ngxson/llama.cpp

Pull Requests Commits

cuda : add F16 kernel support for CONV_2D_DW (#29064)

crion99 committed 1 hour ago

Verified a72e04ab

test: flush status (#28352)

kurquhar committed 2 hours ago

Verified 8212c780

ui : fix missing svg use and animation elements in preview and download (#28962)

nandan2003 committed 2 hours ago

Verified 945064fc

llama: add llama_batch_ext (#24669)

ngxson committed 3 hours ago

Verified fc343a84

server : change default pytest workers to 4 (#29376)

danbev committed 3 hours ago

Verified 308883b3

ci : use hf-jobs-cpu-performance, disable pytest workers (#29369)

ggerganov committed 4 hours ago

Verified 70596c4d

vulkan: int8 coopmat1 matmul implementation for AMD RDNA3 and RDNA4 (#27952)

0cc4m committed 4 hours ago

Verified 70c4e158

vulkan: handle misalignment in conv_2d and conv_3d (#29365)

0cc4m committed 5 hours ago

Verified 6b790a9c

vulkan: tune KHR cooperative matrix support for Adreno GPUs (#29328)

Raman-Raje committed 6 hours ago

Verified 3423f940

cuda : add conv3d with implicit GEMM (#29137)

leejet committed 10 hours ago

Verified 53ed051c

model : add Ling 3.0 VL support (#29151)

aetherbird committed 10 hours ago

Verified f830688e

server,common : fix the GCC 12 stringop-overread false positive (again) (#29325)

angt committed 11 hours ago

Verified 2b705839

test-save-load-state : print a per-model results table in --models mode (#29316)

ggerganov committed 11 hours ago

Verified 4c5957c2

hexagon: reject MUL_MAT_ID when src1 precision is F32 (#29348)

jhen0409 committed 12 hours ago

Verified 9710a321

scripts : make-release-desc - link previous release in changelog title (#29336)

ggerganov committed 13 hours ago

Verified 013b31c0

convert : allow vision target for DFlash/Dspark (#29339)

tdakhran committed 18 hours ago

Verified bd4f514d

server: allow preset to set log file (#29334)

ngxson committed 18 hours ago

Verified b9ae43a5

tests: add `-b/--backend` option to test-llama-archs for testing a specific backend (#27372)

yomaytk committed 21 hours ago

Verified d2e54583

ci : use hf-jobs-cpu-xl runner in server sanitize workflow (#29297)

ggerganov committed 23 hours ago

Verified 6e60f356

opencl: add A8 Q6_K non-MoE dp4a binary kernel (#29057)

shaofeiqi committed 1 day ago

Verified fee39dd9

llama.cpp : bump version to 0.5.0 (#29333)

ggerganov committed 1 day ago

Verified 7fe450e1

ggerganov committed 1 day ago

177cd8cc

ggml : bump version to 0.25.1 (ggml/1637)

ggerganov committed 1 day ago

e4e2f623

CUDA: add a reserve to avoid spurious warning on older GCC builds (#29317)

am17an committed 1 day ago

Verified 66fba63a

common : keep HF cache dir as path, expose UTF-8 only for logs (#29320)

angt committed 1 day ago

Verified bddf8263

metal: add the missing f32 x bf16 mul_mv variants (#28741)

ServeurpersoCom committed 1 day ago

Verified 95753896

CUDA: enable sparse-fa for dsv4 prefill (again) (#29298)

am17an committed 1 day ago

Verified dc9879cf

server: fix token counting API crash on sleep (#29309)

willweimike committed 1 day ago

Verified 42916d83

jinja : parse unary +/- before variables (#29244)

cs-fisha committed 1 day ago

Verified 4e416ee7

metal : key the fa-vec tuned table by family instead of SKU (#29075)

forforever73 committed 1 day ago

Verified ee3ecce0

Older