ggml-org/llama.cpp

Pull Requests Commits

ci: fix sanitizer tests (#28583)

netrunnereve committed 4 days ago

Verified 434ddbbc

vulkan: use spec constant for matrix matrix multiplication A-type (#25773)

0cc4m committed 4 days ago

Verified 91f6a6cf

hexagon: rope updates (#28628)

tboinovski1 committed 4 days ago

Verified 6d9c82ea

vulkan: Convert FILL to distribute workgroups in 2D to avoid exceeding maxComputeWorkGroupCount (#28592)

rillomas committed 4 days ago

Verified 22397c31

py : lower numpy to 2.2.6 (#28654)

CISC committed 4 days ago

Verified 4b98ab80

py : bump numpy to 2.4.6 (#28649)

CISC committed 5 days ago

Verified 9cf3bf25

llama : use int32_t for llama_sampler_chain_n return type (#28631)

linsen458-spec committed 5 days ago

Verified 4850c772

CUDA: size routed MoE MMQ N-tiles from typical expert width on RDNA3 (#28552)

pwilkin committed 5 days ago

Verified d4abd573

CUDA: replace GGML_FA_ALL_QUANTS with GGML_FA_QUANTS, more control over what is compiled (#28079)

pwilkin committed 5 days ago

Verified 5a4d0fec

args: officially deprecate --mmap|mlock|dio (#28334)

taronaeo committed 5 days ago

Verified 14a9d09f

model: fix granite3 moe unknown parameter count (#28632)

taronaeo committed 5 days ago

Verified e2d2c0d6

mtmd: propagate video ID to bitmap (#28601)

ngxson committed 5 days ago

Verified 6de9cdb2

jinja: treat a null left operand of in as a plain lookup (#28620)

ServeurpersoCom committed 5 days ago

Verified b31b71f3

vulkan: add dedicated iq4_xs mat-vec shader (#28426)

WakeUpMorty committed 5 days ago

Verified df750f76

vulkan: add f16 B-type matmul pipelines and warp tile size tuning for Intel coopmat1 (#27471)

fish-jiang committed 5 days ago

Verified 1945e092

tests : use less threads for data initialization (#28325)

ggerganov committed 5 days ago

Verified 30b6a755

Add IQ type handling for MoE (#28476)

cwriter committed 5 days ago

Verified 304665fe

llama: disable lazy tensor loading by default on iGPUs (#28326)

0cc4m committed 5 days ago

Verified f3f1a8f2

ggml : fix msvc+clang ggml_vld1q_u32 (#28284)

sarahwu185 committed 5 days ago

Verified 9113cc18

Revert "ggml-cuda : restore prop.integrated on HIP builds (#24233)" (#28604)

IMbackK committed 5 days ago

Verified d4389a4d

server : apply checkpoint min-step eviction only when the checkpoint list is full (#28302)

masterFoad committed 6 days ago

Verified 5d806aa2

metal : fix idle threads in mul_mv_iq3_xxs for ne00 < 1024 (#28086)

masterFoad committed 6 days ago

Verified 88ada91c

spec: single device drafter should create meta backend wrapper (#28390)

am17an committed 6 days ago

Verified 415e909d

ci : disable npm gha cache (#28600)

CISC committed 6 days ago

Verified 03fa73cb

ci : add PYTEST_WORKERS=1 to fix server-self-hosted job (#28603)

danbev committed 6 days ago

Verified 1744c6bd

llama : add missing headers (#28566)

peppergrayxyz committed 6 days ago

Verified ca86fb22

vulkan : fuse UNARY(GELU|SIGMOID|SILU|SOFTPLUS) + MUL (#27220)

Ankk98 committed 6 days ago

Verified 64e9bceb

Fix Vulkan-Hpp handle usage on 32-bit targets. (#22892)

miyanyan committed 6 days ago

Verified f014bfef

chat : split specialized parsers into common/parsers (#27764)

pwilkin committed 6 days ago

Verified 895c045f

opencl: properly handle non-contiguous inputs to conv2d (#28503)

lhez committed 6 days ago

Verified 7d701b59

Newer Older