ngxson/llama.cpp

Pull Requests Commits

move transpose to outside of loop

ngxson committed 114 days ago

d871ac85

model: (qwen3next) correct vectorized key_gdiff calculation

ngxson committed 114 days ago

4bfbf0b7

metal : add solve_tri (#19302)

ggerganov committed 115 days ago

Verified 44008ce8

ci : add sanitizer runs for server (#19291)

ggerganov committed 115 days ago

Verified 6a9bf2f7

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

ggerganov committed 115 days ago

Verified faa1bc26

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

0cc4m committed 115 days ago

Verified 32b17abd

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

am17an committed 115 days ago

Verified 8bece2eb

models : remove unnecessary cont in openelm (#19289)

CISC committed 115 days ago

Verified a6fd8ca1

metal : minor cleanup (#19251)

ggerganov committed 115 days ago

Verified c55bce41

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

ORippler committed 115 days ago

Verified 1f1e57f2

ggml: added cleanups in ggml_quantize_free (#19278)

noctrex committed 115 days ago

Verified e9a859db

cuda : revert CUDA_SCALE_LAUNCH_QUEUES override until investigated (#19227)

gaugarg-nv committed 115 days ago

Verified 41e3f026

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

vhsw committed 115 days ago

Verified 1efb5f7a

spec : simplify time measurement using common_time_meas (#19262)

ggerganov committed 115 days ago

Verified aeb827a3

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

lhez committed 116 days ago

Verified 91ea44e8

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

sidmohan0 committed 116 days ago

Verified 0dfcd3b6

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

ngxson committed 116 days ago

Verified 07a7412a

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

am17an committed 116 days ago

Verified 9f682fb6

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

teto committed 116 days ago

Verified a3fa0358

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

am17an committed 116 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 116 days ago

Verified bf38346d

sycl: implement GGML_OP_TOP_K (#19242)

tdevelope committed 116 days ago

Verified 4d5e9726

metal : support virtual devices (#18919)

ggerganov committed 116 days ago

Verified 6fdddb49

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

danbev committed 116 days ago

Verified 6156ae51

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

JohannesGaessler committed 116 days ago

Verified 59377a6c

authors : update (#19263)

ggerganov committed 116 days ago

Verified 1239267c

docs : Minor cleanups (#19252)

ckastner committed 116 days ago

Verified 7a4ca3cb

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

srogmann committed 116 days ago

Verified b4d05a3d

Remove pipeline cache mutexes (#19195)

nikhilJain17 committed 116 days ago

Verified 2dc3ce21

Bump cmake max version (needed for Windows on Snapdragon builds) (#19188)

max-krasnyansky committed 117 days ago

Verified 3bc8d2cf

Older