml-cpu: Optimized x86 and generic cpu q1_0 dot (follow up) (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21636\"\u003e#21636\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a6cc43c286a2ebc429aa69b9a4d16de082cedb51\"\u003e\u003ccode\u003ea6cc43c\u003c/code\u003e\u003c/a\u003e ggml-webgpu: updated matrix-vector multiplication (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21738\"\u003e#21738\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a678916623ddef89c2a43776df24e00a52b17638\"\u003e\u003ccode\u003ea678916\u003c/code\u003e\u003c/a\u003e mtmd: refactor mtmd_decode_use_mrope (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22161\"\u003e#22161\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/81df3f7cfaa6f99de14e792b38d5771bf427383e\"\u003e\u003ccode\u003e81df3f7\u003c/code\u003e\u003c/a\u003e fix: GLM-DSA crash in llama-tokenize when using vocab_only (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22102\"\u003e#22102\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/de71b5f81c3b6b9f8bdaf1b2a21198e1eede3fda\"\u003e\u003ccode\u003ede71b5f\u003c/code\u003e\u003c/a\u003e server : refactor \u0026quot;use checkpoint\u0026quot; logic (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22114\"\u003e#22114\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/5d14e5d19bd6af7fc38eb92d96aa185e5948a03d...97895129e5f2bde94d13dc01ca41ee79e9b629f2\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/MiloDevs/go-llama.cpp/pull/5","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/MiloDevs%2Fgo-llama.cpp/issues/5","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/5/packages"},{"uuid":"4275850052","node_id":"PR_kwDOQit6R87S-HXe","number":102,"state":"closed","title":"build(deps): bump llama.cpp from `0893f50` to `b572d1e`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-17T12:55:07.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-16T12:55:31.000Z","updated_at":"2026-04-17T12:55:08.000Z","time_to_close":86376,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`0893f50`","new_version":"`b572d1e`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `0893f50` to `b572d1e`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b572d1ecd62210229e04cdeffd3ae80dd59f0921\"\u003e\u003ccode\u003eb572d1e\u003c/code\u003e\u003c/a\u003e codeowners: add team member comments (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21714\"\u003e#21714\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/03b3d077988dffdf08bf628fab78904526745115\"\u003e\u003ccode\u003e03b3d07\u003c/code\u003e\u003c/a\u003e Convert: Fix NemotronH Config Parsing (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21664\"\u003e#21664\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3f7c29d318e317b63f54c558bc69803963d7d88c\"\u003e\u003ccode\u003e3f7c29d\u003c/code\u003e\u003c/a\u003e ggml: add graph_reused (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21764\"\u003e#21764\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ae2d34899e2a9a172c7f2090ed4dd366bbf25d0d\"\u003e\u003ccode\u003eae2d348\u003c/code\u003e\u003c/a\u003e metal: Implement ROLL op (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21946\"\u003e#21946\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1e796eb41fb51950ada45811a303e57a5f4ea974\"\u003e\u003ccode\u003e1e796eb\u003c/code\u003e\u003c/a\u003e ggml-cpu: add 128-bit RVV implementation for Quantization Vector Dot (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/20633\"\u003e#20633\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/5637536517ae4ed3eaa22b39c0d479e049097a9b\"\u003e\u003ccode\u003e5637536\u003c/code\u003e\u003c/a\u003e ggml : implemented simd_gemm kernel for riscv vector extension (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/20627\"\u003e#20627\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/90fb96a7b3c3dd97420d49603aefe773a612c05a\"\u003e\u003ccode\u003e90fb96a\u003c/code\u003e\u003c/a\u003e devops : added spirv-headers to nix (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21965\"\u003e#21965\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/82677a6ede7927d2286ef1c9e481ce4caf52866f\"\u003e\u003ccode\u003e82677a6\u003c/code\u003e\u003c/a\u003e ggml-webgpu: compute pass batching and removing profiling overhead (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21873\"\u003e#21873\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/8612ed18b7d2896009f255c11eb002aa7bfa9057\"\u003e\u003ccode\u003e8612ed1\u003c/code\u003e\u003c/a\u003e ci : Use ggml-org/ccache-action on RISC-V as well (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21632\"\u003e#21632\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b1be68e8cab67f5c2fb7d8e3e90291a8805ece0e\"\u003e\u003ccode\u003eb1be68e\u003c/code\u003e\u003c/a\u003e [SYCL] Fix Q8_0 reorder: garbage on 2nd prompt + crash on full VRAM (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21638\"\u003e#21638\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/0893f50f2dc14fcc046e10d4f76a1ac7a62c0490...b572d1ecd62210229e04cdeffd3ae80dd59f0921\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/102","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/102","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/102/packages"},{"uuid":"4249310905","node_id":"PR_kwDOQ9H31s7RzZDd","number":24,"state":"closed","title":"build(deps): bump llama.cpp from `67a2209` to `1e9d771`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-19T18:12:42.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-12T18:12:46.000Z","updated_at":"2026-04-19T18:12:43.000Z","time_to_close":604796,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`67a2209`","new_version":"`1e9d771`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `67a2209` to `1e9d771`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1e9d771e2c2f1113a5ebdd0dc15bafe57dce64be\"\u003e\u003ccode\u003e1e9d771\u003c/code\u003e\u003c/a\u003e convert : force f16 or f32 on step3-vl conv weights (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21646\"\u003e#21646\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/aa4695c5e5bf0abda8942c08e94cb804a7ea0347\"\u003e\u003ccode\u003eaa4695c\u003c/code\u003e\u003c/a\u003e mtmd: add gemma 4 test (vision + audio) [no ci] (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21806\"\u003e#21806\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/547765a93e5ad7b4e8ca84d78f6d83f36ad8ee25\"\u003e\u003ccode\u003e547765a\u003c/code\u003e\u003c/a\u003e mtmd: add Gemma 4 audio conformer encoder support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21421\"\u003e#21421\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9e209c5aee8effac146463c8dc32984a4b4d2672\"\u003e\u003ccode\u003e9e209c5\u003c/code\u003e\u003c/a\u003e fix: Proper messages rendering for \u0026quot;Show raw output\u0026quot; (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21672\"\u003e#21672\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6313acbef016d9e4d8e83d3647082329949d958b\"\u003e\u003ccode\u003e6313acb\u003c/code\u003e\u003c/a\u003e docs: add guide on how to add multimodal support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21778\"\u003e#21778\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ff5ef8278615a2462b79b50abdf3cc95cfb31c6f\"\u003e\u003ccode\u003eff5ef82\u003c/code\u003e\u003c/a\u003e CUDA: skip compilation of superfluous FA kernels (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21768\"\u003e#21768\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/073bb2c20b5b2c919469653214aaa1a9895816a2\"\u003e\u003ccode\u003e073bb2c\u003c/code\u003e\u003c/a\u003e mtmd : add MERaLiON-2 multimodal audio support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21756\"\u003e#21756\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/af1127d3c49e41a606bac7c2b3897489aa71b918\"\u003e\u003ccode\u003eaf1127d\u003c/code\u003e\u003c/a\u003e opencl: add basic support for q5_k (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21593\"\u003e#21593\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/865ff06b2ffa2f91f30cbec1f8c73d66cc6642aa\"\u003e\u003ccode\u003e865ff06\u003c/code\u003e\u003c/a\u003e TP: fix Qwen 3 Next data split (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21732\"\u003e#21732\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2b2cd57de64e96c9ebfca6ba7e6bdbab6fe51482\"\u003e\u003ccode\u003e2b2cd57\u003c/code\u003e\u003c/a\u003e ggml : fix a few instances of missing GGML_TYPE_Q1_0 cases (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21716\"\u003e#21716\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/67a2209fabe2e3498d458561933d5380655085d2...1e9d771e2c2f1113a5ebdd0dc15bafe57dce64be\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/Jota-project/jota-inference/pull/24","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/Jota-project%2Fjota-inference/issues/24","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/24/packages"},{"uuid":"4208255933","node_id":"PR_kwDOQ9H31s7QEP9P","number":22,"state":"closed","title":"build(deps): bump llama.cpp from `67a2209` to `5d3a4a7`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-12T18:12:48.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-05T18:12:42.000Z","updated_at":"2026-04-12T18:12:49.000Z","time_to_close":604806,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`67a2209`","new_version":"`5d3a4a7`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `67a2209` to `5d3a4a7`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/5d3a4a7da5e3dd42f5922aba2fe21b520e96e830\"\u003e\u003ccode\u003e5d3a4a7\u003c/code\u003e\u003c/a\u003e server : fix logging of build + system info (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21460\"\u003e#21460\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c08d28d08871715fd68accffaeeb76ddcaede658\"\u003e\u003ccode\u003ec08d28d\u003c/code\u003e\u003c/a\u003e ci: lower cuda12 floor to 12.8.1 for broader host compatibility (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21438\"\u003e#21438\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/661e9acb36768d0d4ddb6f2eb674fbb1be185823\"\u003e\u003ccode\u003e661e9ac\u003c/code\u003e\u003c/a\u003e ci: fix vulkan workflow referencing non-existent action (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21442\"\u003e#21442\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b8635075ffe27b135c49afb9a8b5c434bd42c502\"\u003e\u003ccode\u003eb863507\u003c/code\u003e\u003c/a\u003e common : add gemma 4 specialized parser (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21418\"\u003e#21418\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9c699074c97191754c8a966298f84c79f90fce38\"\u003e\u003ccode\u003e9c69907\u003c/code\u003e\u003c/a\u003e server: Fix undefined timing measurement errors in server context (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21201\"\u003e#21201\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d01f6274c01e111be2ccc39443f79884796e48fb\"\u003e\u003ccode\u003ed01f627\u003c/code\u003e\u003c/a\u003e common : respect specified tag, only fallback when tag is empty (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21413\"\u003e#21413\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/650bf14eb9a922de0f88c9523a271159cc5ae469\"\u003e\u003ccode\u003e650bf14\u003c/code\u003e\u003c/a\u003e llama-model: read final_logit_softcapping for Gemma 4 (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21390\"\u003e#21390\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b7ad48ebda2287c778fd826606d7b3b3570f60ab\"\u003e\u003ccode\u003eb7ad48e\u003c/code\u003e\u003c/a\u003e llama: add custom newline split for Gemma 4 (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21406\"\u003e#21406\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d006858316d4650bb4da0c6923294ccd741caefd\"\u003e\u003ccode\u003ed006858\u003c/code\u003e\u003c/a\u003e ggml-webgpu: move from parameter buffer pool to single buffer with offsets (#...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e43970099269b5b6da36b8977ad47697602e4e54\"\u003e\u003ccode\u003ee439700\u003c/code\u003e\u003c/a\u003e ci: Add Windows Vulkan backend testing on Intel (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21292\"\u003e#21292\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/67a2209fabe2e3498d458561933d5380655085d2...5d3a4a7da5e3dd42f5922aba2fe21b520e96e830\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/Jota-project/jota-inference/pull/22","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/Jota-project%2Fjota-inference/issues/22","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/22/packages"},{"uuid":"4200436429","node_id":"PR_kwDOQit6R87Py9zK","number":93,"state":"closed","title":"build(deps): bump llama.cpp from `8710e5f` to `887535c`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-06T12:57:05.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-03T12:54:31.000Z","updated_at":"2026-04-06T12:57:07.000Z","time_to_close":259354,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`8710e5f`","new_version":"`887535c`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `8710e5f` to `887535c`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/887535c33f9e3bf57532f31bc3d749b264751a2b\"\u003e\u003ccode\u003e887535c\u003c/code\u003e\u003c/a\u003e ci: add more binary checks (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21349\"\u003e#21349\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d3416a4aa9a37d9a0ca547e18c0e126bfe8a07ea\"\u003e\u003ccode\u003ed3416a4\u003c/code\u003e\u003c/a\u003e fix: remove stale assert (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21369\"\u003e#21369\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/43a4ee4a2cf25de0428d618544b877731d4d3713\"\u003e\u003ccode\u003e43a4ee4\u003c/code\u003e\u003c/a\u003e HIP: build eatch ci build test for a different architecture (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21337\"\u003e#21337\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f851fa5ab056c9bada48ad7208fe122fc0574e44\"\u003e\u003ccode\u003ef851fa5\u003c/code\u003e\u003c/a\u003e fix: add openssl to nix dependencies (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21353\"\u003e#21353\u003c/a\u003e) (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21355\"\u003e#21355\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f1ac84119ccc8e72dafd9e9f8fc3b9399917ce11\"\u003e\u003ccode\u003ef1ac841\u003c/code\u003e\u003c/a\u003e ggml-zendnn : add MUL_MAT_ID op support for MoE models (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21315\"\u003e#21315\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b069b10ab48f25ba119e59d0b8bf35d4f06e093f\"\u003e\u003ccode\u003eb069b10\u003c/code\u003e\u003c/a\u003e vocab: fix Gemma4 tokenizer (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21343\"\u003e#21343\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/0c58ba3365d2bc717b447b5d70e4d6be09ff3c40\"\u003e\u003ccode\u003e0c58ba3\u003c/code\u003e\u003c/a\u003e rpc : reuse compute graph buffers (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21299\"\u003e#21299\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/57ace0d612a11133ac86edcc7af1b323bf05f12f\"\u003e\u003ccode\u003e57ace0d\u003c/code\u003e\u003c/a\u003e chat : avoid including json in chat.h (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21306\"\u003e#21306\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/39b27f0da0271c06986cb31b68bc0fe68e780616\"\u003e\u003ccode\u003e39b27f0\u003c/code\u003e\u003c/a\u003e (revert) kv-cache : do not quantize SWA KV cache (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21332\"\u003e#21332\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f49e9178767d557a522618b16ce8694f9ddac628\"\u003e\u003ccode\u003ef49e917\u003c/code\u003e\u003c/a\u003e ci : add AMD ZenDNN label to PR labeler (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21345\"\u003e#21345\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/8710e5f9b9bd7246608808ccd3626bde8abf6ff9...887535c33f9e3bf57532f31bc3d749b264751a2b\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/93","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/93","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/93/packages"},{"uuid":"4194414495","node_id":"PR_kwDOQit6R87PjY4K","number":92,"state":"closed","title":"build(deps): bump llama.cpp from `8710e5f` to `e15efe0`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-03T12:54:33.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-02T12:55:17.000Z","updated_at":"2026-04-03T12:54:35.000Z","time_to_close":86356,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`8710e5f`","new_version":"`e15efe0`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `8710e5f` to `e15efe0`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e15efe007dc1c0d79afa347190dba91de3bd659b\"\u003e\u003ccode\u003ee15efe0\u003c/code\u003e\u003c/a\u003e Relax prefill parser to allow space. (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21240\"\u003e#21240\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6137c325a16073c8bf68a52396a815006ccaa9a9\"\u003e\u003ccode\u003e6137c32\u003c/code\u003e\u003c/a\u003e chat : add Granite 4.0 chat template with correct tool_call role mapping (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/20\"\u003e#20\u003c/a\u003e...