ggml-org/llama.cpp

Pull Requests Commits

Update tools/mtmd/clip.cpp

ngxson committed 5 hours ago

Verified c807c2ce

mtmd: fix gemma 4 audio rms norm eps

ngxson committed 5 hours ago

2528c658

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

jeffbolznv committed 5 hours ago

Verified 48e7078e

chat : add Granite 4.1 chat template (#23518)

jesus-talavera-ibm committed 6 hours ago

Verified bb771cbd

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

winstonma committed 6 hours ago

Verified 7c48fb81

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

winstonma committed 6 hours ago

Verified 91eb8f4f

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

mtavenrath committed 6 hours ago

Verified d205df68

docker : add ZenDNN Dockerfile (#23716)

z-sachin committed 7 hours ago

Verified e8d25674

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

fairydreaming committed 8 hours ago

Verified 09e7b76c

perplexity : fix format specifier in LOG_ERR (#23788)

angt committed 9 hours ago

Verified 48e7eae4

convert : add FP8 to Q8 conversion (#23250)

ynankani committed 10 hours ago

Verified c5229087

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

martin-klacer-arm committed 10 hours ago

Verified e31cdaa0

ci : refactor (#23789)

ggerganov committed 10 hours ago

Verified 491c4d7d

Hexagon: OP_GATED_DELTA_NET K>1 support (#23531)

ymcki committed 11 hours ago

Verified 939a7dd6

opencl: OP_GATED_DELTA_NET (#23312)

ymcki committed 12 hours ago

Verified 8ad8aef4

ggml-webgpu: remove legacy constants (#23672)

reeselevine committed 19 hours ago

Verified f12cc6d0

hexagon: add support for Q4_1 in MUL_MAT and MUL_MAT_ID (#23647)

max-krasnyansky committed 23 hours ago

Verified aa50b2c2

ggml-webgpu: Fix how to dispatch WG to some ops (#23750)

yomaytk committed 1 day ago

Verified c40006a6

vulkan: Switch MUL_MAT_VEC to 4 K per iteration for F16/32 (#22887)

TheBlueMatt committed 1 day ago

Verified c6e40883

vulkan: use GL_NV_cooperative_matrix_decode_vector for faster matmul (#23541)

jeffbolznv committed 1 day ago

Verified b36eefc1

vulkan: add REPEAT op support for f16 to f16. (#23298)

l8bloom committed 1 day ago

Verified 837bb6b4

ci : move ARM jobs to self-hosted + disable kleidiai mac release (#23780)

ggerganov committed 1 day ago

Verified ba4dd0bc

vendor : update cpp-httplib to 0.46.0 (#23650)

cabelo committed 1 day ago

Verified 617255d4

pyproject : add conversion folder and update dependencies (#23746)

CISC committed 1 day ago

Verified 87b0a60c

CUDA: restrict PDL to CTK >= 12.3 due to MSVC issues (#23742)

ORippler committed 1 day ago

Verified fda8528a

ci : bump cuda release to 13.3 (#23749)

CISC committed 1 day ago

Verified 2d0656fb

common : fix env names to all have LLAMA_ARG_ prefix (#23778)

ggerganov committed 1 day ago

Verified 6b4e4bd5

ci : fix windows ccaches (#23777)

ggerganov committed 1 day ago

Verified 9f0e4b14

ci : remove wasm test (#23733)

CISC committed 1 day ago

Verified b3a739c9

vulkan: avoid preferring transfer queue on AMD UMA devices (#22455)

winstonma committed 1 day ago

Verified 4d8cc0c5

Older