ngxson/llama.cpp

Pull Requests Commits

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

chansikpark committed 111 days ago

Verified 333f2595

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

jeffbolznv committed 111 days ago

Verified 53d7d21e

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

gSUz92nc committed 111 days ago

Verified eeee367d

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

am17an committed 111 days ago

Verified 64fe17fb

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

am17an committed 112 days ago

Verified c1b18768

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

SavicStefan committed 112 days ago

Verified b8a5cfd1

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

AlekseiNikiforovIBM committed 112 days ago

Verified 08416ebe

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

jeffbolznv committed 112 days ago

Verified b4e335d8

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

jeffbolznv committed 112 days ago

Verified d6fe40fa

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

JohannesGaessler committed 112 days ago

Verified e14e842e

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

reeselevine committed 112 days ago

Verified 647b960b

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

bssrdf committed 112 days ago

Verified 299f5d78

vulkan : refactor buffer handling in vk_op_f32 (#16840)

Acly committed 112 days ago

Verified ac76d362

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

JohannesGaessler committed 112 days ago

Verified 65156105

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

ggerganov committed 112 days ago

Verified 7956bb4d

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

CISC committed 112 days ago

Verified 9008027a

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

ggerganov committed 112 days ago

Verified 16bcc125

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

angt committed 112 days ago

Verified 9eb9a133

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

lizhenneng committed 112 days ago

Verified 7c23f3f0

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

ggerganov committed 113 days ago

Verified 8c0d6bb4

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

ngxson committed 113 days ago

Verified 5c9a18e6

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

xctan committed 113 days ago

Verified 7f09a680

CUDA: fix crash on uneven context without FA (#16988)

JohannesGaessler committed 113 days ago

Verified aa374175

metal : initial Metal4 tensor API support (#16634)

ggerganov committed 114 days ago

Verified 5b180c3d

server : disable checkpoints with mtmd (#17045)

ggerganov committed 114 days ago

Verified b7f9010d

clip: implement minicpm-v sinusoidal embd using GGML (#17036)

ngxson committed 114 days ago

Verified 4882f0ff

sycl: add CONCAT operator support (#16047)

ye-NX committed 114 days ago

Verified 9d7c518d

docs: explain CUDA 11 compilation [no ci] (#16824)

JohannesGaessler committed 114 days ago

Verified 22c8c3c6

ggml-hexagon: graceful fallback for older socs where rpcmem_alloc2 and FASTRPC_GET_URI is unsupported (#16987)

l3utterfly committed 114 days ago

Verified 6db3d1ff

improve CUDA cpy memory bandwidth when copying transposed tensor (#16841)

bssrdf committed 114 days ago

Verified 230d1169

Newer Older