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/17193cce34036a6488b092ca79313d4ee1f895f5\"\u003e\u003ccode\u003e17193cc\u003c/code\u003e\u003c/a\u003e kv-cache : do not quantize SWA KV cache (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21277\"\u003e#21277\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d6dac92bfdf6797b74eff25493c5e525561f70fb\"\u003e\u003ccode\u003ed6dac92\u003c/code\u003e\u003c/a\u003e Ignore Transfer-Encoding header. (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/20269\"\u003e#20269\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/dae2bf41c91a0d8eea7b6c7ded08d452eb8aeb79\"\u003e\u003ccode\u003edae2bf4\u003c/code\u003e\u003c/a\u003e sync : ggml\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bc07d559224260f65f40595121d9e2ebe60ee99e\"\u003e\u003ccode\u003ebc07d55\u003c/code\u003e\u003c/a\u003e ggml : bump version to 0.9.11 (ggml/1456)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/4888137b1736b706e39806025d24e4ca342f1e4a\"\u003e\u003ccode\u003e4888137\u003c/code\u003e\u003c/a\u003e sycl : fix llama_kv_cache hang when kv_cache is huge: 5GB (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21283\"\u003e#21283\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/fbd441c37933550c1e3365dc84dd73232334c15d\"\u003e\u003ccode\u003efbd441c\u003c/code\u003e\u003c/a\u003e hexagon : add cumsum op support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21246\"\u003e#21246\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c30e012253dd9e322c8e3424f808a5c74ecc46bf\"\u003e\u003ccode\u003ec30e012\u003c/code\u003e\u003c/a\u003e contrib : rewrite AGENTS.md, make it more clear about project values (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21270\"\u003e#21270\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/95a6ebabb277c4cc18247e7bc2a5502133caca63\"\u003e\u003ccode\u003e95a6eba\u003c/code\u003e\u003c/a\u003e opencl: fix leak in Adreno q8_0 path (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21212\"\u003e#21212\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/8710e5f9b9bd7246608808ccd3626bde8abf6ff9...e15efe007dc1c0d79afa347190dba91de3bd659b\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/92","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/92","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/92/packages"},{"uuid":"3968435681","node_id":"PR_kwDOQit6R87FKPts","number":63,"state":"closed","title":"build(deps): bump llama.cpp from `e6267a9` to `b908baf`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-02-23T13:46:52.000Z","author_association":null,"state_reason":null,"created_at":"2026-02-20T12:55:07.000Z","updated_at":"2026-02-23T13:46:54.000Z","time_to_close":262305,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`e6267a9`","new_version":"`b908baf`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `e6267a9` to `b908baf`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b908baf1825b1a89afef87b09e22c32af2ca6548\"\u003e\u003ccode\u003eb908baf\u003c/code\u003e\u003c/a\u003e ggml-cpu: add RVV vec dot kernels for quantization types (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18784\"\u003e#18784\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/492bc319782b1f13f302911f4c73437382cc8bb9\"\u003e\u003ccode\u003e492bc31\u003c/code\u003e\u003c/a\u003e quantize : add --dry-run option (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19526\"\u003e#19526\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/77d6ae4ac89bb879ada3989a748276dfe4553674\"\u003e\u003ccode\u003e77d6ae4\u003c/code\u003e\u003c/a\u003e test: mul_mat tests with huge batch size (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19519\"\u003e#19519\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/10b26ee23a2d1b563a62db1ea4710cf8b723791a\"\u003e\u003ccode\u003e10b26ee\u003c/code\u003e\u003c/a\u003e WebUI hide models in router mode (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19374\"\u003e#19374\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3dadc88b589ca43b8fca0e1beb22d4b78a09b4dd\"\u003e\u003ccode\u003e3dadc88\u003c/code\u003e\u003c/a\u003e common : fix Step-3.5-Flash format detection and thinking support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19635\"\u003e#19635\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/39e4b1dc9bd00eb21a4e9cc6950855f94bc66de0\"\u003e\u003ccode\u003e39e4b1d\u003c/code\u003e\u003c/a\u003e common : fix gpt-oss Jinja error when assistant message has both content and ...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/11c325c6e0666a30590cde390d5746a405e536b9\"\u003e\u003ccode\u003e11c325c\u003c/code\u003e\u003c/a\u003e ggml-webgpu: Add unary op (SQR, SQRT, SIN, COS) support. (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19700\"\u003e#19700\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/237958db339300bdd8028608cc08b2ba2685ec33\"\u003e\u003ccode\u003e237958d\u003c/code\u003e\u003c/a\u003e model: Add PaddleOCR-VL model support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18825\"\u003e#18825\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/abb9f3c42b5e6acee9e8e37836ef691d1a41bdb8\"\u003e\u003ccode\u003eabb9f3c\u003c/code\u003e\u003c/a\u003e vulkan: fix MMQ shader push constants and multi-dispatch (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19732\"\u003e#19732\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/da348c9dfbcfab16584f4640ee53146fdf85a741\"\u003e\u003ccode\u003eda348c9\u003c/code\u003e\u003c/a\u003e models : fix qwen3.5 beta/gate shapes (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19730\"\u003e#19730\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eSee full diff in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/e6267a935901313dc727ec74d159fc66e206e9c4...b908baf1825b1a89afef87b09e22c32af2ca6548\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/63","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/63","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/63/packages"},{"uuid":"3847341072","node_id":"PR_kwDOQit6R86-7PoZ","number":43,"state":"closed","title":"build(deps): bump llama.cpp from `b70d251` to `cb6caca`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-01-26T14:04:15.000Z","author_association":null,"state_reason":null,"created_at":"2026-01-23T12:56:05.000Z","updated_at":"2026-01-26T14:04:16.000Z","time_to_close":263290,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`b70d251`","new_version":"`cb6caca`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `b70d251` to `cb6caca`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/cb6caca191b9a3a9a4eaa13dd9e465225d127034\"\u003e\u003ccode\u003ecb6caca\u003c/code\u003e\u003c/a\u003e [SYCL] use malloc to support both iGPU and dGPU in same time (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18992\"\u003e#18992\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b5b8fa1c8b3b27683b2965a22f9985eec683d384\"\u003e\u003ccode\u003eb5b8fa1\u003c/code\u003e\u003c/a\u003e chat : fix translategemma crash on common_chat_format_example (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19019\"\u003e#19019\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a14b960bc70a0b48405409bbe3e0d6238473a0f8\"\u003e\u003ccode\u003ea14b960\u003c/code\u003e\u003c/a\u003e model-conversion : use BUILD_DIR variable in all scripts (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19015\"\u003e#19015\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/091a46cb8d43c0e662d04b80a3d11320d25b7d49\"\u003e\u003ccode\u003e091a46c\u003c/code\u003e\u003c/a\u003e ggml-cpu: aarm64: q5_K repack gemm and gemv (and generic) implementations (i8...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a3e812811d8f12f4236efa41287dc3dcd5c3c2f6\"\u003e\u003ccode\u003ea3e8128\u003c/code\u003e\u003c/a\u003e cli : load parser definition (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19031\"\u003e#19031\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/51fa458a92d6a3f305f8fd76fc8f702e3e87ddb5\"\u003e\u003ccode\u003e51fa458\u003c/code\u003e\u003c/a\u003e server : support preserving reasoning_content in assistant message (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18994\"\u003e#18994\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a5eaa1d6a3732bc0f460b02b61c95680bba5a012\"\u003e\u003ccode\u003ea5eaa1d\u003c/code\u003e\u003c/a\u003e mla : make the V tensor a view of K (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18986\"\u003e#18986\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e2baf02162382a14c9f4fc15d7681a715256453c\"\u003e\u003ccode\u003ee2baf02\u003c/code\u003e\u003c/a\u003e CUDA: fix alignment check for FA (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19023\"\u003e#19023\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e34d6d03b25d9e8d07f3bd0190b27d0d01a7e416\"\u003e\u003ccode\u003ee34d6d0\u003c/code\u003e\u003c/a\u003e convert_hf_to_gguf.py: refactor modify_tensors to call super (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18866\"\u003e#18866\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9c96465f99e47a3a568c50969ff5c6b672ab2714\"\u003e\u003ccode\u003e9c96465\u003c/code\u003e\u003c/a\u003e opencl: enable the general fp mm for non-cont input and as a fallback for spe...\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/b70d251076ac7c3ac1cd5d39dbb167f6ff3b6880...cb6caca191b9a3a9a4eaa13dd9e465225d127034\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/43","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/43","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/43/packages"},{"uuid":"3696163650","node_id":"PR_kwDOQit6R863JhQJ","number":1,"state":"open","title":"Bump llama.cpp from `ac43576` to `bde188d`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":2,"pull_request":true,"closed_at":null,"author_association":null,"state_reason":null,"created_at":"2025-12-04T19:26:06.000Z","updated_at":"2025-12-04T19:29:19.000Z","time_to_close":null,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`ac43576`","new_version":"`bde188d`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `ac43576` to `bde188d`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bde188d60f58012ada0725c6dd5ba7c69fe4dd87\"\u003e\u003ccode\u003ebde188d\u003c/code\u003e\u003c/a\u003e metal: TRI, FILL, EXPM1, SOFTPLUS (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/16623\"\u003e#16623\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9d0229967a0538840368547ee7ddc637fc28142d\"\u003e\u003ccode\u003e9d02299\u003c/code\u003e\u003c/a\u003e server: strip content-length header on proxy (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17734\"\u003e#17734\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c4c10bfb86569ccb070d0dbe1a621a8f186baa16\"\u003e\u003ccode\u003ec4c10bf\u003c/code\u003e\u003c/a\u003e server: move msg diffs tracking to HTTP thread  (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17740\"\u003e#17740\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/817d743cc17cf644dab8408eb0f1e6eac89562c1\"\u003e\u003ccode\u003e817d743\u003c/code\u003e\u003c/a\u003e examples : add missing code block end marker [no ci] (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17756\"\u003e#17756\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bd4ef134763d81e251fd097019578f2df571dfef\"\u003e\u003ccode\u003ebd4ef13\u003c/code\u003e\u003c/a\u003e common : skip model validation when --help is requested (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17755\"\u003e#17755\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/87a2084c45188d54a554c305a397e778759545ed\"\u003e\u003ccode\u003e87a2084\u003c/code\u003e\u003c/a\u003e ggml-cpu : remove asserts always evaluating to false (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17728\"\u003e#17728\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3659aa28e963ef3f782cd27258e97ddef678c776\"\u003e\u003ccode\u003e3659aa2\u003c/code\u003e\u003c/a\u003e convert: use existing local chat_template if mistral-format model has one. (#...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2a73f81f8a810783db5794256e5ba79f298adee7\"\u003e\u003ccode\u003e2a73f81\u003c/code\u003e\u003c/a\u003e cmake : simplify build info detection using standard variables (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17423\"\u003e#17423\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7dba049b0707ae395c59b085c5fd52cae7b74fe0\"\u003e\u003ccode\u003e7dba049\u003c/code\u003e\u003c/a\u003e ci : disable ggml-ci-x64-amd-* (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17753\"\u003e#17753\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/83c1171529a30c5e018779339690e21430aae372\"\u003e\u003ccode\u003e83c1171\u003c/code\u003e\u003c/a\u003e common: use native MultiByteToWideChar (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17738\"\u003e#17738\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/ac43576124a75c2de6e333ac31a3444ff9eb9458...bde188d60f58012ada0725c6dd5ba7c69fe4dd87\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/1","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/1","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/1/packages"},{"uuid":"3605942282","node_id":"PR_kwDOOKZoKM6ybK-f","number":31,"state":"closed","title":"Bump llama.cpp from `2be60cb` to `b8595b1`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2025-11-17T03:19:58.000Z","author_association":null,"state_reason":null,"created_at":"2025-11-10T03:31:52.000Z","updated_at":"2025-11-17T03:20:00.000Z","time_to_close":604086,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`b8595b1`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `b8595b1`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b8595b16e69e3029e06be3b8f6635f9812b2bc3f\"\u003e\u003ccode\u003eb8595b1\u003c/code\u003e\u003c/a\u003e mtmd : fix embedding size for image input (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17123\"\u003e#17123\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/392e09a60852d0e879d4bbedd5ace3e6852f719e\"\u003e\u003ccode\u003e392e09a\u003c/code\u003e\u003c/a\u003e vulkan: fix memory allocations (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17122\"\u003e#17122\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/802cef44bfaa80987076d621c8bf5875627c197b\"\u003e\u003ccode\u003e802cef4\u003c/code\u003e\u003c/a\u003e convert : parse safetensors directly (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15667\"\u003e#15667\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1c07c0c68c692d39b83f491bad9447af852bb652\"\u003e\u003ccode\u003e1c07c0c\u003c/code\u003e\u003c/a\u003e convert : handle compressed-tensors quant method (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17069\"\u003e#17069\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/cb1adf885105da7ce23db746b4202f4e987aa3e8\"\u003e\u003ccode\u003ecb1adf8\u003c/code\u003e\u003c/a\u003e server : handle failures to restore host cache (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17078\"\u003e#17078\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ef1d8269972bd086bee2554fd31a865c6da84f33\"\u003e\u003ccode\u003eef1d826\u003c/code\u003e\u003c/a\u003e benches : add folder with benchmarks (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16931\"\u003e#16931\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/86fde91e62c3f72ab7ed8a540dc1be049b735477\"\u003e\u003ccode\u003e86fde91\u003c/code\u003e\u003c/a\u003e Switch to using Ubuntu 25.10 vulkan/mesa (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16497\"\u003e#16497\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7f3e9d339c99d96d6df9833c63ec27dbbc96f003\"\u003e\u003ccode\u003e7f3e9d3\u003c/code\u003e\u003c/a\u003e vulkan: iGPU memory reporting fix (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17110\"\u003e#17110\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/8a3519b70898b07ec05c391418a05aaa6b377c83\"\u003e\u003ccode\u003e8a3519b\u003c/code\u003e\u003c/a\u003e vulkan: fix mmq out of bounds reads (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17108\"\u003e#17108\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/80a6cf63473b95742444a1b27d45164591282a7d\"\u003e\u003ccode\u003e80a6cf6\u003c/code\u003e\u003c/a\u003e vulkan: fuse mul_mat_id + mul (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17095\"\u003e#17095\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...b8595b16e69e3029e06be3b8f6635f9812b2bc3f\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/31","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/31","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/31/packages"},{"uuid":"3580402939","node_id":"PR_kwDOOKZoKM6xGfhG","number":30,"state":"closed","title":"Bump llama.cpp from `2be60cb` to `7e99416`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2025-11-10T03:31:55.000Z","author_association":null,"state_reason":null,"created_at":"2025-11-03T03:27:28.000Z","updated_at":"2025-11-10T03:31:56.000Z","time_to_close":605067,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`7e99416`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `7e99416`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7e994168b1ccc12337ba8de939c4fd466107c1fb\"\u003e\u003ccode\u003e7e99416\u003c/code\u003e\u003c/a\u003e SYCL: optimized repeat_back kernel (3× fewer asm instructions, 2× faster)Feat...