ggerganov/llama.cpp

Pull Requests Commits

qwen3next : fix chunking

ggerganov committed 20 days ago

Verified 1213a035

completion : simplify batch (embd) processing (#19286)

danbev committed 20 days ago

Verified 25f40ca6

ggml-virtgpu: make the code thread safe (#19204)

kpouget committed 21 days ago

Verified 015deb90

ggml-cpu: use LUT for converting e8->f32 scales on x86 (#19288)

am17an committed 21 days ago

Verified 2ceda3f6

metal : add solve_tri (#19302)

ggerganov committed 21 days ago

Verified 44008ce8

ci : add sanitizer runs for server (#19291)

ggerganov committed 21 days ago

Verified 6a9bf2f7

sampling : delegate input allocation to the scheduler (#19266)

ggerganov committed 21 days ago

Verified faa1bc26

vulkan: disable coopmat1 fa on Nvidia Turing (#19290)

0cc4m committed 21 days ago

Verified 32b17abd

CUDA: use mmvq for mul-mat-id for small batch sizes (#18958)

am17an committed 21 days ago

Verified 8bece2eb

models : remove unnecessary cont in openelm (#19289)

CISC committed 21 days ago

Verified a6fd8ca1

metal : minor cleanup (#19251)

ggerganov committed 21 days ago

Verified c55bce41

CUDA: Fix loop unrolling for BW in mul_mat_q_stream_k_fixup (#19053)

ORippler committed 21 days ago

Verified 1f1e57f2

ggml: added cleanups in ggml_quantize_free (#19278)

noctrex committed 21 days ago

Verified e9a859db

cuda : revert CUDA_SCALE_LAUNCH_QUEUES override until investigated (#19227)

gaugarg-nv committed 21 days ago

Verified 41e3f026

vocab: add Falcon-H1-Tiny-Coder FIM tokens (#19249)

vhsw committed 21 days ago

Verified 1efb5f7a

spec : simplify time measurement using common_time_meas (#19262)

ggerganov committed 21 days ago

Verified aeb827a3

opencl: refactor some ops, concat, repeat, tanh and scale (#19226)

lhez committed 22 days ago

Verified 91ea44e8

jinja : add missing 'in' test to template engine (#19004) (#19239)

sidmohan0 committed 22 days ago

Verified 0dfcd3b6

mtmd: add min/max pixels gguf metadata (#19273)

ngxson committed 22 days ago

Verified 07a7412a

ggml-cpu: FA split across kv for faster TG (#19209)

am17an committed 22 days ago

Verified 9f682fb6

server: print actual model name in 'model not found" error (#19117)

teto committed 22 days ago

Verified a3fa0358

ci: add test-backend-ops test for CPU (#19268)

am17an committed 22 days ago

Verified 15818ac4

Remove support for Nvidia & AMD GPU, because the oneAPI plugin for Nvidia & AMD GPU is unavailable: download/installation channels are out of work. (#19246)

arthw committed 22 days ago

Verified bf38346d

sycl: implement GGML_OP_TOP_K (#19242)

tdevelope committed 22 days ago

Verified 4d5e9726

metal : support virtual devices (#18919)

ggerganov committed 22 days ago

Verified 6fdddb49

model-conversion : add debug option to conversion script (#19265)

danbev committed 22 days ago

Verified 6156ae51

ggml-backend: fix async set/get fallback sync (#19179)

JohannesGaessler committed 22 days ago

Verified 59377a6c

authors : update (#19263)

ggerganov committed 22 days ago

Verified 1239267c

docs : Minor cleanups (#19252)

ckastner committed 22 days ago

Verified 7a4ca3cb

spec : various improvements ton ngram-map + docs (#19253)

srogmann committed 22 days ago

Verified b4d05a3d

Older