ggerganov/llama.cpp

Pull Requests Commits

Use dot instruction to multiply two values at once, to enable dual issue instructions

0cc4m committed 151 days ago

ab27292b

Vulkan: Remove dedicated aligned matrix matrix multiplication shaders

0cc4m committed 151 days ago

039cc074

vulkan: fix NaN issue in flash attention shader (#12776)

jeffbolznv committed 151 days ago

Verified 0c74b043

vulkan: Use unclamped loads for flash attention mask (#12720)

jeffbolznv committed 151 days ago

Verified 80b717d4

Vulkan: Tune Vulkan mmq int dot shader for performance (#12767)

0cc4m committed 152 days ago

Verified 6bf28f01

common : fix includes in arg.cpp and gemma3-cli.cpp (#12766)

barracuda156 committed 152 days ago

Verified f1e3eb42

clip : refactor clip_init, add tests (#12757)

ngxson committed 152 days ago

Verified 0364178c

common: custom hf endpoint support (#12769)

eternaphia committed 152 days ago

Verified c6ff5d2a

sync: minja (#12739)

ochafik committed 153 days ago

Verified 7a84777f

kv-cache : simplify + fix warning for recurrent models (#12756)

ggerganov committed 153 days ago

Verified 3e1d2934

ci: add Linux cross-compile build (#12428)

bandoti committed 153 days ago

Verified 1be76e46

server : webui : Upgrade daisyui, tailwindcss. (#12735)

nauful committed 153 days ago

Verified b7723942

gguf-split : --merge now respects --dry-run option (#12681)

nickhuang99 committed 153 days ago

Verified 23106f94

sycl: allow ggml-sycl configuration and compilation using Visual Studio project/solution (#12625)

s-Nick committed 153 days ago

Verified 94148ba3

cmake: fix ggml-shaders-gen compiler paths containing spaces (#12747)

hydroo committed 153 days ago

Verified 9ac4d611

docs : add XCFramework section to README.md [no ci] (#12746)

danbev committed 153 days ago

Verified 348888e0

vulkan: Hybrid waitForFences/getFenceStatus to reduce fence latency (#12630)

jeffbolznv committed 154 days ago

Verified 74d4f5b0

vulkan: set cmake minimum and project name in vulkan-shaders (#12744)

jeffbolznv committed 154 days ago

Verified 35e592eb

opencl: update doc for OpenCL (#12702)

lhez committed 154 days ago

Verified 7d7b1baf

CUDA: Prefer vector flash decoding kernel for Gemma models (#12738)

gaugarg-nv committed 154 days ago

Verified c262bedd

vocab : use string_view::find() to avoid unnecessary looking up beyond the fragment range (#12706)

yumeyao committed 154 days ago

Verified 5dd5d1ab

vulkan: Fix missing cmake logic for dot product extension (#12721)

jeffbolznv committed 154 days ago

Verified 1c059995

ci : add env variable in ggml-ci and document the same in SYCL.md (#12736)

AD2605 committed 154 days ago

Verified 2004644b

sync : minja (inclusionAI/Ling) and update tests (#12699)

yeahdongcn committed 154 days ago

Verified 5f696e88

fix MUSA compiler warning (#12704)

A3shTnT committed 154 days ago

Verified 193c3e03

CANN: Support operator SIN COS ARGMAX (#12709)

noemotiovon committed 154 days ago

Verified 65cfe136

Simplify and improve CUDA graphs through use of indirect copy pointers (#9017)

agray3 committed 155 days ago

Verified 3f9da22c

CANN: Fix failed test cases (#12708)

hipudding committed 155 days ago

Verified 2a0dc97e

opencl: use `max_alloc_size` in backend ctx instead of querying again (#12705)

lhez committed 155 days ago

Verified 97a20c01

vulkan: Implement split_k for coopmat2 flash attention. (#12627)

jeffbolznv committed 155 days ago

Verified f01bd023

Older