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bcfa87622ae46be6345a8e3dfdbdc5ba5414042b\"\u003e\u003ccode\u003ebcfa876\u003c/code\u003e\u003c/a\u003e feat(webui): improve LaTeX rendering with currency detection (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16508\"\u003e#16508\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a2054e3a8ff0da3978a4acc18c349ff58554d336\"\u003e\u003ccode\u003ea2054e3\u003c/code\u003e\u003c/a\u003e test-backend-ops : fix segfault in moe-expert-reduce test in support mode and...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/dd5286805004db1f9ac3176a1cbbfe373bdda0f8\"\u003e\u003ccode\u003edd52868\u003c/code\u003e\u003c/a\u003e ci : disable failing riscv cross build (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16952\"\u003e#16952\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6b9a52422bac0f50dd8f1f8386744fa3ce9783bf\"\u003e\u003ccode\u003e6b9a524\u003c/code\u003e\u003c/a\u003e model: add Janus Pro for image understanding (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16906\"\u003e#16906\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2f966b8ed87514e74bb96592217226cb6a6974dd\"\u003e\u003ccode\u003e2f966b8\u003c/code\u003e\u003c/a\u003e clip : use FA (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16837\"\u003e#16837\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/cd5e3b57541ecc52421130742f4d89acbcf77cd4\"\u003e\u003ccode\u003ecd5e3b5\u003c/code\u003e\u003c/a\u003e server : support unified cache across slots (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16736\"\u003e#16736\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/87c9efc3b297b8a498716b1db3d061842e6fc85b\"\u003e\u003ccode\u003e87c9efc\u003c/code\u003e\u003c/a\u003e common : move gpt-oss reasoning processing to init params (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16937\"\u003e#16937\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/76af40aaaad78c42faecd8016a88362c788b84b0\"\u003e\u003ccode\u003e76af40a\u003c/code\u003e\u003c/a\u003e docs: remove llama_sampler_accept reference in sampling sample usage (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16920\"\u003e#16920\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7db35a7958a943be1693879f42d166f152613979\"\u003e\u003ccode\u003e7db35a7\u003c/code\u003e\u003c/a\u003e CUDA: add FLOOR, CEIL, ROUND, TRUNC unary ops (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16917\"\u003e#16917\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...7e994168b1ccc12337ba8de939c4fd466107c1fb\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/30","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/30","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/30/packages"},{"uuid":"3554906682","node_id":"PR_kwDOOKZoKM6vyucH","number":29,"state":"closed","title":"Bump llama.cpp from `2be60cb` to `75cbdd3`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2025-11-03T03:27:31.000Z","author_association":null,"state_reason":null,"created_at":"2025-10-27T03:32:29.000Z","updated_at":"2025-11-03T03:27:32.000Z","time_to_close":604502,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`75cbdd3`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `75cbdd3`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/75cbdd3fce38ea12d50cd19e73a069aa5dbbd5fa\"\u003e\u003ccode\u003e75cbdd3\u003c/code\u003e\u003c/a\u003e test-backend-ops: print failed tests at the end (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16785\"\u003e#16785\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2b9bd9bf4e759c05db629ec1c391dc8aeaa71887\"\u003e\u003ccode\u003e2b9bd9b\u003c/code\u003e\u003c/a\u003e sycl: add ROLL operation support (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16665\"\u003e#16665\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/59fc1ec8e83b14354c1a3a8acf8c5c2cbf9af42f\"\u003e\u003ccode\u003e59fc1ec\u003c/code\u003e\u003c/a\u003e sycl: add REPEAT_BACK operation support (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16734\"\u003e#16734\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/75d33b9302f84a5b89f82205d2bcd8def5a64e0a\"\u003e\u003ccode\u003e75d33b9\u003c/code\u003e\u003c/a\u003e CUDA: support for weight clamp in top-k norm (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16702\"\u003e#16702\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3470a5c891dcc94363e492a3760af92b6b07241c\"\u003e\u003ccode\u003e3470a5c\u003c/code\u003e\u003c/a\u003e ggml-alloc : make gallocr prefer chunks that allow memory reuse (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16788\"\u003e#16788\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bd562fe4f7bd55625511d5f9d639c4fb1db1d440\"\u003e\u003ccode\u003ebd562fe\u003c/code\u003e\u003c/a\u003e cuda : use fast copy when src and dst are of different type and contiguous (#...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bbac6a26b2bd7f7c1f0831cb1e7b52734c66673b\"\u003e\u003ccode\u003ebbac6a2\u003c/code\u003e\u003c/a\u003e ggml: fix cuda kernel launch configuration for k_compute_batched_ptrs to supp...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/73a48c9790d320476b3e5ef75bda09f2f8269e6e\"\u003e\u003ccode\u003e73a48c9\u003c/code\u003e\u003c/a\u003e convert : enable expert group selection for all models with it (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16691\"\u003e#16691\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f696428ce8e4d16c17acbffeaa7feac3b0fb9061\"\u003e\u003ccode\u003ef696428\u003c/code\u003e\u003c/a\u003e graph : add clamping to ffn_moe_weights_sum to avoid div-by-zero (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16655\"\u003e#16655\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7cce4f8158f0c4c88d8dadd4c23d33938127b897\"\u003e\u003ccode\u003e7cce4f8\u003c/code\u003e\u003c/a\u003e model : set res-\u0026gt;t_embd in SmallThinker models (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16782\"\u003e#16782\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...75cbdd3fce38ea12d50cd19e73a069aa5dbbd5fa\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/29","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/29","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/29/packages"},{"uuid":"3508320929","node_id":"PR_kwDOOKZoKM6tXxdH","number":27,"state":"closed","title":"Bump llama.cpp from `2be60cb` to `f9bc66c`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2025-10-20T03:31:58.000Z","author_association":null,"state_reason":null,"created_at":"2025-10-13T03:22:29.000Z","updated_at":"2025-10-20T03:32:00.000Z","time_to_close":605369,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`f9bc66c`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `f9bc66c`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f9bc66c3ebcfddb5f09e4b21253623caeb8e414a\"\u003e\u003ccode\u003ef9bc66c\u003c/code\u003e\u003c/a\u003e CANN: Update several operators to support FP16 data format (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16251\"\u003e#16251\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a31cf36ad946a13b3a646bf0dadf2a481e89f944\"\u003e\u003ccode\u003ea31cf36\u003c/code\u003e\u003c/a\u003e metal : add opt_step_adamw and op_sum (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16529\"\u003e#16529\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/81d54bbfd599811b354c39f04550888168be7780\"\u003e\u003ccode\u003e81d54bb\u003c/code\u003e\u003c/a\u003e webui: remove client-side context pre-check and rely on backend for limits (#...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c7be9febcbafa9af7d1b9443f86475c59c9c5f87\"\u003e\u003ccode\u003ec7be9fe\u003c/code\u003e\u003c/a\u003e [SYCL] fix UT fault cases: count-equal, argsort, pad OPs (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16521\"\u003e#16521\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/8415f61e23d04427cd0d912fbb9d33b85f849456\"\u003e\u003ccode\u003e8415f61\u003c/code\u003e\u003c/a\u003e ci : add Vulkan on Ubuntu with default packages build (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16532\"\u003e#16532\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2c301e91abb92d03c1a682b4b540ba835562a74b\"\u003e\u003ccode\u003e2c301e9\u003c/code\u003e\u003c/a\u003e common : handle unicode during partial json parsing (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16526\"\u003e#16526\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/4b2dae383df708e2afc49c4859a81cd074f5ac10\"\u003e\u003ccode\u003e4b2dae3\u003c/code\u003e\u003c/a\u003e common : update presets (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16504\"\u003e#16504\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/41aac5c69b5fb281bc1f486afb053f78101bb39e\"\u003e\u003ccode\u003e41aac5c\u003c/code\u003e\u003c/a\u003e ggml : Fix FP16 ELU positive branch (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16519\"\u003e#16519\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a2fba89a426ff8005d303c73f0436e7e67368b70\"\u003e\u003ccode\u003ea2fba89\u003c/code\u003e\u003c/a\u003e hparams : add check for layer index in is_recurrent (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16511\"\u003e#16511\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/20cc625edc2264aae2779e71bef1593e6a4e8c43\"\u003e\u003ccode\u003e20cc625\u003c/code\u003e\u003c/a\u003e ggml: Correct SVE implementation in ggml_vec_dot_f16_unroll (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16518\"\u003e#16518\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...f9bc66c3ebcfddb5f09e4b21253623caeb8e414a\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/27","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/27","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/27/packages"},{"uuid":"2888707654","node_id":"PR_kwDOOKZoKM6sLi5G","number":26,"state":"open","title":"Bump llama.cpp from `2be60cb` to `ca71fb9`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":0,"pull_request":true,"closed_at":null,"author_association":"CONTRIBUTOR","state_reason":null,"created_at":"2025-10-06T03:29:59.000Z","updated_at":"2025-10-06T03:30:00.000Z","time_to_close":null,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`ca71fb9`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `ca71fb9`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ca71fb9b368e3db96e028f80c4c9df6b6b370edd\"\u003e\u003ccode\u003eca71fb9\u003c/code\u003e\u003c/a\u003e model : Granite docling + Idefics3 preprocessing (SmolVLM) (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16206\"\u003e#16206\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/35266573b968e1c947b367782fb4b3eddbb4f3c0\"\u003e\u003ccode\u003e3526657\u003c/code\u003e\u003c/a\u003e ggml webgpu: actually add softmax, fix rms_norm offset (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16400\"\u003e#16400\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/86df2c9ae4f2f1ee63d2558a9dc797b98524639b\"\u003e\u003ccode\u003e86df2c9\u003c/code\u003e\u003c/a\u003e vulkan: use a more appropriate amount of threads when generating shaders (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16\"\u003e#16\u003c/a\u003e...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f39283960b58a92ecc0c72567711318b20e22b55\"\u003e\u003ccode\u003ef392839\u003c/code\u003e\u003c/a\u003e rpc : check src buffer when copying tensor (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16421\"\u003e#16421\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/898acba6816ad23b6a9491347d30e7570bffadfd\"\u003e\u003ccode\u003e898acba\u003c/code\u003e\u003c/a\u003e rpc : add support for multiple devices (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16276\"\u003e#16276\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e29acf74fea996014380d59d31aa504ae8964258\"\u003e\u003ccode\u003ee29acf7\u003c/code\u003e\u003c/a\u003e vulkan : incremental shader builds (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16341\"\u003e#16341\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/128d522c04286e019666bd6ee4d18e3fbf8772e2\"\u003e\u003ccode\u003e128d522\u003c/code\u003e\u003c/a\u003e chat : support Magistral thinking (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16413\"\u003e#16413\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f6dcda390004b627ef30af378d0c01ad2519289e\"\u003e\u003ccode\u003ef6dcda3\u003c/code\u003e\u003c/a\u003e server : context checkpointing for hybrid and recurrent models (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16382\"\u003e#16382\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/606a73f53175077429484b23dcf799f69a31d0bd\"\u003e\u003ccode\u003e606a73f\u003c/code\u003e\u003c/a\u003e metal : fix loop bound in ggml_mem_ranges (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16412\"\u003e#16412\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/946f71ed9ade07e319859b5ce656144140e066fb\"\u003e\u003ccode\u003e946f71e\u003c/code\u003e\u003c/a\u003e llama : fix shapes for bert/mpt q/k norm (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16409\"\u003e#16409\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...ca71fb9b368e3db96e028f80c4c9df6b6b370edd\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/26","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/26","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/26/packages"},{"uuid":"2868919049","node_id":"PR_kwDOOKZoKM6rADsJ","number":25,"state":"open","title":"Bump llama.cpp from `2be60cb` to `b887d2f`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":0,"pull_request":true,"closed_at":null,"author_association":"CONTRIBUTOR","state_reason":null,"created_at":"2025-09-29T03:47:20.000Z","updated_at":"2025-09-29T03:47:21.000Z","time_to_close":null,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`b887d2f`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `b887d2f`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b887d2f3413ac231e3cb5925260c39902af4a70c\"\u003e\u003ccode\u003eb887d2f\u003c/code\u003e\u003c/a\u003e ggml : fix GGML_F32_VEC_FMA argument order in ggml_vec_mad1_f32 (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16307\"\u003e#16307\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bd0af02fc96c2057726f33c0f0daf7bb8f3e462a\"\u003e\u003ccode\u003ebd0af02\u003c/code\u003e\u003c/a\u003e common : fix reasoning before forced tool call via tool_choice = required (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/1\"\u003e#1\u003c/a\u003e...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d9e0e7c8194dfd7d23bf3a86608c9ece68d77c93\"\u003e\u003ccode\u003ed9e0e7c\u003c/code\u003e\u003c/a\u003e ci : fix musa docker build (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16306\"\u003e#16306\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/0124ac989f7e7bf08803788f66dbe4106bdcdd58\"\u003e\u003ccode\u003e0124ac9\u003c/code\u003e\u003c/a\u003e devops: switch to using ubuntu-22.04-s390x image (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16302\"\u003e#16302\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2811c65286ae954bec87049f75b86dc022006dcc\"\u003e\u003ccode\u003e2811c65\u003c/code\u003e\u003c/a\u003e Fixed a few typos in the README of the LLaMA.cpp HTTP Server [no ci] (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16297\"\u003e#16297\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d8359f5fde480da030bf75c7711573c7c4d993ba\"\u003e\u003ccode\u003ed8359f5\u003c/code\u003e\u003c/a\u003e vulkan: 64-bit im2col (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16135\"\u003e#16135\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6a2c6145a0b91b40eb3c3dba7b20ccc4b270490f\"\u003e\u003ccode\u003e6a2c614\u003c/code\u003e\u003c/a\u003e metal : extend mat-mat multiplication support (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16225\"\u003e#16225\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3b53634fe35771e2e318227aa81585726bae7234\"\u003e\u003ccode\u003e3b53634\u003c/code\u003e\u003c/a\u003e metal : fuse non-sequential nodes (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16102\"\u003e#16102\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1384abf8b8d5894d32fada453ccf4d196ffba7de\"\u003e\u003ccode\u003e1384abf\u003c/code\u003e\u003c/a\u003e vulkan: handle mat_mul with A matrix \u0026gt; 4GB (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16176\"\u003e#16176\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e6d65fb02d553bd79cad94e517cdca18b687788d\"\u003e\u003ccode\u003ee6d65fb\u003c/code\u003e\u003c/a\u003e vulkan: support arbitrary KV dimension in flash attention (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16160\"\u003e#16160\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...b887d2f3413ac231e3cb5925260c39902af4a70c\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/25","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/25","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/25/packages"},{"uuid":"2847775100","node_id":"PR_kwDOOKZoKM6pvZl8","number":24,"state":"closed","title":"Bump llama.cpp from `2be60cb` to `51f5a45`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2025-09-29T03:47:22.000Z","author_association":"CONTRIBUTOR","state_reason":null,"created_at":"2025-09-22T03:19:27.000Z","updated_at":"2025-09-29T03:47:22.000Z","time_to_close":606475,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`51f5a45`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `51f5a45`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/51f5a45fbe575dcd54bdd2a339ef8e8424d1c12a\"\u003e\u003ccode\u003e51f5a45\u003c/code\u003e\u003c/a\u003e opencl: fix concat crash on win arm64 with Adreno (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15944\"\u003e#15944\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c4510dc9374e17dcb8726902ab5216067a92b3d3\"\u003e\u003ccode\u003ec4510dc\u003c/code\u003e\u003c/a\u003e opencl: initial \u003ccode\u003eq8_0\u003c/code\u003e mv support (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15732\"\u003e#15732\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/da30ab5f8696cabb2d4620cdc0aa41a298c54fd6\"\u003e\u003ccode\u003eda30ab5\u003c/code\u003e\u003c/a\u003e ci : add label for the RISC-V runner (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16150\"\u003e#16150\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/28baac9c9f491c872e2c37762d3bd90446b005e9\"\u003e\u003ccode\u003e28baac9\u003c/code\u003e\u003c/a\u003e ci : migrate ggml ci to self-hosted runners (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16116\"\u003e#16116\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1eeb523c3e0c7ffbd59469f5463dcbdecba3535e\"\u003e\u003ccode\u003e1eeb523\u003c/code\u003e\u003c/a\u003e vulkan: optimize UMA buffer operations and fix driver hangs (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16059\"\u003e#16059\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/5bb4a3edec297e74b0f7bd4ed5d0fdd12e28d858\"\u003e\u003ccode\u003e5bb4a3e\u003c/code\u003e\u003c/a\u003e vulkan: fix validation error about VK_PIPELINE_CREATE_CAPTURE_STATISTICS_BIT_...