ggml-org/llama.cpp

Pull Requests Commits

fit: take an optional second model into account

ServeurpersoCom committed 26 days ago

ff2e325f

server: make the draft context follow the target context

ServeurpersoCom committed 26 days ago

aed7d0ba

fit: also take into account n_streams

ngxson committed 27 days ago

2913a33e

ui: Settings navigation cleanup (#27241)

allozaur committed 27 days ago

Verified 5b6ddc96

ci : add nightly-tag.txt to make-release (#27485)

ggerganov committed 27 days ago

Verified e467c2ff

ci : release clean-up (#27477)

ggerganov committed 27 days ago

Verified 17197474

kleidiai : add SME2 F32 GEMV kernel support (#26891)

chaxu01 committed 27 days ago

Verified 62b22690

sycl : add Q2_K reordered MMVQ and ESIMD kernels (#26336)

malsbat committed 27 days ago

Verified ff14356e

test : make the FA V-is-view-of-K case a test case parameter (#27394)

ggerganov committed 27 days ago

Verified 5fff1284

sycl : Add Q5_K ESIMD kernel (#26376)

malsbat committed 27 days ago

Verified 9e89a196

opencl: keep the vocab-scale K-quant lm_head on the CPU for Adreno A7X (compiler issue workaround) (#26440)

wanghqc committed 27 days ago

Verified cd26896c

sycl: Update gate logic for Alchemist GPUs regarding OneDNN features. (#26635)

HumerousGorgon committed 27 days ago

Verified 1cb3f5eb

sycl: fix multiple warnings in compiling sycl backend (#26713)

icfaust committed 27 days ago

Verified 6602dd33

sycl : fix load model with mlock issue (#27250)

arthw committed 27 days ago

Verified 9e96cf77

TP: enable tensor split for LFM2/LFM2MOE (#26993)

cdanis committed 27 days ago

Verified b2e5e9b2

docs: fix typos in ET.md (#27457)

itsvedantkumar committed 27 days ago

Verified a298422d

ggml: support ggml_rope_set_offset on opencl, sycl, wgpu, hexagon (#27345)

ngxson committed 27 days ago

Verified 749f688f

ci: use shell script to check cmake pkg (#27414)

netrunnereve committed 27 days ago

Verified 0e1d9185

metal : clamp K extent in tensor API mat-mat kernel for K not a multiple of 32 (#27450)

ggerganov committed 27 days ago

Verified a3027337

opencl: fix q6_K flat mul_mat for Adreno A6x/A7x GPUs with older E031 compilers (#26476)

wanghqc committed 27 days ago

Verified 6503355d

opencl: fix local size for norm (#27339)

lhez committed 27 days ago

Verified 6b4fa88a

ui: Stores split refactor (#27240)

allozaur committed 28 days ago

Verified 521a64cd

mtmd: add --mmproj-device argument (#23255)

Interpause committed 28 days ago

Verified 681c29d3

model : support DSpark for LFM2 models (#27383)

tdakhran committed 28 days ago

Verified 07822bdd

vulkan: FA MMQ should use fp32 for Q quantization calculations (#27413)

jeffbolznv committed 28 days ago

Verified 78ec4c37

metal : dequant kv cache only for large batches (#27438)

ggerganov committed 28 days ago

Verified 63b64a50

CI: Use LLVM's OpenMP over MSVC_DEBUG_non_redist on Windows (#26678)

ORippler committed 28 days ago

Verified bf0040e1

server: (router) lazy-load startup_models after main setup (#27424)

ngxson committed 28 days ago

Verified 9855ad69

server : fix --docker-repo being treated as router mode (#27416)

AriBandyo committed 28 days ago

Verified 8a832e4b

CUDA: adding switch points per HW and quant type to tune the mvq->MMQ decode crossover (#26079)

praneshgo committed 28 days ago

Verified 2b562109

Older