ggerganov/llama.cpp

Pull Requests Commits

mtmd: support webp via ffmpeg

ngxson committed 6 hours ago

5471bc2c

opencl: fold the gpt-oss MoE per-expert bias adds into the epilogue (op/kernel fusion) (#26431)

wanghqc committed 7 hours ago

Verified 3af988fa

docs: improve Windows build instructions (#27381)

nikwen committed 9 hours ago

Verified 9a286ac9

ci : fix empty release_id in make-release upload step (#27516)

ggerganov committed 9 hours ago

Verified a3b9c23e

model: add dots3-note (#27060)

ngxson committed 10 hours ago

Verified 5a32f7b6

model: use ggml_rope_set_offset() (#27382)

ngxson committed 11 hours ago

Verified 873e5d8e

ci : run ccache-clear as the last step of release jobs (#27503)

ggerganov committed 12 hours ago

Verified d7fa69b7

llama.cpp : bump version to 0.2.0 (#27498)

ggerganov committed 16 hours ago

Verified bb4caa75

scripts : add release.sh for release preparation (#27497)

ggerganov committed 16 hours ago

Verified c4b0225d

ggerganov committed 16 hours ago

5de25a74

ggml : bump version to 0.21.0 (ggml/1597)

ggerganov committed 16 hours ago

01ff204f

ci : remove duplicate flag (#27488)

ggerganov committed 17 hours ago

Verified 353b32d8

Revert "sycl : add Q2_K reordered MMVQ and ESIMD kernels (#26336)" (#27486)

ggerganov committed 17 hours ago

Verified 7a0e42fd

ui: Settings navigation cleanup (#27241)

allozaur committed 18 hours ago

Verified 5b6ddc96

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

ggerganov committed 18 hours ago

Verified e467c2ff

ci : release clean-up (#27477)

ggerganov committed 19 hours ago

Verified 17197474

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

chaxu01 committed 19 hours ago

Verified 62b22690

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

malsbat committed 20 hours ago

Verified ff14356e

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

ggerganov committed 21 hours ago

Verified 5fff1284

sycl : Add Q5_K ESIMD kernel (#26376)

malsbat committed 21 hours ago

Verified 9e89a196

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

wanghqc committed 23 hours ago

Verified cd26896c

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

HumerousGorgon committed 23 hours ago

Verified 1cb3f5eb

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

icfaust committed 23 hours ago

Verified 6602dd33

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

arthw committed 23 hours ago

Verified 9e96cf77

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

cdanis committed 23 hours ago

Verified b2e5e9b2

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

itsvedantkumar committed 23 hours ago

Verified a298422d

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

ngxson committed 1 day ago

Verified 749f688f

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

netrunnereve committed 1 day ago

Verified 0e1d9185

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

ggerganov committed 1 day ago

Verified a3027337

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

wanghqc committed 1 day ago

Verified 6503355d

Older