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7f766929ca8e8e01dcceb1c526ee584f7e5e1408\"\u003e\u003ccode\u003e7f76692\u003c/code\u003e\u003c/a\u003e sync : ggml\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/405921dcefdd4e90ed948a4bf179007c2fa92b2d\"\u003e\u003ccode\u003e405921d\u003c/code\u003e\u003c/a\u003e ggml : introduce semantic versioning (ggml/1336)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/fa6383ca7e7ccb8ca3bdfeb37e348ddc4113aa26\"\u003e\u003ccode\u003efa6383c\u003c/code\u003e\u003c/a\u003e CUDA : conditionally add cuda architectures (ggml/1341)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/803dac2e48ef3ba26a504eb27c4e77ec2d21f7d0\"\u003e\u003ccode\u003e803dac2\u003c/code\u003e\u003c/a\u003e vulkan: use vec dot for matrix matrix multiplications (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16056\"\u003e#16056\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...51f5a45fbe575dcd54bdd2a339ef8e8424d1c12a\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/24","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/24","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/24/packages"},{"uuid":"2806712558","node_id":"PR_kwDOOKZoKM6nSwju","number":22,"state":"open","title":"Bump llama.cpp from `2be60cb` to `85ca66a`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":0,"pull_request":true,"closed_at":null,"author_association":"CONTRIBUTOR","state_reason":null,"created_at":"2025-09-08T03:38:55.000Z","updated_at":"2025-09-08T03:38:55.000Z","time_to_close":null,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`85ca66a`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `85ca66a`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/85ca66a74676e6d5df4433016488e039a4b464ae\"\u003e\u003ccode\u003e85ca66a\u003c/code\u003e\u003c/a\u003e CANN: Stream sync between devices for acl_graph (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15809\"\u003e#15809\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3976dfbe00f02a62c0deca32c46138e4f0ca81d8\"\u003e\u003ccode\u003e3976dfb\u003c/code\u003e\u003c/a\u003e vulkan: support im2col_3d (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15795\"\u003e#15795\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d36e61c580bf7fc7879c443c542312a42b718e11\"\u003e\u003ccode\u003ed36e61c\u003c/code\u003e\u003c/a\u003e ggml-cpu: clean up s390x SIMD (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15855\"\u003e#15855\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c97b5e5854b47b18a248d77edb693c63018a0865\"\u003e\u003ccode\u003ec97b5e5\u003c/code\u003e\u003c/a\u003e vulkan: Support pad_ext (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15794\"\u003e#15794\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/267e99867f09bec8bcc2e424ad9bcddd6cccf9d0\"\u003e\u003ccode\u003e267e998\u003c/code\u003e\u003c/a\u003e vulkan: Use larger loads in scalar/coopmat1 matmul (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15729\"\u003e#15729\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3b15924d71237a43bb5ad71f5b885ee66a821342\"\u003e\u003ccode\u003e3b15924\u003c/code\u003e\u003c/a\u003e ggml WebGPU: remove userdata from request adapter callback (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15527\"\u003e#15527\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/79bc429262268ad2ac8a364cfe6c2d6b9c5f008a\"\u003e\u003ccode\u003e79bc429\u003c/code\u003e\u003c/a\u003e CUDA: faster tile FA (Pascal/AMD), headsize 256 (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15769\"\u003e#15769\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c4df49a42d396bdf7344501813e7de53bc9e7bb3\"\u003e\u003ccode\u003ec4df49a\u003c/code\u003e\u003c/a\u003e kleidiai: generalize compute_forward_kv_cache to compute_forward_fp16 (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15817\"\u003e#15817\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3c3635d2f20424d557b5b0605a2a356214ffe048\"\u003e\u003ccode\u003e3c3635d\u003c/code\u003e\u003c/a\u003e server : speed up tests (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15836\"\u003e#15836\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/61bdfd5298a78593be649a1035ee2a120b13c4f0\"\u003e\u003ccode\u003e61bdfd5\u003c/code\u003e\u003c/a\u003e server : implement prompt processing progress report in stream mode (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15827\"\u003e#15827\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...85ca66a74676e6d5df4433016488e039a4b464ae\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/22","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/22","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/22/packages"},{"uuid":"2788776505","node_id":"PR_kwDOOKZoKM6mOVo5","number":21,"state":"open","title":"Bump llama.cpp from `2be60cb` to `b66df9d`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":0,"pull_request":true,"closed_at":null,"author_association":"CONTRIBUTOR","state_reason":null,"created_at":"2025-09-01T07:13:39.000Z","updated_at":"2025-09-01T07:13:40.000Z","time_to_close":null,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`b66df9d`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `b66df9d`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b66df9d9c942254d03209186ef24ed7c994a576e\"\u003e\u003ccode\u003eb66df9d\u003c/code\u003e\u003c/a\u003e CUDA: fix build error from ambiguous __half conversions in conv2d (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15690\"\u003e#15690\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b9382c3877c6067feccf182efe9449a2d1cb24c7\"\u003e\u003ccode\u003eb9382c3\u003c/code\u003e\u003c/a\u003e CANN: Optimize MUL_MAT_ID (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15658\"\u003e#15658\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3dc7397a2799bdc07bccf637ab7ae5a1e786d1a4\"\u003e\u003ccode\u003e3dc7397\u003c/code\u003e\u003c/a\u003e CANN: fix RoPE cache issue on multi-device (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15629\"\u003e#15629\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e92d53b29e393fc4c0f9f1f7c3fe651be8d36faa\"\u003e\u003ccode\u003ee92d53b\u003c/code\u003e\u003c/a\u003e sampling : optimize samplers by reusing bucket sort (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15665\"\u003e#15665\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/0d161f021aa33ec0e90cce96f5d1a88925557327\"\u003e\u003ccode\u003e0d161f0\u003c/code\u003e\u003c/a\u003e server : enable /slots by default and make it secure (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15630\"\u003e#15630\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/4efd5a83163ff383285b3a4c2106feabf5c69557\"\u003e\u003ccode\u003e4efd5a8\u003c/code\u003e\u003c/a\u003e metal : fix checks for available FA kernels (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15700\"\u003e#15700\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/274966226f87f301ac132da898280ca3142b60e5\"\u003e\u003ccode\u003e2749662\u003c/code\u003e\u003c/a\u003e llama : fix fattn reserve call n_seqs parameter (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15699\"\u003e#15699\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9777032dccd67bdc7785aeab7497014a8be8dacc\"\u003e\u003ccode\u003e9777032\u003c/code\u003e\u003c/a\u003e llama : separate compute buffer reserve from fattn check (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15696\"\u003e#15696\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7d3c9f2b217acf0ce5db81ae83d3f375f49ab2c7\"\u003e\u003ccode\u003e7d3c9f2\u003c/code\u003e\u003c/a\u003e ci : explicitly set fa off or on (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15692\"\u003e#15692\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bbbf5ecccb35286521f735239d499eec4279a840\"\u003e\u003ccode\u003ebbbf5ec\u003c/code\u003e\u003c/a\u003e vulkan: handle large sizes for get_rows (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15686\"\u003e#15686\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...b66df9d9c942254d03209186ef24ed7c994a576e\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/21","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/21","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/21/packages"}],"issue_packages":[{"old_version":"`5d14e5d`","new_version":"`1acee6b`","update_type":null,"path":null,"pr_created_at":"2026-05-22T21:37:02.000Z","version_change":"`5d14e5d` → `1acee6b`","issue":{"uuid":"4505731598","node_id":"PR_kwDOSDCeKc7egGk7","number":29,"state":"closed","title":"Bump llama.cpp from `5d14e5d` to `1acee6b`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-05-26T03:08:35.000Z","author_association":null,"state_reason":null,"created_at":"2026-05-22T21:37:02.000Z","updated_at":"2026-05-26T03:08:37.000Z","time_to_close":279093,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`5d14e5d`","new_version":"`1acee6b`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `5d14e5d` to `1acee6b`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1acee6bf8939948f9bcbf4b14034e4b475f06069\"\u003e\u003ccode\u003e1acee6b\u003c/code\u003e\u003c/a\u003e server: only parse empty msg if continuing an assistant msg (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23506\"\u003e#23506\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ef570f63087b6a5a2930210a13f87990e8113927\"\u003e\u003ccode\u003eef570f6\u003c/code\u003e\u003c/a\u003e perplexity : fix integer overflow (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23496\"\u003e#23496\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/cc9e331213b6a9cb186aabe01a4ec6a61419dd80\"\u003e\u003ccode\u003ecc9e331\u003c/code\u003e\u003c/a\u003e SYCL: improve MoE prefill throughput (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23142\"\u003e#23142\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bcfd1989e9a90af74669d94057ff2468682c3f4a\"\u003e\u003ccode\u003ebcfd198\u003c/code\u003e\u003c/a\u003e sycl : Level Zero detection in ggml_sycl_init (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23097\"\u003e#23097\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/56f16f235c4a6ffd0cd316e1d4b5dcfbf2dcb7a4\"\u003e\u003ccode\u003e56f16f2\u003c/code\u003e\u003c/a\u003e SYCL : gated_delta_net K\u0026gt;1 (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23174\"\u003e#23174\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/8cc67efcd4834a46b18a0cf32c9b1c99762daeac\"\u003e\u003ccode\u003e8cc67ef\u003c/code\u003e\u003c/a\u003e SYCL: add BF16 to DMMV kernel path (~4x tg speedup on Intel Arc) (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21580\"\u003e#21580\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/95feeab52e41ceaf71e87b2dd01895f6d8815b60\"\u003e\u003ccode\u003e95feeab\u003c/code\u003e\u003c/a\u003e docs: Update documentation with Granite 4.0/4.1 (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23404\"\u003e#23404\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/99d4026b116605ed8e1f3ab179b3c63bc4637195\"\u003e\u003ccode\u003e99d4026\u003c/code\u003e\u003c/a\u003e ggml-zendnn : add Q8_0 quantization support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23414\"\u003e#23414\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9c92e96a64fe0f03f5f3e5ab720a151941da1de5\"\u003e\u003ccode\u003e9c92e96\u003c/code\u003e\u003c/a\u003e cmake : build router app only during standalone builds (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23521\"\u003e#23521\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/afcda09d154a285cd366135f98ffc1d357f7ddbd\"\u003e\u003ccode\u003eafcda09\u003c/code\u003e\u003c/a\u003e vocab : fix HybridDNA tokenizer (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23466\"\u003e#23466\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/5d14e5d19bd6af7fc38eb92d96aa185e5948a03d...1acee6bf8939948f9bcbf4b14034e4b475f06069\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/MiloDevs/go-llama.cpp/pull/29","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/MiloDevs%2Fgo-llama.cpp/issues/29","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/29/packages"}},{"old_version":"`5d14e5d`","new_version":"`6a257d4`","update_type":null,"path":null,"pr_created_at":"2026-05-20T23:56:07.000Z","version_change":"`5d14e5d` → `6a257d4`","issue":{"uuid":"4490664723","node_id":"PR_kwDOSDCeKc7dvMRO","number":27,"state":"closed","title":"Bump llama.cpp from `5d14e5d` to `6a257d4`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-05-21T21:37:37.000Z","author_association":null,"state_reason":null,"created_at":"2026-05-20T23:56:07.000Z","updated_at":"2026-05-21T21:37:39.000Z","time_to_close":78090,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`5d14e5d`","new_version":"`6a257d4`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `5d14e5d` to `6a257d4`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6a257d44633d4a752183ed778b88d2924d0a6b9d\"\u003e\u003ccode\u003e6a257d4\u003c/code\u003e\u003c/a\u003e mtmd, model : merge HunyuanOCR into HunyuanVL and fix OCR vision precision (#...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3a479c9132072815cb70a443b4efa45bb66b3f59\"\u003e\u003ccode\u003e3a479c9\u003c/code\u003e\u003c/a\u003e ui: Add max image size option (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22849\"\u003e#22849\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ad277572619fcfb6ddd38f4c6437283a4b2b8636\"\u003e\u003ccode\u003ead27757\u003c/code\u003e\u003c/a\u003e Move to backend sampling for MTP draft path (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23287\"\u003e#23287\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3a6db741a8189a45260536581f4ebb0a7f051f3c\"\u003e\u003ccode\u003e3a6db74\u003c/code\u003e\u003c/a\u003e opencl: refactor backend initilization (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23318\"\u003e#23318\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/510b5c2a35652390c71327ecb29c2fb14bfe0e8c\"\u003e\u003ccode\u003e510b5c2\u003c/code\u003e\u003c/a\u003e common/speculative : fix nullptr crash in get_devices_str (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23386\"\u003e#23386\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a8681a0ed2e3f2a7452e642639671bcec20b865c\"\u003e\u003ccode\u003ea8681a0\u003c/code\u003e\u003c/a\u003e mtmd : DeepSeek-OCR image processing fixes, img_tool::resize padding refactor...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/acd604fb277044e07c2bff01f4c169167b45f478\"\u003e\u003ccode\u003eacd604f\u003c/code\u003e\u003c/a\u003e vulkan: optimize operations in the IM2COL shader (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22685\"\u003e#22685\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6ce96713de33e3e4c1599025c0bacf3c3e524c6a\"\u003e\u003ccode\u003e6ce9671\u003c/code\u003e\u003c/a\u003e feat: Add WAV MIME type variants and improve audio format detection (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23396\"\u003e#23396\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c9872a2575acc65834deb15a1f5155f6dbc75229\"\u003e\u003ccode\u003ec9872a2\u003c/code\u003e\u003c/a\u003e hexagon: HMX quantized matmul rework (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/23368\"\u003e#23368\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e947228222147356bc7e64154d3439e142481632\"\u003e\u003ccode\u003ee947228\u003c/code\u003e\u003c/a\u003e Programmatic Dependent Launch (PDL) for more performance on newer NVIDIA GPUs...