ggml-org/llama.cpp

Pull Requests Commits

ggml-openvino : reject BF16 SCALE and mixed-type BF16 ADD/MUL/SUB

am17an committed 24 days ago

71d72e14

CUDA: use ggml_cuda_cast in binbcast and unary kernels to fix the HIP bf16 build

am17an committed 24 days ago

f575038e

ggml-cpu : use per-op _bf16 functions for BF16 unary and GLU ops

am17an committed 24 days ago

30a6e87b

ggml : add BF16 unary, GLU, binary and scale ops (CPU, CUDA)

am17an committed 24 days ago

248fecc0

CUDA: bitonic argsort handles rows wider than one block (#28957)

ServeurpersoCom committed 24 days ago

Verified 6a2743f0

ggml-cuda: HIP: optimize packed byte subtraction (`__vsubss4` -> `__vsub4`) (#29478)

thelittlefireman committed 24 days ago

Verified 748d4225

ci: add zdnn backend build but not test (#29541)

taronaeo committed 24 days ago

Verified cee37ffe

opencl: fix get_tensor for q5_K adreno gemm_nonshuffle kernel (#29555)

lhez committed 24 days ago

Verified 6dbbac44

vulkan: Tune GDN kernel, fix Intel performance (#29476)

0cc4m committed 24 days ago

Verified 5c200e0c

vulkan : Load F32 A matrix 2 at a time when its 2-aligned (#29254)

TheBlueMatt committed 24 days ago

Verified 83dd71f8

vulkan: MOE aware mat_mul_id tile selection (#29182)

Ankk98 committed 24 days ago

Verified 94a0ae3e

ggml : fix c++ odr by properly using GGML_COMMON_DECL_CPP (#29504)

XZiar committed 24 days ago

Verified da89bb3c

vocab : keep </s> NORMAL in PLaMo-2 and PLaMo-3 (#29580)

tokinasin committed 24 days ago

Verified a3f84faf

ggml : accumulate f16 dot products in f32 on AVX512-FP16 (#29545)

angt committed 24 days ago

Verified 284153e0

ggml : require input tensors to be GGML_OP_NONE (#29647)

ggerganov committed 24 days ago

Verified b5cf8ce0

hexagon: add FP32 GELU_ERF and GEGLU_ERF support (#29631)

aparmp-quic committed 24 days ago

Verified e904318a

common : stop accepting draft tokens at EOG (#29638)

ethanhq committed 24 days ago

Verified d280808f

server : remove the built-in UI's service worker when the UI is not served (#29565)

erusev committed 25 days ago

Verified ba0ba54d

common : use fs::path for config dir (#29649)

angt committed 25 days ago

Verified 00af6356

tests : adjust server string regex to also match m2 utlra results (#29648)

ggerganov committed 25 days ago

Verified c85b92c6

common : add fs_write_atomic() (#29642)

angt committed 25 days ago

Verified 31385c9c

ggml : collect all input tensors into graph_inputs (#29634)

ggerganov committed 25 days ago

Verified 8019dc56

ggml-zdnn: fix 0-row tensor crash (#29636)

taronaeo committed 25 days ago

Verified 86ea01d0

musa: build the docker image and CI container from the MUSA SDK images (#29624)

yeahdongcn committed 25 days ago

Verified 18b74ff6

ggml : speed up model loading (#29598)

angt committed 25 days ago

Verified c13e04e1

ci: remove gpu-rocm keyed directory logs (#28940)

taronaeo committed 25 days ago

Verified c8cda8b4

llama : fix init in several tools/examples (#29632)

ggerganov committed 25 days ago

Verified 6d78fb07

metal: FWHT perf optimizations (#29602)

bri-prism committed 25 days ago

Verified 18bbc46b

vulkan : reuse descriptor sets when bindings are constant (#29280)

realrasengan committed 25 days ago

Verified 0bc845d3

chat : fix Muse Glimmer ignoring response_format json_schema with --jinja (#29615)

vaibhavdedhia committed 25 days ago

Verified 139997d8

Older