Search results

ggml : support broadcast for ggml_soft_max_ext and ggml_flash_attn_ext

Apple Metalhttps://en.wikipedia.org/wiki/Metal_(API)https://en.wikipedia.org/wiki/Metal_(API)

Ascend NPUissues specific to Ascend NPUsissues specific to Ascend NPUs

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

SYCLhttps://en.wikipedia.org/wiki/SYCL - GPU programming languagehttps://en.wikipedia.org/wiki/SYCL - GPU programming language

testingEverything test relatedEverything test related

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open (in progress).

ggml-org/llama.cppnumber 14435

#14435 In ggml-org/llama.cpp;

· ggerganov opened on Jun 28, 2025

vulkan: Add fusion support for RMS_NORM+MUL

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

testingEverything test relatedEverything test related

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open (in progress).

ggml-org/llama.cppnumber 14366

#14366 In ggml-org/llama.cpp;

· jeffbolznv opened on Jun 24, 2025

ggml: adds CONV_2D op and direct GEMM Vulkan implementation

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

testingEverything test relatedEverything test related

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open (in progress).

ggml-org/llama.cppnumber 14316

#14316 In ggml-org/llama.cpp;

· etasnadi opened on Jun 21, 2025

ggml : implement REGLU/GEGLU/SWIGLU ops

Apple Metalhttps://en.wikipedia.org/wiki/Metal_(API)https://en.wikipedia.org/wiki/Metal_(API)

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

help wantedExtra attention is neededExtra attention is needed

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

SYCLhttps://en.wikipedia.org/wiki/SYCL - GPU programming languagehttps://en.wikipedia.org/wiki/SYCL - GPU programming language

testingEverything test relatedEverything test related

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Draft (not ready).

ggml-org/llama.cppnumber 14158

#14158 In ggml-org/llama.cpp;

· CISC opened on Jun 12, 2025

Fix Vulkan glslc invocation command lines

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open (in progress).

ggml-org/llama.cppnumber 13289

#13289 In ggml-org/llama.cpp;

· kangalio opened on May 3, 2025

llama : try loading tensors with pre-computed hashes

Apple Metalhttps://en.wikipedia.org/wiki/Metal_(API)https://en.wikipedia.org/wiki/Metal_(API)

examples

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Komputehttps://github.com/KomputeProject/kompute/https://github.com/KomputeProject/kompute/

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

SYCLhttps://en.wikipedia.org/wiki/SYCL - GPU programming languagehttps://en.wikipedia.org/wiki/SYCL - GPU programming language

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open (in progress).

ggml-org/llama.cppnumber 13106

#13106 In ggml-org/llama.cpp;

· rgerganov opened on Apr 25, 2025

llama-server : implement universal assisted decoding

androidIssues specific to AndroidIssues specific to Android

Apple Metalhttps://en.wikipedia.org/wiki/Metal_(API)https://en.wikipedia.org/wiki/Metal_(API)

Ascend NPUissues specific to Ascend NPUsissues specific to Ascend NPUs

buildCompilation issuesCompilation issues

devopsimprovements to build systems and github actionsimprovements to build systems and github actions

documentationImprovements or additions to documentationImprovements or additions to documentation

examples

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

pythonpython script changespython script changes

scriptScript relatedScript related

server

SYCLhttps://en.wikipedia.org/wiki/SYCL - GPU programming languagehttps://en.wikipedia.org/wiki/SYCL - GPU programming language

testingEverything test relatedEverything test related

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open (in progress).

ggml-org/llama.cppnumber 12635

#12635 In ggml-org/llama.cpp;

· g2mt opened on Mar 28, 2025

Vulkan: Remove dedicated aligned matrix matrix multiplication shaders

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

testingEverything test relatedEverything test related

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Draft (not ready).

ggml-org/llama.cppnumber 12515

#12515 In ggml-org/llama.cpp;

· 0cc4m opened on Mar 22, 2025

Fixed Eval Bug: 12163 : Fallback to CPU when loading model: vk::PhysicalDevice::createDevice: ErrorExtensionNotPresent.

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open (in progress).

ggml-org/llama.cppnumber 12329

#12329 In ggml-org/llama.cpp;

· ashwini778 opened on Mar 11, 2025

<code>tool-call</code>: Phi-4 support

androidIssues specific to AndroidIssues specific to Android

Apple Metalhttps://en.wikipedia.org/wiki/Metal_(API)https://en.wikipedia.org/wiki/Metal_(API)

devopsimprovements to build systems and github actionsimprovements to build systems and github actions

documentationImprovements or additions to documentationImprovements or additions to documentation

examples

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

pythonpython script changespython script changes

server

SYCLhttps://en.wikipedia.org/wiki/SYCL - GPU programming languagehttps://en.wikipedia.org/wiki/SYCL - GPU programming language

testingEverything test relatedEverything test related

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open (in progress).

ggml-org/llama.cppnumber 12288

#12288 In ggml-org/llama.cpp;

· jpohhhh opened on Mar 9, 2025

vulkan: optimization proposals for coopmat1 mul_mm

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Draft (not ready).

ggml-org/llama.cppnumber 12260

#12260 In ggml-org/llama.cpp;

· remyoudompheng opened on Mar 7, 2025

vulkan : add GGML_VK_FORCE_HEAP_INDEX env var

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

VulkanIssues specific to the Vulkan backendIssues specific to the Vulkan backend

Status: Open (in progress).

ggml-org/llama.cppnumber 9734

#9734 In ggml-org/llama.cpp;

· gyf304 opened on Oct 4, 2024