\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/5d14e5d19bd6af7fc38eb92d96aa185e5948a03d...6a257d44633d4a752183ed778b88d2924d0a6b9d\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/MiloDevs/go-llama.cpp/pull/27","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/MiloDevs%2Fgo-llama.cpp/issues/27","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/27/packages"}},{"old_version":"`5d14e5d`","new_version":"`9789512`","update_type":null,"path":null,"pr_created_at":"2026-04-20T23:07:23.000Z","version_change":"`5d14e5d` → `9789512`","issue":{"uuid":"4299204512","node_id":"PR_kwDOSDCeKc7UIHD1","number":5,"state":"closed","title":"Bump llama.cpp from `5d14e5d` to `9789512`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-21T21:37:19.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-20T23:07:23.000Z","updated_at":"2026-04-21T21:37:21.000Z","time_to_close":80996,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`5d14e5d`","new_version":"`9789512`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `5d14e5d` to `9789512`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/97895129e5f2bde94d13dc01ca41ee79e9b629f2\"\u003e\u003ccode\u003e9789512\u003c/code\u003e\u003c/a\u003e ggml-cuda: flush legacy pool on OOM and retry (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22155\"\u003e#22155\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/86f8daacfe5e202b99ed396aa574a8b41a982048\"\u003e\u003ccode\u003e86f8daa\u003c/code\u003e\u003c/a\u003e mtmd: correct get_n_pos / get_decoder_pos (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22175\"\u003e#22175\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/cf8b0dbda9ac0eac30ee33f87bc6702ead1c4664\"\u003e\u003ccode\u003ecf8b0db\u003c/code\u003e\u003c/a\u003e server : remove /api endpoints (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22165\"\u003e#22165\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/fd6ae4ca1cd5446442f6c2e5e73a2a4c9bc44993\"\u003e\u003ccode\u003efd6ae4c\u003c/code\u003e\u003c/a\u003e Tensor-parallel: Fix delayed AllReduce on Gemma-4 MoE (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22129\"\u003e#22129\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/fb19f94c715c466230c72d2a32822f8a9e113708\"\u003e\u003ccode\u003efb19f94\u003c/code\u003e\u003c/a\u003e TP: fix 0-sized tensor slices, AllReduce fallback (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21808\"\u003e#21808\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7f251fdbce614a50141005dc70ce3787b7777a8e\"\u003e\u003ccode\u003e7f251fd\u003c/code\u003e\u003c/a\u003e ggml-cpu: Optimized x86 and generic cpu q1_0 dot (follow up) (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21636\"\u003e#21636\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a6cc43c286a2ebc429aa69b9a4d16de082cedb51\"\u003e\u003ccode\u003ea6cc43c\u003c/code\u003e\u003c/a\u003e ggml-webgpu: updated matrix-vector multiplication (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21738\"\u003e#21738\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a678916623ddef89c2a43776df24e00a52b17638\"\u003e\u003ccode\u003ea678916\u003c/code\u003e\u003c/a\u003e mtmd: refactor mtmd_decode_use_mrope (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22161\"\u003e#22161\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/81df3f7cfaa6f99de14e792b38d5771bf427383e\"\u003e\u003ccode\u003e81df3f7\u003c/code\u003e\u003c/a\u003e fix: GLM-DSA crash in llama-tokenize when using vocab_only (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22102\"\u003e#22102\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/de71b5f81c3b6b9f8bdaf1b2a21198e1eede3fda\"\u003e\u003ccode\u003ede71b5f\u003c/code\u003e\u003c/a\u003e server : refactor \u0026quot;use checkpoint\u0026quot; logic (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/22114\"\u003e#22114\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/5d14e5d19bd6af7fc38eb92d96aa185e5948a03d...97895129e5f2bde94d13dc01ca41ee79e9b629f2\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/MiloDevs/go-llama.cpp/pull/5","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/MiloDevs%2Fgo-llama.cpp/issues/5","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/5/packages"}},{"old_version":"`0893f50`","new_version":"`b572d1e`","update_type":null,"path":null,"pr_created_at":"2026-04-16T12:55:31.000Z","version_change":"`0893f50` → `b572d1e`","issue":{"uuid":"4275850052","node_id":"PR_kwDOQit6R87S-HXe","number":102,"state":"closed","title":"build(deps): bump llama.cpp from `0893f50` to `b572d1e`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-17T12:55:07.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-16T12:55:31.000Z","updated_at":"2026-04-17T12:55:08.000Z","time_to_close":86376,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`0893f50`","new_version":"`b572d1e`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `0893f50` to `b572d1e`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b572d1ecd62210229e04cdeffd3ae80dd59f0921\"\u003e\u003ccode\u003eb572d1e\u003c/code\u003e\u003c/a\u003e codeowners: add team member comments (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21714\"\u003e#21714\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/03b3d077988dffdf08bf628fab78904526745115\"\u003e\u003ccode\u003e03b3d07\u003c/code\u003e\u003c/a\u003e Convert: Fix NemotronH Config Parsing (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21664\"\u003e#21664\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3f7c29d318e317b63f54c558bc69803963d7d88c\"\u003e\u003ccode\u003e3f7c29d\u003c/code\u003e\u003c/a\u003e ggml: add graph_reused (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21764\"\u003e#21764\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ae2d34899e2a9a172c7f2090ed4dd366bbf25d0d\"\u003e\u003ccode\u003eae2d348\u003c/code\u003e\u003c/a\u003e metal: Implement ROLL op (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21946\"\u003e#21946\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1e796eb41fb51950ada45811a303e57a5f4ea974\"\u003e\u003ccode\u003e1e796eb\u003c/code\u003e\u003c/a\u003e ggml-cpu: add 128-bit RVV implementation for Quantization Vector Dot (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/20633\"\u003e#20633\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/5637536517ae4ed3eaa22b39c0d479e049097a9b\"\u003e\u003ccode\u003e5637536\u003c/code\u003e\u003c/a\u003e ggml : implemented simd_gemm kernel for riscv vector extension (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/20627\"\u003e#20627\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/90fb96a7b3c3dd97420d49603aefe773a612c05a\"\u003e\u003ccode\u003e90fb96a\u003c/code\u003e\u003c/a\u003e devops : added spirv-headers to nix (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21965\"\u003e#21965\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/82677a6ede7927d2286ef1c9e481ce4caf52866f\"\u003e\u003ccode\u003e82677a6\u003c/code\u003e\u003c/a\u003e ggml-webgpu: compute pass batching and removing profiling overhead (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21873\"\u003e#21873\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/8612ed18b7d2896009f255c11eb002aa7bfa9057\"\u003e\u003ccode\u003e8612ed1\u003c/code\u003e\u003c/a\u003e ci : Use ggml-org/ccache-action on RISC-V as well (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21632\"\u003e#21632\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b1be68e8cab67f5c2fb7d8e3e90291a8805ece0e\"\u003e\u003ccode\u003eb1be68e\u003c/code\u003e\u003c/a\u003e [SYCL] Fix Q8_0 reorder: garbage on 2nd prompt + crash on full VRAM (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21638\"\u003e#21638\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/0893f50f2dc14fcc046e10d4f76a1ac7a62c0490...b572d1ecd62210229e04cdeffd3ae80dd59f0921\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/102","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/102","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/102/packages"}},{"old_version":"`67a2209`","new_version":"`1e9d771`","update_type":null,"path":null,"pr_created_at":"2026-04-12T18:12:46.000Z","version_change":"`67a2209` → `1e9d771`","issue":{"uuid":"4249310905","node_id":"PR_kwDOQ9H31s7RzZDd","number":24,"state":"closed","title":"build(deps): bump llama.cpp from `67a2209` to `1e9d771`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-19T18:12:42.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-12T18:12:46.000Z","updated_at":"2026-04-19T18:12:43.000Z","time_to_close":604796,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`67a2209`","new_version":"`1e9d771`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `67a2209` to `1e9d771`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1e9d771e2c2f1113a5ebdd0dc15bafe57dce64be\"\u003e\u003ccode\u003e1e9d771\u003c/code\u003e\u003c/a\u003e convert : force f16 or f32 on step3-vl conv weights (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21646\"\u003e#21646\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/aa4695c5e5bf0abda8942c08e94cb804a7ea0347\"\u003e\u003ccode\u003eaa4695c\u003c/code\u003e\u003c/a\u003e mtmd: add gemma 4 test (vision + audio) [no ci] (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21806\"\u003e#21806\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/547765a93e5ad7b4e8ca84d78f6d83f36ad8ee25\"\u003e\u003ccode\u003e547765a\u003c/code\u003e\u003c/a\u003e mtmd: add Gemma 4 audio conformer encoder support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21421\"\u003e#21421\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9e209c5aee8effac146463c8dc32984a4b4d2672\"\u003e\u003ccode\u003e9e209c5\u003c/code\u003e\u003c/a\u003e fix: Proper messages rendering for \u0026quot;Show raw output\u0026quot; (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21672\"\u003e#21672\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6313acbef016d9e4d8e83d3647082329949d958b\"\u003e\u003ccode\u003e6313acb\u003c/code\u003e\u003c/a\u003e docs: add guide on how to add multimodal support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21778\"\u003e#21778\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ff5ef8278615a2462b79b50abdf3cc95cfb31c6f\"\u003e\u003ccode\u003eff5ef82\u003c/code\u003e\u003c/a\u003e CUDA: skip compilation of superfluous FA kernels (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21768\"\u003e#21768\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/073bb2c20b5b2c919469653214aaa1a9895816a2\"\u003e\u003ccode\u003e073bb2c\u003c/code\u003e\u003c/a\u003e mtmd : add MERaLiON-2 multimodal audio support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21756\"\u003e#21756\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/af1127d3c49e41a606bac7c2b3897489aa71b918\"\u003e\u003ccode\u003eaf1127d\u003c/code\u003e\u003c/a\u003e opencl: add basic support for q5_k (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21593\"\u003e#21593\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/865ff06b2ffa2f91f30cbec1f8c73d66cc6642aa\"\u003e\u003ccode\u003e865ff06\u003c/code\u003e\u003c/a\u003e TP: fix Qwen 3 Next data split (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21732\"\u003e#21732\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2b2cd57de64e96c9ebfca6ba7e6bdbab6fe51482\"\u003e\u003ccode\u003e2b2cd57\u003c/code\u003e\u003c/a\u003e ggml : fix a few instances of missing GGML_TYPE_Q1_0 cases (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21716\"\u003e#21716\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/67a2209fabe2e3498d458561933d5380655085d2...1e9d771e2c2f1113a5ebdd0dc15bafe57dce64be\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/Jota-project/jota-inference/pull/24","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/Jota-project%2Fjota-inference/issues/24","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/24/packages"}},{"old_version":"`67a2209`","new_version":"`5d3a4a7`","update_type":null,"path":null,"pr_created_at":"2026-04-05T18:12:42.000Z","version_change":"`67a2209` → `5d3a4a7`","issue":{"uuid":"4208255933","node_id":"PR_kwDOQ9H31s7QEP9P","number":22,"state":"closed","title":"build(deps): bump llama.cpp from `67a2209` to `5d3a4a7`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-12T18:12:48.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-05T18:12:42.000Z","updated_at":"2026-04-12T18:12:49.000Z","time_to_close":604806,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`67a2209`","new_version":"`5d3a4a7`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `67a2209` to `5d3a4a7`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/5d3a4a7da5e3dd42f5922aba2fe21b520e96e830\"\u003e\u003ccode\u003e5d3a4a7\u003c/code\u003e\u003c/a\u003e server : fix logging of build + system info (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21460\"\u003e#21460\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c08d28d08871715fd68accffaeeb76ddcaede658\"\u003e\u003ccode\u003ec08d28d\u003c/code\u003e\u003c/a\u003e ci: lower cuda12 floor to 12.8.1 for broader host compatibility (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21438\"\u003e#21438\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/661e9acb36768d0d4ddb6f2eb674fbb1be185823\"\u003e\u003ccode\u003e661e9ac\u003c/code\u003e\u003c/a\u003e ci: fix vulkan workflow referencing non-existent action (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21442\"\u003e#21442\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b8635075ffe27b135c49afb9a8b5c434bd42c502\"\u003e\u003ccode\u003eb863507\u003c/code\u003e\u003c/a\u003e common : add gemma 4 specialized parser (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21418\"\u003e#21418\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9c699074c97191754c8a966298f84c79f90fce38\"\u003e\u003ccode\u003e9c69907\u003c/code\u003e\u003c/a\u003e server: Fix undefined timing measurement errors in server context (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21201\"\u003e#21201\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d01f6274c01e111be2ccc39443f79884796e48fb\"\u003e\u003ccode\u003ed01f627\u003c/code\u003e\u003c/a\u003e common : respect specified tag, only fallback when tag is empty (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21413\"\u003e#21413\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/650bf14eb9a922de0f88c9523a271159cc5ae469\"\u003e\u003ccode\u003e650bf14\u003c/code\u003e\u003c/a\u003e llama-model: read final_logit_softcapping for Gemma 4 (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21390\"\u003e#21390\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b7ad48ebda2287c778fd826606d7b3b3570f60ab\"\u003e\u003ccode\u003eb7ad48e\u003c/code\u003e\u003c/a\u003e llama: add custom newline split for Gemma 4 (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21406\"\u003e#21406\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d006858316d4650bb4da0c6923294ccd741caefd\"\u003e\u003ccode\u003ed006858\u003c/code\u003e\u003c/a\u003e ggml-webgpu: move from parameter buffer pool to single buffer with offsets (#...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e43970099269b5b6da36b8977ad47697602e4e54\"\u003e\u003ccode\u003ee439700\u003c/code\u003e\u003c/a\u003e ci: Add Windows Vulkan backend testing on Intel (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21292\"\u003e#21292\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/67a2209fabe2e3498d458561933d5380655085d2...5d3a4a7da5e3dd42f5922aba2fe21b520e96e830\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/Jota-project/jota-inference/pull/22","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/Jota-project%2Fjota-inference/issues/22","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/22/packages"}},{"old_version":"`8710e5f`","new_version":"`887535c`","update_type":null,"path":null,"pr_created_at":"2026-04-03T12:54:31.000Z","version_change":"`8710e5f` → `887535c`","issue":{"uuid":"4200436429","node_id":"PR_kwDOQit6R87Py9zK","number":93,"state":"closed","title":"build(deps): bump llama.cpp from `8710e5f` to `887535c`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-06T12:57:05.