ggerganov/llama.cpp

Pull Requests Commits

rpc : fix alloc size logic

ggerganov committed 10 hours ago

Verified e118dc9b

benches : add folder with benchmarks (#16931)

ggerganov committed 10 hours ago

Verified ef1d8269

Switch to using Ubuntu 25.10 vulkan/mesa (#16497)

ericcurtin committed 12 hours ago

Verified 86fde91e

vulkan: iGPU memory reporting fix (#17110)

0cc4m committed 12 hours ago

Verified 7f3e9d33

vulkan: fix mmq out of bounds reads (#17108)

0cc4m committed 12 hours ago

Verified 8a3519b7

vulkan: fuse mul_mat_id + mul (#17095)

jeffbolznv committed 13 hours ago

Verified 80a6cf63

metal : retain src and dst buffers during async ops (#17101)

ggerganov committed 15 hours ago

Verified 0750a599

arg: add --cache-list argument to list cached models (#17073)

ngxson committed 1 day ago

Verified aa3b7a90

webui: fix keyboard shortcuts for new chat & edit chat title (#17007)

chansikpark committed 1 day ago

Verified 333f2595

vulkan: Use spec constants for conv2d s/d/p and kernel W/H (#16978)

jeffbolznv committed 1 day ago

Verified 53d7d21e

server: fix correct time_ms calculation in prompt_progress (#17093)

gSUz92nc committed 1 day ago

Verified eeee367d

Revert "CUDA: add expert reduce kernel (#16857)" (#17100)

am17an committed 1 day ago

Verified 64fe17fb

CUDA: skip fusion for repeating adds in bias (#17080)

am17an committed 1 day ago

Verified c1b18768

vulkan: Increase BK to 32; use BK/4 for non-CM mul_mm.comp (#16636)

SavicStefan committed 1 day ago

Verified b8a5cfd1

ggml: disable vxe for cross-compilation by default (#16966)

AlekseiNikiforovIBM committed 1 day ago

Verified 08416ebe

vulkan: fuse rms_norm + mul + rope (+ view + set_rows) (#16977)

jeffbolznv committed 1 day ago

Verified b4e335d8

vulkan: Fix test-thread-safety crashes (#17024)

jeffbolznv committed 1 day ago

Verified d6fe40fa

CUDA: fix MMQ stream-k fixup ne1 indices (#17089)

JohannesGaessler committed 1 day ago

Verified e14e842e

ggml webgpu: faster matrix multiplication/matrix-vector multiplication (#17031)

reeselevine committed 1 day ago

Verified 647b960b

CUDA: properly handle nb00=nb02 case for cpy (#17081)

bssrdf committed 1 day ago

Verified 299f5d78

vulkan : refactor buffer handling in vk_op_f32 (#16840)

Acly committed 2 days ago

Verified ac76d362

CUDA: fix should_use_mmvf for ne11 == 1 (#17085)

JohannesGaessler committed 2 days ago

Verified 65156105

bench : cache the llama_context state at computed depth (#16944)

ggerganov committed 2 days ago

Verified 7956bb4d

hparams : add n_embd_inp() to support extended embed (#16928)

CISC committed 2 days ago

Verified 9008027a

kv-cache : pad the cache size to 256 for performance (#17046)

ggerganov committed 2 days ago

Verified 16bcc125

Revert "ggml-cpu: detect correct cpu flags for arm64 (#16229) (#16239)" (#17084)

angt committed 2 days ago

Verified 9eb9a133

ggml-cpu: detect correct cpu flags for arm64 (#16229) (#16239)

lizhenneng committed 2 days ago

Verified 7c23f3f0

server : print the samplers chain for each request (#17070)

ggerganov committed 2 days ago

Verified 8c0d6bb4

common: move download functions to download.(cpp|h) (#17059)

ngxson committed 2 days ago

Verified 5c9a18e6

ggml-cpu : optimize RVV q2_k and q3_k kernels (#16887)

xctan committed 3 days ago

Verified 7f09a680

Older