ggml-org/llama.cpp

Pull Requests Commits

server: create checkpoint on task cancel

ngxson committed 41 days ago

24c307d2

mtmd-debug: add color and rainbow mode (#23829)

ngxson committed 41 days ago

Verified 751ebd17

mtmd: fix gemma 4 projector pre_norm (#23822)

ngxson committed 41 days ago

Verified c8914ad4

opencl: move backend info printing into its own function (#23702)

lhez committed 41 days ago

Verified 408ae2b9

ci : run ui publish on ubuntu-slim (#23818)

CISC committed 41 days ago

Verified 3ef23695

ui: fix audio and video modality detection (#23756)

ValdikSS committed 41 days ago

Verified 2f6c815d

ci : releases use Github-hosted builds for the UI (#23823)

ggerganov committed 41 days ago

Verified 445b7cef

app : improve help output (#23805)

angt committed 41 days ago

Verified 479a9a1b

mtmd: n_head_kv defaults to n_head (#23782)

sfallah committed 41 days ago

Verified 0b56d283

mtmd: fix gemma 4 audio rms norm eps (#23815)

ngxson committed 41 days ago

Verified d6be3158

ci : change Vulkan builds to Release to reduce ccache (#23820)

ggerganov committed 41 days ago

Verified dd155790

arg: Add LLAMA_ARG_API_KEY_FILE environment variable for --api-key-file (#23167)

kucharskim committed 41 days ago

Verified 7fb1e70b

test-llama-archs: fix table format [no release] (#23810)

JohannesGaessler committed 41 days ago

Verified d374e71e

ggml: auto apply iGPU flag CUDA/HIP if integrated device (#23007)

fl0rianr committed 41 days ago

Verified 30af6e2b

mmvq Optim: add MMVQ_PARAMETERS_TURING(mmvq_parameter_table_id) for … (#23729)

yaohengxu committed 41 days ago

Verified d7be4618

CUDA: route batch>=4 quantized matmul to MMQ on AMD MFMA hardware (#23227)

jadenmach2 committed 41 days ago

Verified bc81d47a

server: minor tweaks to use more cpp features (#23785)

mfuntowicz committed 41 days ago

Verified 0b246862

hexagon: minor refresh for HMX FA and MM (#23796)

max-krasnyansky committed 41 days ago

Verified a9190011

vulkan: fast path for walsh-hadamard transform (#23687)

jeffbolznv committed 41 days ago

Verified 48e7078e

chat : add Granite 4.1 chat template (#23518)

jesus-talavera-ibm committed 41 days ago

Verified bb771cbd

vulkan: fix wrong index variable in inner loop (#23665)

winstonma committed 42 days ago

Verified 7c48fb81

vulkan: Fix memory logger unsafe iterator access (#23667)

winstonma committed 42 days ago

Verified 91eb8f4f

server, ui : Add support for HTTP ETags in llama-server (#23701)

mtavenrath committed 42 days ago

Verified d205df68

docker : add ZenDNN Dockerfile (#23716)

z-sachin committed 42 days ago

Verified e8d25674

cuda : fix KQ mask offset integer overflow in fattn MMA kernel (#23610)

fairydreaming committed 42 days ago

Verified 09e7b76c

perplexity : fix format specifier in LOG_ERR (#23788)

angt committed 42 days ago

Verified 48e7eae4

convert : add FP8 to Q8 conversion (#23250)

ynankani committed 42 days ago

Verified c5229087

ggml: fixed Arm SVE usage bug in vec.h, vec.cpp (#22841)

martin-klacer-arm committed 42 days ago

Verified e31cdaa0

ci : refactor (#23789)

ggerganov committed 42 days ago

Verified 491c4d7d

Hexagon: OP_GATED_DELTA_NET K>1 support (#23531)

ymcki committed 42 days ago

Verified 939a7dd6

Older