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-03T12:54:31.000Z","updated_at":"2026-04-06T12:57:07.000Z","time_to_close":259354,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`8710e5f`","new_version":"`887535c`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `8710e5f` to `887535c`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/887535c33f9e3bf57532f31bc3d749b264751a2b\"\u003e\u003ccode\u003e887535c\u003c/code\u003e\u003c/a\u003e ci: add more binary checks (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21349\"\u003e#21349\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d3416a4aa9a37d9a0ca547e18c0e126bfe8a07ea\"\u003e\u003ccode\u003ed3416a4\u003c/code\u003e\u003c/a\u003e fix: remove stale assert (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21369\"\u003e#21369\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/43a4ee4a2cf25de0428d618544b877731d4d3713\"\u003e\u003ccode\u003e43a4ee4\u003c/code\u003e\u003c/a\u003e HIP: build eatch ci build test for a different architecture (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21337\"\u003e#21337\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f851fa5ab056c9bada48ad7208fe122fc0574e44\"\u003e\u003ccode\u003ef851fa5\u003c/code\u003e\u003c/a\u003e fix: add openssl to nix dependencies (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21353\"\u003e#21353\u003c/a\u003e) (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21355\"\u003e#21355\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f1ac84119ccc8e72dafd9e9f8fc3b9399917ce11\"\u003e\u003ccode\u003ef1ac841\u003c/code\u003e\u003c/a\u003e ggml-zendnn : add MUL_MAT_ID op support for MoE models (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21315\"\u003e#21315\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b069b10ab48f25ba119e59d0b8bf35d4f06e093f\"\u003e\u003ccode\u003eb069b10\u003c/code\u003e\u003c/a\u003e vocab: fix Gemma4 tokenizer (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21343\"\u003e#21343\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/0c58ba3365d2bc717b447b5d70e4d6be09ff3c40\"\u003e\u003ccode\u003e0c58ba3\u003c/code\u003e\u003c/a\u003e rpc : reuse compute graph buffers (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21299\"\u003e#21299\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/57ace0d612a11133ac86edcc7af1b323bf05f12f\"\u003e\u003ccode\u003e57ace0d\u003c/code\u003e\u003c/a\u003e chat : avoid including json in chat.h (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21306\"\u003e#21306\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/39b27f0da0271c06986cb31b68bc0fe68e780616\"\u003e\u003ccode\u003e39b27f0\u003c/code\u003e\u003c/a\u003e (revert) kv-cache : do not quantize SWA KV cache (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21332\"\u003e#21332\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f49e9178767d557a522618b16ce8694f9ddac628\"\u003e\u003ccode\u003ef49e917\u003c/code\u003e\u003c/a\u003e ci : add AMD ZenDNN label to PR labeler (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21345\"\u003e#21345\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/8710e5f9b9bd7246608808ccd3626bde8abf6ff9...887535c33f9e3bf57532f31bc3d749b264751a2b\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/93","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/93","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/93/packages"}},{"old_version":"`8710e5f`","new_version":"`e15efe0`","update_type":null,"path":null,"pr_created_at":"2026-04-02T12:55:17.000Z","version_change":"`8710e5f` → `e15efe0`","issue":{"uuid":"4194414495","node_id":"PR_kwDOQit6R87PjY4K","number":92,"state":"closed","title":"build(deps): bump llama.cpp from `8710e5f` to `e15efe0`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-04-03T12:54:33.000Z","author_association":null,"state_reason":null,"created_at":"2026-04-02T12:55:17.000Z","updated_at":"2026-04-03T12:54:35.000Z","time_to_close":86356,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`8710e5f`","new_version":"`e15efe0`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `8710e5f` to `e15efe0`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e15efe007dc1c0d79afa347190dba91de3bd659b\"\u003e\u003ccode\u003ee15efe0\u003c/code\u003e\u003c/a\u003e Relax prefill parser to allow space. (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21240\"\u003e#21240\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6137c325a16073c8bf68a52396a815006ccaa9a9\"\u003e\u003ccode\u003e6137c32\u003c/code\u003e\u003c/a\u003e chat : add Granite 4.0 chat template with correct tool_call role mapping (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/20\"\u003e#20\u003c/a\u003e...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/17193cce34036a6488b092ca79313d4ee1f895f5\"\u003e\u003ccode\u003e17193cc\u003c/code\u003e\u003c/a\u003e kv-cache : do not quantize SWA KV cache (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21277\"\u003e#21277\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d6dac92bfdf6797b74eff25493c5e525561f70fb\"\u003e\u003ccode\u003ed6dac92\u003c/code\u003e\u003c/a\u003e Ignore Transfer-Encoding header. (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/20269\"\u003e#20269\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/dae2bf41c91a0d8eea7b6c7ded08d452eb8aeb79\"\u003e\u003ccode\u003edae2bf4\u003c/code\u003e\u003c/a\u003e sync : ggml\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bc07d559224260f65f40595121d9e2ebe60ee99e\"\u003e\u003ccode\u003ebc07d55\u003c/code\u003e\u003c/a\u003e ggml : bump version to 0.9.11 (ggml/1456)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/4888137b1736b706e39806025d24e4ca342f1e4a\"\u003e\u003ccode\u003e4888137\u003c/code\u003e\u003c/a\u003e sycl : fix llama_kv_cache hang when kv_cache is huge: 5GB (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21283\"\u003e#21283\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/fbd441c37933550c1e3365dc84dd73232334c15d\"\u003e\u003ccode\u003efbd441c\u003c/code\u003e\u003c/a\u003e hexagon : add cumsum op support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21246\"\u003e#21246\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c30e012253dd9e322c8e3424f808a5c74ecc46bf\"\u003e\u003ccode\u003ec30e012\u003c/code\u003e\u003c/a\u003e contrib : rewrite AGENTS.md, make it more clear about project values (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21270\"\u003e#21270\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/95a6ebabb277c4cc18247e7bc2a5502133caca63\"\u003e\u003ccode\u003e95a6eba\u003c/code\u003e\u003c/a\u003e opencl: fix leak in Adreno q8_0 path (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/21212\"\u003e#21212\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/8710e5f9b9bd7246608808ccd3626bde8abf6ff9...e15efe007dc1c0d79afa347190dba91de3bd659b\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/92","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/92","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/92/packages"}},{"old_version":"`e6267a9`","new_version":"`b908baf`","update_type":null,"path":null,"pr_created_at":"2026-02-20T12:55:07.000Z","version_change":"`e6267a9` → `b908baf`","issue":{"uuid":"3968435681","node_id":"PR_kwDOQit6R87FKPts","number":63,"state":"closed","title":"build(deps): bump llama.cpp from `e6267a9` to `b908baf`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-02-23T13:46:52.000Z","author_association":null,"state_reason":null,"created_at":"2026-02-20T12:55:07.000Z","updated_at":"2026-02-23T13:46:54.000Z","time_to_close":262305,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`e6267a9`","new_version":"`b908baf`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `e6267a9` to `b908baf`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b908baf1825b1a89afef87b09e22c32af2ca6548\"\u003e\u003ccode\u003eb908baf\u003c/code\u003e\u003c/a\u003e ggml-cpu: add RVV vec dot kernels for quantization types (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18784\"\u003e#18784\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/492bc319782b1f13f302911f4c73437382cc8bb9\"\u003e\u003ccode\u003e492bc31\u003c/code\u003e\u003c/a\u003e quantize : add --dry-run option (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19526\"\u003e#19526\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/77d6ae4ac89bb879ada3989a748276dfe4553674\"\u003e\u003ccode\u003e77d6ae4\u003c/code\u003e\u003c/a\u003e test: mul_mat tests with huge batch size (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19519\"\u003e#19519\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/10b26ee23a2d1b563a62db1ea4710cf8b723791a\"\u003e\u003ccode\u003e10b26ee\u003c/code\u003e\u003c/a\u003e WebUI hide models in router mode (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19374\"\u003e#19374\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3dadc88b589ca43b8fca0e1beb22d4b78a09b4dd\"\u003e\u003ccode\u003e3dadc88\u003c/code\u003e\u003c/a\u003e common : fix Step-3.5-Flash format detection and thinking support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19635\"\u003e#19635\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/39e4b1dc9bd00eb21a4e9cc6950855f94bc66de0\"\u003e\u003ccode\u003e39e4b1d\u003c/code\u003e\u003c/a\u003e common : fix gpt-oss Jinja error when assistant message has both content and ...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/11c325c6e0666a30590cde390d5746a405e536b9\"\u003e\u003ccode\u003e11c325c\u003c/code\u003e\u003c/a\u003e ggml-webgpu: Add unary op (SQR, SQRT, SIN, COS) support. (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19700\"\u003e#19700\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/237958db339300bdd8028608cc08b2ba2685ec33\"\u003e\u003ccode\u003e237958d\u003c/code\u003e\u003c/a\u003e model: Add PaddleOCR-VL model support (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18825\"\u003e#18825\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/abb9f3c42b5e6acee9e8e37836ef691d1a41bdb8\"\u003e\u003ccode\u003eabb9f3c\u003c/code\u003e\u003c/a\u003e vulkan: fix MMQ shader push constants and multi-dispatch (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19732\"\u003e#19732\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/da348c9dfbcfab16584f4640ee53146fdf85a741\"\u003e\u003ccode\u003eda348c9\u003c/code\u003e\u003c/a\u003e models : fix qwen3.5 beta/gate shapes (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19730\"\u003e#19730\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eSee full diff in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/e6267a935901313dc727ec74d159fc66e206e9c4...b908baf1825b1a89afef87b09e22c32af2ca6548\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/63","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/63","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/63/packages"}},{"old_version":"`b70d251`","new_version":"`cb6caca`","update_type":null,"path":null,"pr_created_at":"2026-01-23T12:56:05.000Z","version_change":"`b70d251` → `cb6caca`","issue":{"uuid":"3847341072","node_id":"PR_kwDOQit6R86-7PoZ","number":43,"state":"closed","title":"build(deps): bump llama.cpp from `b70d251` to `cb6caca`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2026-01-26T14:04:15.000Z","author_association":null,"state_reason":null,"created_at":"2026-01-23T12:56:05.000Z","updated_at":"2026-01-26T14:04:16.000Z","time_to_close":263290,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"build(deps)","packages":[{"name":"llama.cpp","old_version":"`b70d251`","new_version":"`cb6caca`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `b70d251` to `cb6caca`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/cb6caca191b9a3a9a4eaa13dd9e465225d127034\"\u003e\u003ccode\u003ecb6caca\u003c/code\u003e\u003c/a\u003e [SYCL] use malloc to support both iGPU and dGPU in same time (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18992\"\u003e#18992\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b5b8fa1c8b3b27683b2965a22f9985eec683d384\"\u003e\u003ccode\u003eb5b8fa1\u003c/code\u003e\u003c/a\u003e chat : fix translategemma crash on common_chat_format_example (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19019\"\u003e#19019\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a14b960bc70a0b48405409bbe3e0d6238473a0f8\"\u003e\u003ccode\u003ea14b960\u003c/code\u003e\u003c/a\u003e model-conversion : use BUILD_DIR variable in all scripts (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19015\"\u003e#19015\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/091a46cb8d43c0e662d04b80a3d11320d25b7d49\"\u003e\u003ccode\u003e091a46c\u003c/code\u003e\u003c/a\u003e ggml-cpu: aarm64: q5_K repack gemm and gemv (and generic) implementations (i8...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a3e812811d8f12f4236efa41287dc3dcd5c3c2f6\"\u003e\u003ccode\u003ea3e8128\u003c/code\u003e\u003c/a\u003e cli : load parser definition (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19031\"\u003e#19031\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/51fa458a92d6a3f305f8fd76fc8f702e3e87ddb5\"\u003e\u003ccode\u003e51fa458\u003c/code\u003e\u003c/a\u003e server : support preserving reasoning_content in assistant message (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18994\"\u003e#18994\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a5eaa1d6a3732bc0f460b02b61c95680bba5a012\"\u003e\u003ccode\u003ea5eaa1d\u003c/code\u003e\u003c/a\u003e mla : make the V tensor a view of K (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18986\"\u003e#18986\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e2baf02162382a14c9f4fc15d7681a715256453c\"\u003e\u003ccode\u003ee2baf02\u003c/code\u003e\u003c/a\u003e CUDA: fix alignment check for FA (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/19023\"\u003e#19023\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e34d6d03b25d9e8d07f3bd0190b27d0d01a7e416\"\u003e\u003ccode\u003ee34d6d0\u003c/code\u003e\u003c/a\u003e convert_hf_to_gguf.py: refactor modify_tensors to call super (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/18866\"\u003e#18866\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9c96465f99e47a3a568c50969ff5c6b672ab2714\"\u003e\u003ccode\u003e9c96465\u003c/code\u003e\u003c/a\u003e opencl: enable the general fp mm for non-cont input and as a fallback for spe...\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/b70d251076ac7c3ac1cd5d39dbb167f6ff3b6880...cb6caca191b9a3a9a4eaa13dd9e465225d127034\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/43","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/43","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/43/packages"}},{"old_version":"`ac43576`","new_version":"`bde188d`","update_type":null,"path":null,"pr_created_at":"2025-12-04T19:26:06.000Z","version_change":"`ac43576` → `bde188d`","issue":{"uuid":"3696163650","node_id":"PR_kwDOQit6R863JhQJ","number":1,"state":"open","title":"Bump llama.cpp from `ac43576` to `bde188d`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":2,"pull_request":true,"closed_at":null,"author_association":null,"state_reason":null,"created_at":"2025-12-04T19:26:06.000Z","updated_at":"2025-12-04T19:29:19.000Z","time_to_close":null,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`ac43576`","new_version":"`bde188d`","repository_url":"https://github.com/ggerganov/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggerganov/llama.cpp) from `ac43576` to `bde188d`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bde188d60f58012ada0725c6dd5ba7c69fe4dd87\"\u003e\u003ccode\u003ebde188d\u003c/code\u003e\u003c/a\u003e metal: TRI, FILL, EXPM1, SOFTPLUS (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/16623\"\u003e#16623\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9d0229967a0538840368547ee7ddc637fc28142d\"\u003e\u003ccode\u003e9d02299\u003c/code\u003e\u003c/a\u003e server: strip content-length header on proxy (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17734\"\u003e#17734\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c4c10bfb86569ccb070d0dbe1a621a8f186baa16\"\u003e\u003ccode\u003ec4c10bf\u003c/code\u003e\u003c/a\u003e server: move msg diffs tracking to HTTP thread  (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17740\"\u003e#17740\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/817d743cc17cf644dab8408eb0f1e6eac89562c1\"\u003e\u003ccode\u003e817d743\u003c/code\u003e\u003c/a\u003e examples : add missing code block end marker [no ci] (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17756\"\u003e#17756\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bd4ef134763d81e251fd097019578f2df571dfef\"\u003e\u003ccode\u003ebd4ef13\u003c/code\u003e\u003c/a\u003e common : skip model validation when --help is requested (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17755\"\u003e#17755\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/87a2084c45188d54a554c305a397e778759545ed\"\u003e\u003ccode\u003e87a2084\u003c/code\u003e\u003c/a\u003e ggml-cpu : remove asserts always evaluating to false (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17728\"\u003e#17728\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3659aa28e963ef3f782cd27258e97ddef678c776\"\u003e\u003ccode\u003e3659aa2\u003c/code\u003e\u003c/a\u003e convert: use existing local chat_template if mistral-format model has one. (#...