ggml-org/llama.cpp

Pull Requests Commits

llama: fix whole source code rebuilt on each new commit

ngxson committed 26 days ago

ea14703b

mtmd: fix idefics3 preproc (#28273)

ngxson committed 26 days ago

Verified 67a17c17

finetune: fix no KV cache (#27199)

ngxson committed 26 days ago

Verified 159b7414

server : accept data: URLs for input_video and input_audio (#27735)

geckguy committed 27 days ago

Verified 9cffdcc8

ggml-hexagon: add F16 support for unary ops (#28228)

cqderek committed 27 days ago

Verified f027c4f1

mtmd: add mtmd_tokenize_from_parts() (#28250)

ngxson committed 27 days ago

Verified 73390547

metal : add fa-vec tunings for M3 (#28236)

init-22 committed 27 days ago

Verified 9cc33944

metal : fix memory query under low-memory conditions (#27701)

madsmtm committed 27 days ago

Verified 8c0b9cd0

ci : check for missing autoreleasepools (#27884)

nikwen committed 27 days ago

Verified 03dbcc53

Update ROCm to 10.0.0 release (#27803)

superm1 committed 27 days ago

Verified cff18443

model: correctly support input vision for deepseek4 (#28154)

ngxson committed 27 days ago

Verified 9400c894

ci : enable hf-jobs on server-cuda (#28258)

CISC committed 27 days ago

Verified d5fec32a

ggml-cuda : remove unused vars (#28235)

angt committed 27 days ago

Verified 3d3d7c81

common, server : enable preserve_reasoning kwarg by default, log its effective state (#28174)

ggerganov committed 27 days ago

Verified e750b887

mtmd: support DeepSeek-V4-Flash-Vision-Exp (#28133)

ngxson committed 27 days ago

Verified 7798007a

CUDA + ggml: add sparse-fa for DSV4/GLM (#27970)

am17an committed 27 days ago

Verified 8e93a977

mtmd: Fix Qwen3-tts-0.6b (#28231)

ServeurpersoCom committed 27 days ago

Verified 0f3a71be

ggml: avoid KleidiAI buffer type init on dispatch (#27891)

ac-mmi committed 27 days ago

Verified b81c99b4

hexagon: MUL_MAT and MUL_MAT_ID fusion and fixes (#28202)

max-krasnyansky committed 27 days ago

Verified 960dffab

vulkan: handle larger batch sizes (>4) efficiently for IQ3_S mat-vec (#27449)

LaurentZuijdwijk committed 27 days ago

Verified ba8818cb

vulkan : only request VK_KHR_shader_bfloat16 extension if supported (#28155)

madsmtm committed 27 days ago

Verified 56dd8150

ggml-cpu : conditionally add SpacemiT IME kernel sources (#27961)

alanhc committed 27 days ago

Verified 2637dfe3

opencl: fix out‐of‐bound reads in the Adreno image kernels (#27632)

wanghqc committed 27 days ago

Verified 43d87ff2

hexagon: add missing FARF logs for cpy/get_rows/set_rows/gdn ops (#28217)

trivikram-reddy1 committed 27 days ago

Verified 69320fef

metal : add metallib build support for xcframework (#28163)

jhen0409 committed 27 days ago

Verified b96806d9

cuda: fuse MoE weighted expert reduction (#25952)

anujj committed 28 days ago

Verified 3466812d

kv-cells: look up the n-gram history in the sequence position index (#28040)

ServeurpersoCom committed 28 days ago

Verified b356fa26

context : autoscale n_ctx_train when yarn scaling specified (#28030)

CISC committed 28 days ago

Verified dfc29b64

models : appropriately flag noscan ssm_a tensors (#28121)

CISC committed 28 days ago

Verified f28493c7

model : fix gemma4-assistant (#28183)

CISC committed 28 days ago

Verified 73159c30

Older