ggml-org/llama.cpp

Pull Requests Commits

vulkan: Change the mul_mm shared memory and register caching system to use vec2 instead of scalars, to enable using dot2 instructions

0cc4m committed 2 hours ago

be5a0a8b

CUDA: fix FA occupancy, optimize tile kernel (#15982)

JohannesGaessler committed 5 hours ago

Verified c959b676

common : Fix corrupted memory error on json grammar initialization (#16038)

dralves committed 10 hours ago

Verified cd08fc3e

vulkan: automatically remove unsupported devices (#15976)

netrunnereve committed 11 hours ago

Verified cb5bb6cc

ci : revert back to macos-13 for macOS-latest-cmake-x64 (#16040)

danbev committed 11 hours ago

Verified a91d035b

llama-quant : fix the verification of attention layers for encoder-decoder models (#16023)

DamonFool committed 11 hours ago

Verified 745cbcf2

examples : support encoder-decoder models in the simple example (#16002)

DamonFool committed 11 hours ago

Verified 1cbd80f8

model : add OLMo3 support (#16015)

2015aroras committed 11 hours ago

Verified 85286f35

CANN: Optimize ggml_cann_set_device (#15935)

noemotiovon committed 12 hours ago

Verified d5fabe36

llama-bench: add --n-cpu-moe support (#15952)

jacekpoplawski committed 1 day ago

Verified 8ff20609

ci : use macos-latest for arm64 webgpu build (#16029)

danbev committed 1 day ago

Verified 77475530

ggml : fix padding in timestep embedding kernels (#15932)

danbev committed 1 day ago

Verified 3913f873

ci : upload xcframework artifact from ios-xcode-build job (#16010)

danbev committed 1 day ago

Verified 76888d20

fix: apply clang-format to CUDA macros (#16017)

bugparty committed 1 day ago

Verified f1fbffb5

ci : update macos-latest* jobs to use macos-latest (#15938)

danbev committed 1 day ago

Verified 51abc96b

cmake : Do not install tools on iOS targets (#15903)

ykhrustalev committed 1 day ago

Verified 07808ebb

Add LLaDA-7b-MoE diffusion model (#16003)

am17an committed 1 day ago

Verified 6d758839

CUDA: fix im2col_3d to respect non-contiguous inputs (views) (#15956)

jakekarnes42 committed 1 day ago

Verified 3d4053f7

docker : enable rocWMMA in ROCm images, add gfx1151 (#15997)

slaren committed 1 day ago

Verified dc381aa9

releases : switch to rocWMMA develop branch, add gfx1151 (#15992)

slaren committed 1 day ago

Verified 10d19740

SYCL: Add COUNT_EQUAL operator support (#15991)

yael-works committed 2 days ago

Verified b907255f

llama-run: Fix model download on Windows (#15988)

npopov-vst committed 2 days ago

Verified 28c39da7

CUDA: some micro-optimizations in mmf.cuh for mul_mat_id (#15926)

am17an committed 2 days ago

Verified 10622056

fix KLD percentile output (#15999)

ddh0 committed 2 days ago

Verified a68f31ed

model : add grok-2 support (#15539)

CISC committed 2 days ago

Verified b8e09f08

server : only attempt to enable thinking if using jinja (#15967)

CISC committed 2 days ago

Verified 6c019cb0

metal : remove memory pools (#15966)

ggerganov committed 2 days ago

Verified 9dcd200d

rocm.Dockerfile: added gfx1200,gfx1201 architectures to support AMD Radeon RX 9000 series (#15994)

channeladam committed 3 days ago

Verified 0fa154e3

Vulkan: Clean up mul_mm shader (#15987)

0cc4m committed 3 days ago

Verified 261e6a20

build: fix the build failures of Windows HIP release job (#15984)

lcy0321 committed 3 days ago

Verified a0e13dcb

Older