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2a73f81f8a810783db5794256e5ba79f298adee7\"\u003e\u003ccode\u003e2a73f81\u003c/code\u003e\u003c/a\u003e cmake : simplify build info detection using standard variables (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17423\"\u003e#17423\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7dba049b0707ae395c59b085c5fd52cae7b74fe0\"\u003e\u003ccode\u003e7dba049\u003c/code\u003e\u003c/a\u003e ci : disable ggml-ci-x64-amd-* (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17753\"\u003e#17753\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/83c1171529a30c5e018779339690e21430aae372\"\u003e\u003ccode\u003e83c1171\u003c/code\u003e\u003c/a\u003e common: use native MultiByteToWideChar (\u003ca href=\"https://redirect.github.com/ggerganov/llama.cpp/issues/17738\"\u003e#17738\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggerganov/llama.cpp/compare/ac43576124a75c2de6e333ac31a3444ff9eb9458...bde188d60f58012ada0725c6dd5ba7c69fe4dd87\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/AshkanYarmoradi/go-llama.cpp/pull/1","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/AshkanYarmoradi%2Fgo-llama.cpp/issues/1","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/1/packages"}},{"old_version":"`2be60cb`","new_version":"`b8595b1`","update_type":null,"path":null,"pr_created_at":"2025-11-10T03:31:52.000Z","version_change":"`2be60cb` → `b8595b1`","issue":{"uuid":"3605942282","node_id":"PR_kwDOOKZoKM6ybK-f","number":31,"state":"closed","title":"Bump llama.cpp from `2be60cb` to `b8595b1`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2025-11-17T03:19:58.000Z","author_association":null,"state_reason":null,"created_at":"2025-11-10T03:31:52.000Z","updated_at":"2025-11-17T03:20:00.000Z","time_to_close":604086,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`b8595b1`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `b8595b1`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b8595b16e69e3029e06be3b8f6635f9812b2bc3f\"\u003e\u003ccode\u003eb8595b1\u003c/code\u003e\u003c/a\u003e mtmd : fix embedding size for image input (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17123\"\u003e#17123\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/392e09a60852d0e879d4bbedd5ace3e6852f719e\"\u003e\u003ccode\u003e392e09a\u003c/code\u003e\u003c/a\u003e vulkan: fix memory allocations (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17122\"\u003e#17122\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/802cef44bfaa80987076d621c8bf5875627c197b\"\u003e\u003ccode\u003e802cef4\u003c/code\u003e\u003c/a\u003e convert : parse safetensors directly (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15667\"\u003e#15667\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1c07c0c68c692d39b83f491bad9447af852bb652\"\u003e\u003ccode\u003e1c07c0c\u003c/code\u003e\u003c/a\u003e convert : handle compressed-tensors quant method (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17069\"\u003e#17069\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/cb1adf885105da7ce23db746b4202f4e987aa3e8\"\u003e\u003ccode\u003ecb1adf8\u003c/code\u003e\u003c/a\u003e server : handle failures to restore host cache (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17078\"\u003e#17078\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ef1d8269972bd086bee2554fd31a865c6da84f33\"\u003e\u003ccode\u003eef1d826\u003c/code\u003e\u003c/a\u003e benches : add folder with benchmarks (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16931\"\u003e#16931\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/86fde91e62c3f72ab7ed8a540dc1be049b735477\"\u003e\u003ccode\u003e86fde91\u003c/code\u003e\u003c/a\u003e Switch to using Ubuntu 25.10 vulkan/mesa (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16497\"\u003e#16497\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7f3e9d339c99d96d6df9833c63ec27dbbc96f003\"\u003e\u003ccode\u003e7f3e9d3\u003c/code\u003e\u003c/a\u003e vulkan: iGPU memory reporting fix (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17110\"\u003e#17110\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/8a3519b70898b07ec05c391418a05aaa6b377c83\"\u003e\u003ccode\u003e8a3519b\u003c/code\u003e\u003c/a\u003e vulkan: fix mmq out of bounds reads (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17108\"\u003e#17108\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/80a6cf63473b95742444a1b27d45164591282a7d\"\u003e\u003ccode\u003e80a6cf6\u003c/code\u003e\u003c/a\u003e vulkan: fuse mul_mat_id + mul (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/17095\"\u003e#17095\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...b8595b16e69e3029e06be3b8f6635f9812b2bc3f\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/31","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/31","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/31/packages"}},{"old_version":"`2be60cb`","new_version":"`7e99416`","update_type":null,"path":null,"pr_created_at":"2025-11-03T03:27:28.000Z","version_change":"`2be60cb` → `7e99416`","issue":{"uuid":"3580402939","node_id":"PR_kwDOOKZoKM6xGfhG","number":30,"state":"closed","title":"Bump llama.cpp from `2be60cb` to `7e99416`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2025-11-10T03:31:55.000Z","author_association":null,"state_reason":null,"created_at":"2025-11-03T03:27:28.000Z","updated_at":"2025-11-10T03:31:56.000Z","time_to_close":605067,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`7e99416`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `7e99416`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7e994168b1ccc12337ba8de939c4fd466107c1fb\"\u003e\u003ccode\u003e7e99416\u003c/code\u003e\u003c/a\u003e SYCL: optimized repeat_back kernel (3× fewer asm instructions, 2× faster)Feat...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bcfa87622ae46be6345a8e3dfdbdc5ba5414042b\"\u003e\u003ccode\u003ebcfa876\u003c/code\u003e\u003c/a\u003e feat(webui): improve LaTeX rendering with currency detection (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16508\"\u003e#16508\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a2054e3a8ff0da3978a4acc18c349ff58554d336\"\u003e\u003ccode\u003ea2054e3\u003c/code\u003e\u003c/a\u003e test-backend-ops : fix segfault in moe-expert-reduce test in support mode and...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/dd5286805004db1f9ac3176a1cbbfe373bdda0f8\"\u003e\u003ccode\u003edd52868\u003c/code\u003e\u003c/a\u003e ci : disable failing riscv cross build (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16952\"\u003e#16952\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6b9a52422bac0f50dd8f1f8386744fa3ce9783bf\"\u003e\u003ccode\u003e6b9a524\u003c/code\u003e\u003c/a\u003e model: add Janus Pro for image understanding (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16906\"\u003e#16906\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2f966b8ed87514e74bb96592217226cb6a6974dd\"\u003e\u003ccode\u003e2f966b8\u003c/code\u003e\u003c/a\u003e clip : use FA (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16837\"\u003e#16837\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/cd5e3b57541ecc52421130742f4d89acbcf77cd4\"\u003e\u003ccode\u003ecd5e3b5\u003c/code\u003e\u003c/a\u003e server : support unified cache across slots (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16736\"\u003e#16736\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/87c9efc3b297b8a498716b1db3d061842e6fc85b\"\u003e\u003ccode\u003e87c9efc\u003c/code\u003e\u003c/a\u003e common : move gpt-oss reasoning processing to init params (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16937\"\u003e#16937\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/76af40aaaad78c42faecd8016a88362c788b84b0\"\u003e\u003ccode\u003e76af40a\u003c/code\u003e\u003c/a\u003e docs: remove llama_sampler_accept reference in sampling sample usage (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16920\"\u003e#16920\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7db35a7958a943be1693879f42d166f152613979\"\u003e\u003ccode\u003e7db35a7\u003c/code\u003e\u003c/a\u003e CUDA: add FLOOR, CEIL, ROUND, TRUNC unary ops (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16917\"\u003e#16917\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...7e994168b1ccc12337ba8de939c4fd466107c1fb\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/30","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/30","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/30/packages"}},{"old_version":"`2be60cb`","new_version":"`75cbdd3`","update_type":null,"path":null,"pr_created_at":"2025-10-27T03:32:29.000Z","version_change":"`2be60cb` → `75cbdd3`","issue":{"uuid":"3554906682","node_id":"PR_kwDOOKZoKM6vyucH","number":29,"state":"closed","title":"Bump llama.cpp from `2be60cb` to `75cbdd3`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2025-11-03T03:27:31.000Z","author_association":null,"state_reason":null,"created_at":"2025-10-27T03:32:29.000Z","updated_at":"2025-11-03T03:27:32.000Z","time_to_close":604502,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`75cbdd3`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `75cbdd3`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/75cbdd3fce38ea12d50cd19e73a069aa5dbbd5fa\"\u003e\u003ccode\u003e75cbdd3\u003c/code\u003e\u003c/a\u003e test-backend-ops: print failed tests at the end (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16785\"\u003e#16785\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2b9bd9bf4e759c05db629ec1c391dc8aeaa71887\"\u003e\u003ccode\u003e2b9bd9b\u003c/code\u003e\u003c/a\u003e sycl: add ROLL operation support (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16665\"\u003e#16665\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/59fc1ec8e83b14354c1a3a8acf8c5c2cbf9af42f\"\u003e\u003ccode\u003e59fc1ec\u003c/code\u003e\u003c/a\u003e sycl: add REPEAT_BACK operation support (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16734\"\u003e#16734\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/75d33b9302f84a5b89f82205d2bcd8def5a64e0a\"\u003e\u003ccode\u003e75d33b9\u003c/code\u003e\u003c/a\u003e CUDA: support for weight clamp in top-k norm (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16702\"\u003e#16702\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3470a5c891dcc94363e492a3760af92b6b07241c\"\u003e\u003ccode\u003e3470a5c\u003c/code\u003e\u003c/a\u003e ggml-alloc : make gallocr prefer chunks that allow memory reuse (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16788\"\u003e#16788\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bd562fe4f7bd55625511d5f9d639c4fb1db1d440\"\u003e\u003ccode\u003ebd562fe\u003c/code\u003e\u003c/a\u003e cuda : use fast copy when src and dst are of different type and contiguous (#...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bbac6a26b2bd7f7c1f0831cb1e7b52734c66673b\"\u003e\u003ccode\u003ebbac6a2\u003c/code\u003e\u003c/a\u003e ggml: fix cuda kernel launch configuration for k_compute_batched_ptrs to supp...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/73a48c9790d320476b3e5ef75bda09f2f8269e6e\"\u003e\u003ccode\u003e73a48c9\u003c/code\u003e\u003c/a\u003e convert : enable expert group selection for all models with it (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16691\"\u003e#16691\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f696428ce8e4d16c17acbffeaa7feac3b0fb9061\"\u003e\u003ccode\u003ef696428\u003c/code\u003e\u003c/a\u003e graph : add clamping to ffn_moe_weights_sum to avoid div-by-zero (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16655\"\u003e#16655\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7cce4f8158f0c4c88d8dadd4c23d33938127b897\"\u003e\u003ccode\u003e7cce4f8\u003c/code\u003e\u003c/a\u003e model : set res-\u0026gt;t_embd in SmallThinker models (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16782\"\u003e#16782\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...75cbdd3fce38ea12d50cd19e73a069aa5dbbd5fa\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/29","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/29","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/29/packages"}},{"old_version":"`2be60cb`","new_version":"`f9bc66c`","update_type":null,"path":null,"pr_created_at":"2025-10-13T03:22:29.000Z","version_change":"`2be60cb` → `f9bc66c`","issue":{"uuid":"3508320929","node_id":"PR_kwDOOKZoKM6tXxdH","number":27,"state":"closed","title":"Bump llama.cpp from `2be60cb` to `f9bc66c`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2025-10-20T03:31:58.000Z","author_association":null,"state_reason":null,"created_at":"2025-10-13T03:22:29.000Z","updated_at":"2025-10-20T03:32:00.000Z","time_to_close":605369,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`f9bc66c`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `f9bc66c`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f9bc66c3ebcfddb5f09e4b21253623caeb8e414a\"\u003e\u003ccode\u003ef9bc66c\u003c/code\u003e\u003c/a\u003e CANN: Update several operators to support FP16 data format (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16251\"\u003e#16251\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a31cf36ad946a13b3a646bf0dadf2a481e89f944\"\u003e\u003ccode\u003ea31cf36\u003c/code\u003e\u003c/a\u003e metal : add opt_step_adamw and op_sum (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16529\"\u003e#16529\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/81d54bbfd599811b354c39f04550888168be7780\"\u003e\u003ccode\u003e81d54bb\u003c/code\u003e\u003c/a\u003e webui: remove client-side context pre-check and rely on backend for limits (#...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c7be9febcbafa9af7d1b9443f86475c59c9c5f87\"\u003e\u003ccode\u003ec7be9fe\u003c/code\u003e\u003c/a\u003e [SYCL] fix UT fault cases: count-equal, argsort, pad OPs (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16521\"\u003e#16521\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/8415f61e23d04427cd0d912fbb9d33b85f849456\"\u003e\u003ccode\u003e8415f61\u003c/code\u003e\u003c/a\u003e ci : add Vulkan on Ubuntu with default packages build (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16532\"\u003e#16532\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2c301e91abb92d03c1a682b4b540ba835562a74b\"\u003e\u003ccode\u003e2c301e9\u003c/code\u003e\u003c/a\u003e common : handle unicode during partial json parsing (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16526\"\u003e#16526\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/4b2dae383df708e2afc49c4859a81cd074f5ac10\"\u003e\u003ccode\u003e4b2dae3\u003c/code\u003e\u003c/a\u003e common : update presets (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16504\"\u003e#16504\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/41aac5c69b5fb281bc1f486afb053f78101bb39e\"\u003e\u003ccode\u003e41aac5c\u003c/code\u003e\u003c/a\u003e ggml : Fix FP16 ELU positive branch (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16519\"\u003e#16519\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/a2fba89a426ff8005d303c73f0436e7e67368b70\"\u003e\u003ccode\u003ea2fba89\u003c/code\u003e\u003c/a\u003e hparams : add check for layer index in is_recurrent (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16511\"\u003e#16511\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/20cc625edc2264aae2779e71bef1593e6a4e8c43\"\u003e\u003ccode\u003e20cc625\u003c/code\u003e\u003c/a\u003e ggml: Correct SVE implementation in ggml_vec_dot_f16_unroll (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16518\"\u003e#16518\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...f9bc66c3ebcfddb5f09e4b21253623caeb8e414a\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/27","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/27","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/27/packages"}},{"old_version":"`2be60cb`","new_version":"`ca71fb9`","update_type":null,"path":null,"pr_created_at":"2025-10-06T03:29:59.000Z","version_change":"`2be60cb` → `ca71fb9`","issue":{"uuid":"2888707654","node_id":"PR_kwDOOKZoKM6sLi5G","number":26,"state":"open","title":"Bump llama.cpp from `2be60cb` to `ca71fb9`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":0,"pull_request":true,"closed_at":null,"author_association":"CONTRIBUTOR","state_reason":null,"created_at":"2025-10-06T03:29:59.000Z","updated_at":"2025-10-06T03:30:00.000Z","time_to_close":null,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`ca71fb9`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `ca71fb9`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/ca71fb9b368e3db96e028f80c4c9df6b6b370edd\"\u003e\u003ccode\u003eca71fb9\u003c/code\u003e\u003c/a\u003e model : Granite docling + Idefics3 preprocessing (SmolVLM) (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16206\"\u003e#16206\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/35266573b968e1c947b367782fb4b3eddbb4f3c0\"\u003e\u003ccode\u003e3526657\u003c/code\u003e\u003c/a\u003e ggml webgpu: actually add softmax, fix rms_norm offset (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16400\"\u003e#16400\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/86df2c9ae4f2f1ee63d2558a9dc797b98524639b\"\u003e\u003ccode\u003e86df2c9\u003c/code\u003e\u003c/a\u003e vulkan: use a more appropriate amount of threads when generating shaders (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16\"\u003e#16\u003c/a\u003e...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f39283960b58a92ecc0c72567711318b20e22b55\"\u003e\u003ccode\u003ef392839\u003c/code\u003e\u003c/a\u003e rpc : check src buffer when copying tensor (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16421\"\u003e#16421\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/898acba6816ad23b6a9491347d30e7570bffadfd\"\u003e\u003ccode\u003e898acba\u003c/code\u003e\u003c/a\u003e rpc : add support for multiple devices (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16276\"\u003e#16276\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e29acf74fea996014380d59d31aa504ae8964258\"\u003e\u003ccode\u003ee29acf7\u003c/code\u003e\u003c/a\u003e vulkan : incremental shader builds (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16341\"\u003e#16341\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/128d522c04286e019666bd6ee4d18e3fbf8772e2\"\u003e\u003ccode\u003e128d522\u003c/code\u003e\u003c/a\u003e chat : support Magistral thinking (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16413\"\u003e#16413\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/f6dcda390004b627ef30af378d0c01ad2519289e\"\u003e\u003ccode\u003ef6dcda3\u003c/code\u003e\u003c/a\u003e server : context checkpointing for hybrid and recurrent models (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16382\"\u003e#16382\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/606a73f53175077429484b23dcf799f69a31d0bd\"\u003e\u003ccode\u003e606a73f\u003c/code\u003e\u003c/a\u003e metal : fix loop bound in ggml_mem_ranges (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16412\"\u003e#16412\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/946f71ed9ade07e319859b5ce656144140e066fb\"\u003e\u003ccode\u003e946f71e\u003c/code\u003e\u003c/a\u003e llama : fix shapes for bert/mpt q/k norm (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16409\"\u003e#16409\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...ca71fb9b368e3db96e028f80c4c9df6b6b370edd\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/26","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/26","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/26/packages"}},{"old_version":"`2be60cb`","new_version":"`b887d2f`","update_type":null,"path":null,"pr_created_at":"2025-09-29T03:47:20.000Z","version_change":"`2be60cb` → `b887d2f`","issue":{"uuid":"2868919049","node_id":"PR_kwDOOKZoKM6rADsJ","number":25,"state":"open","title":"Bump llama.cpp from `2be60cb` to `b887d2f`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":0,"pull_request":true,"closed_at":null,"author_association":"CONTRIBUTOR","state_reason":null,"created_at":"2025-09-29T03:47:20.000Z","updated_at":"2025-09-29T03:47:21.000Z","time_to_close":null,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`b887d2f`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `b887d2f`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b887d2f3413ac231e3cb5925260c39902af4a70c\"\u003e\u003ccode\u003eb887d2f\u003c/code\u003e\u003c/a\u003e ggml : fix GGML_F32_VEC_FMA argument order in ggml_vec_mad1_f32 (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16307\"\u003e#16307\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bd0af02fc96c2057726f33c0f0daf7bb8f3e462a\"\u003e\u003ccode\u003ebd0af02\u003c/code\u003e\u003c/a\u003e common : fix reasoning before forced tool call via tool_choice = required (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/1\"\u003e#1\u003c/a\u003e...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d9e0e7c8194dfd7d23bf3a86608c9ece68d77c93\"\u003e\u003ccode\u003ed9e0e7c\u003c/code\u003e\u003c/a\u003e ci : fix musa docker build (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16306\"\u003e#16306\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/0124ac989f7e7bf08803788f66dbe4106bdcdd58\"\u003e\u003ccode\u003e0124ac9\u003c/code\u003e\u003c/a\u003e devops: switch to using ubuntu-22.04-s390x image (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16302\"\u003e#16302\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/2811c65286ae954bec87049f75b86dc022006dcc\"\u003e\u003ccode\u003e2811c65\u003c/code\u003e\u003c/a\u003e Fixed a few typos in the README of the LLaMA.cpp HTTP Server [no ci] (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16297\"\u003e#16297\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d8359f5fde480da030bf75c7711573c7c4d993ba\"\u003e\u003ccode\u003ed8359f5\u003c/code\u003e\u003c/a\u003e vulkan: 64-bit im2col (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16135\"\u003e#16135\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/6a2c6145a0b91b40eb3c3dba7b20ccc4b270490f\"\u003e\u003ccode\u003e6a2c614\u003c/code\u003e\u003c/a\u003e metal : extend mat-mat multiplication support (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16225\"\u003e#16225\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3b53634fe35771e2e318227aa81585726bae7234\"\u003e\u003ccode\u003e3b53634\u003c/code\u003e\u003c/a\u003e metal : fuse non-sequential nodes (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16102\"\u003e#16102\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1384abf8b8d5894d32fada453ccf4d196ffba7de\"\u003e\u003ccode\u003e1384abf\u003c/code\u003e\u003c/a\u003e vulkan: handle mat_mul with A matrix \u0026gt; 4GB (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16176\"\u003e#16176\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e6d65fb02d553bd79cad94e517cdca18b687788d\"\u003e\u003ccode\u003ee6d65fb\u003c/code\u003e\u003c/a\u003e vulkan: support arbitrary KV dimension in flash attention (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16160\"\u003e#16160\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...b887d2f3413ac231e3cb5925260c39902af4a70c\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/25","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/25","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/25/packages"}},{"old_version":"`2be60cb`","new_version":"`51f5a45`","update_type":null,"path":null,"pr_created_at":"2025-09-22T03:19:27.000Z","version_change":"`2be60cb` → `51f5a45`","issue":{"uuid":"2847775100","node_id":"PR_kwDOOKZoKM6pvZl8","number":24,"state":"closed","title":"Bump llama.cpp from `2be60cb` to `51f5a45`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":1,"pull_request":true,"closed_at":"2025-09-29T03:47:22.000Z","author_association":"CONTRIBUTOR","state_reason":null,"created_at":"2025-09-22T03:19:27.000Z","updated_at":"2025-09-29T03:47:22.000Z","time_to_close":606475,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`51f5a45`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `51f5a45`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/51f5a45fbe575dcd54bdd2a339ef8e8424d1c12a\"\u003e\u003ccode\u003e51f5a45\u003c/code\u003e\u003c/a\u003e opencl: fix concat crash on win arm64 with Adreno (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15944\"\u003e#15944\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c4510dc9374e17dcb8726902ab5216067a92b3d3\"\u003e\u003ccode\u003ec4510dc\u003c/code\u003e\u003c/a\u003e opencl: initial \u003ccode\u003eq8_0\u003c/code\u003e mv support (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15732\"\u003e#15732\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/da30ab5f8696cabb2d4620cdc0aa41a298c54fd6\"\u003e\u003ccode\u003eda30ab5\u003c/code\u003e\u003c/a\u003e ci : add label for the RISC-V runner (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16150\"\u003e#16150\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/28baac9c9f491c872e2c37762d3bd90446b005e9\"\u003e\u003ccode\u003e28baac9\u003c/code\u003e\u003c/a\u003e ci : migrate ggml ci to self-hosted runners (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16116\"\u003e#16116\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/1eeb523c3e0c7ffbd59469f5463dcbdecba3535e\"\u003e\u003ccode\u003e1eeb523\u003c/code\u003e\u003c/a\u003e vulkan: optimize UMA buffer operations and fix driver hangs (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16059\"\u003e#16059\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/5bb4a3edec297e74b0f7bd4ed5d0fdd12e28d858\"\u003e\u003ccode\u003e5bb4a3e\u003c/code\u003e\u003c/a\u003e vulkan: fix validation error about VK_PIPELINE_CREATE_CAPTURE_STATISTICS_BIT_...\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7f766929ca8e8e01dcceb1c526ee584f7e5e1408\"\u003e\u003ccode\u003e7f76692\u003c/code\u003e\u003c/a\u003e sync : ggml\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/405921dcefdd4e90ed948a4bf179007c2fa92b2d\"\u003e\u003ccode\u003e405921d\u003c/code\u003e\u003c/a\u003e ggml : introduce semantic versioning (ggml/1336)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/fa6383ca7e7ccb8ca3bdfeb37e348ddc4113aa26\"\u003e\u003ccode\u003efa6383c\u003c/code\u003e\u003c/a\u003e CUDA : conditionally add cuda architectures (ggml/1341)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/803dac2e48ef3ba26a504eb27c4e77ec2d21f7d0\"\u003e\u003ccode\u003e803dac2\u003c/code\u003e\u003c/a\u003e vulkan: use vec dot for matrix matrix multiplications (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/16056\"\u003e#16056\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...51f5a45fbe575dcd54bdd2a339ef8e8424d1c12a\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/24","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/24","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/24/packages"}},{"old_version":"`2be60cb`","new_version":"`85ca66a`","update_type":null,"path":null,"pr_created_at":"2025-09-08T03:38:55.000Z","version_change":"`2be60cb` → `85ca66a`","issue":{"uuid":"2806712558","node_id":"PR_kwDOOKZoKM6nSwju","number":22,"state":"open","title":"Bump llama.cpp from `2be60cb` to `85ca66a`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":0,"pull_request":true,"closed_at":null,"author_association":"CONTRIBUTOR","state_reason":null,"created_at":"2025-09-08T03:38:55.000Z","updated_at":"2025-09-08T03:38:55.000Z","time_to_close":null,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`85ca66a`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `85ca66a`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/85ca66a74676e6d5df4433016488e039a4b464ae\"\u003e\u003ccode\u003e85ca66a\u003c/code\u003e\u003c/a\u003e CANN: Stream sync between devices for acl_graph (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15809\"\u003e#15809\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3976dfbe00f02a62c0deca32c46138e4f0ca81d8\"\u003e\u003ccode\u003e3976dfb\u003c/code\u003e\u003c/a\u003e vulkan: support im2col_3d (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15795\"\u003e#15795\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/d36e61c580bf7fc7879c443c542312a42b718e11\"\u003e\u003ccode\u003ed36e61c\u003c/code\u003e\u003c/a\u003e ggml-cpu: clean up s390x SIMD (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15855\"\u003e#15855\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c97b5e5854b47b18a248d77edb693c63018a0865\"\u003e\u003ccode\u003ec97b5e5\u003c/code\u003e\u003c/a\u003e vulkan: Support pad_ext (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15794\"\u003e#15794\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/267e99867f09bec8bcc2e424ad9bcddd6cccf9d0\"\u003e\u003ccode\u003e267e998\u003c/code\u003e\u003c/a\u003e vulkan: Use larger loads in scalar/coopmat1 matmul (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15729\"\u003e#15729\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3b15924d71237a43bb5ad71f5b885ee66a821342\"\u003e\u003ccode\u003e3b15924\u003c/code\u003e\u003c/a\u003e ggml WebGPU: remove userdata from request adapter callback (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15527\"\u003e#15527\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/79bc429262268ad2ac8a364cfe6c2d6b9c5f008a\"\u003e\u003ccode\u003e79bc429\u003c/code\u003e\u003c/a\u003e CUDA: faster tile FA (Pascal/AMD), headsize 256 (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15769\"\u003e#15769\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/c4df49a42d396bdf7344501813e7de53bc9e7bb3\"\u003e\u003ccode\u003ec4df49a\u003c/code\u003e\u003c/a\u003e kleidiai: generalize compute_forward_kv_cache to compute_forward_fp16 (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15817\"\u003e#15817\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3c3635d2f20424d557b5b0605a2a356214ffe048\"\u003e\u003ccode\u003e3c3635d\u003c/code\u003e\u003c/a\u003e server : speed up tests (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15836\"\u003e#15836\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/61bdfd5298a78593be649a1035ee2a120b13c4f0\"\u003e\u003ccode\u003e61bdfd5\u003c/code\u003e\u003c/a\u003e server : implement prompt processing progress report in stream mode (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15827\"\u003e#15827\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...85ca66a74676e6d5df4433016488e039a4b464ae\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/22","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/22","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/22/packages"}},{"old_version":"`2be60cb`","new_version":"`b66df9d`","update_type":null,"path":null,"pr_created_at":"2025-09-01T07:13:39.000Z","version_change":"`2be60cb` → `b66df9d`","issue":{"uuid":"2788776505","node_id":"PR_kwDOOKZoKM6mOVo5","number":21,"state":"open","title":"Bump llama.cpp from `2be60cb` to `b66df9d`","user":"dependabot[bot]","labels":["dependencies","submodules"],"assignees":[],"locked":false,"comments_count":0,"pull_request":true,"closed_at":null,"author_association":"CONTRIBUTOR","state_reason":null,"created_at":"2025-09-01T07:13:39.000Z","updated_at":"2025-09-01T07:13:40.000Z","time_to_close":null,"merged_at":null,"merged_by":null,"closed_by":null,"dependency_metadata":{"prefix":"Bump","packages":[{"name":"llama.cpp","old_version":"`2be60cb`","new_version":"`b66df9d`","repository_url":"https://github.com/ggml-org/llama.cpp"}],"path":null,"ecosystem":"submodules"},"body":"Bumps [llama.cpp](https://github.com/ggml-org/llama.cpp) from `2be60cb` to `b66df9d`.\n\u003cdetails\u003e\n\u003csummary\u003eCommits\u003c/summary\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b66df9d9c942254d03209186ef24ed7c994a576e\"\u003e\u003ccode\u003eb66df9d\u003c/code\u003e\u003c/a\u003e CUDA: fix build error from ambiguous __half conversions in conv2d (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15690\"\u003e#15690\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/b9382c3877c6067feccf182efe9449a2d1cb24c7\"\u003e\u003ccode\u003eb9382c3\u003c/code\u003e\u003c/a\u003e CANN: Optimize MUL_MAT_ID (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15658\"\u003e#15658\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/3dc7397a2799bdc07bccf637ab7ae5a1e786d1a4\"\u003e\u003ccode\u003e3dc7397\u003c/code\u003e\u003c/a\u003e CANN: fix RoPE cache issue on multi-device (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15629\"\u003e#15629\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/e92d53b29e393fc4c0f9f1f7c3fe651be8d36faa\"\u003e\u003ccode\u003ee92d53b\u003c/code\u003e\u003c/a\u003e sampling : optimize samplers by reusing bucket sort (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15665\"\u003e#15665\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/0d161f021aa33ec0e90cce96f5d1a88925557327\"\u003e\u003ccode\u003e0d161f0\u003c/code\u003e\u003c/a\u003e server : enable /slots by default and make it secure (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15630\"\u003e#15630\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/4efd5a83163ff383285b3a4c2106feabf5c69557\"\u003e\u003ccode\u003e4efd5a8\u003c/code\u003e\u003c/a\u003e metal : fix checks for available FA kernels (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15700\"\u003e#15700\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/274966226f87f301ac132da898280ca3142b60e5\"\u003e\u003ccode\u003e2749662\u003c/code\u003e\u003c/a\u003e llama : fix fattn reserve call n_seqs parameter (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15699\"\u003e#15699\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/9777032dccd67bdc7785aeab7497014a8be8dacc\"\u003e\u003ccode\u003e9777032\u003c/code\u003e\u003c/a\u003e llama : separate compute buffer reserve from fattn check (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15696\"\u003e#15696\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/7d3c9f2b217acf0ce5db81ae83d3f375f49ab2c7\"\u003e\u003ccode\u003e7d3c9f2\u003c/code\u003e\u003c/a\u003e ci : explicitly set fa off or on (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15692\"\u003e#15692\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://github.com/ggml-org/llama.cpp/commit/bbbf5ecccb35286521f735239d499eec4279a840\"\u003e\u003ccode\u003ebbbf5ec\u003c/code\u003e\u003c/a\u003e vulkan: handle large sizes for get_rows (\u003ca href=\"https://redirect.github.com/ggml-org/llama.cpp/issues/15686\"\u003e#15686\u003c/a\u003e)\u003c/li\u003e\n\u003cli\u003eAdditional commits viewable in \u003ca href=\"https://github.com/ggml-org/llama.cpp/compare/2be60cbc2707359241c2784f9d2e30d8fc7cdabb...b66df9d9c942254d03209186ef24ed7c994a576e\"\u003ecompare view\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/details\u003e\n\u003cbr /\u003e\n\n\nDependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.\n\n[//]: # (dependabot-automerge-start)\n[//]: # (dependabot-automerge-end)\n\n---\n\n\u003cdetails\u003e\n\u003csummary\u003eDependabot commands and options\u003c/summary\u003e\n\u003cbr /\u003e\n\nYou can trigger Dependabot actions by commenting on this PR:\n- `@dependabot rebase` will rebase this PR\n- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it\n- `@dependabot merge` will merge this PR after your CI passes on it\n- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it\n- `@dependabot cancel merge` will cancel a previously requested merge and block automerging\n- `@dependabot reopen` will reopen this PR if it is closed\n- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually\n- `@dependabot show \u003cdependency name\u003e ignore conditions` will show all of the ignore conditions of the specified dependency\n- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)\n- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)\n\n\n\u003c/details\u003e","html_url":"https://github.com/TheCBaH/devcontainer.llama.cpp/pull/21","url":"https://dependabot.ecosyste.ms/api/v1/hosts/GitHub/repositories/TheCBaH%2Fdevcontainer.llama.cpp/issues/21","packages_url":"https://dependabot.ecosyste.ms/api/v1/issues/21/packages"}}]}