ggerganov/llama.cpp

Pull Requests Commits

ggml: aarch64: Implement SVE in Gemm q4_k 8x8 q8_k Kernel (#19132)

abhijain1204fujitsu committed 47 days ago

Verified 267ba5a1

ggerganov committed 47 days ago

ff4affb4

ggml : bump version to 0.9.7 (ggml/1425)

ggerganov committed 47 days ago

55d58599

ggml : bump version to 0.9.6 (ggml/1423)

ggerganov committed 47 days ago

1a8c700b

cuda: optimize iq2xxs/iq2xs/iq3xxs dequantization (#19624)

dfriehs committed 48 days ago

Verified 27b93cbd

docs: update s390x build docs (#19643)

taronaeo committed 48 days ago

Verified 6e67fd21

build : remove LLAMA_HTTPLIB option (#19623)

angt committed 48 days ago

Verified 9e118b97

cmake : check if KleidiAI API has been fetched (#19640)

danbev committed 48 days ago

Verified 57088276

context : fix output reorder with backend sampling (#19638)

ggerganov committed 48 days ago

Verified 341bc7d2

ggml : avoid UB in gemm ukernel (#19642)

ggerganov committed 48 days ago

Verified 08e6d914

ggml-cpu: optimize ggml_vec_dot_bf16 for s390x (#19399)

taronaeo committed 48 days ago

Verified 184c694f

ggml-cpu: FA add GEMM microkernel (#19422)

am17an committed 48 days ago

Verified 684b3610

cmake : fix KleidiAI install target failure with EXCLUDE_FROM_ALL (#19581)

ssam18 committed 48 days ago

Verified 3a00c985

convert : ensure all models handle new experts count (#19621)

CISC committed 48 days ago

Verified 079feab9

mtmd : Add Nemotron Nano 12B v2 VL support (#19547)

anavp-nvidia committed 49 days ago

Verified 01d8eaa2

models : optimize qwen3next graph (#19375)

ggerganov committed 49 days ago

Verified 1725e316

ggml : fix GGML_DEBUG with OpenMP (#19599)

angt committed 49 days ago

Verified b7742cf3

NetBSD build support (#19589)

iMilnb committed 49 days ago

Verified badba893

webui: Architecture and UI improvements (#19596)

allozaur committed 49 days ago

Verified baa12f38

llama : update LoRA API. + fix excessive graph reserves (#19280)

agent-enemy-2 committed 49 days ago

Verified 2d8015e8

mmap: Fix Windows handle lifetime (#19598)

noctrex committed 49 days ago

Verified eb145c07

metal : fix ACC op (#19427)

ggerganov committed 49 days ago

Verified 6e473fb3

scripts : use official split.py for cpp-httplib (#19588)

angt committed 49 days ago

Verified c7db95f1

convert : store ffn_gate_inp_shexp as F32 (#19606)

CISC committed 49 days ago

Verified 0d00ef65

build : fix libtool call in build-xcframework.sh (#19605)

angt committed 49 days ago

Verified 91ea5d67

vulkan: support L2_NORM with contiguous rows (#19604)

jeffbolznv committed 49 days ago

Verified dbb02333

vulkan: support GGML_OP_SET (#19584)

jeffbolznv committed 49 days ago

Verified 53aef25a

vulkan: Add vendor id for Qualcomm drivers (#19569)

strongtz committed 49 days ago

Verified 2dec5480

hexagon: further optimizations and refactoring for flash attention (#19583)

max-krasnyansky committed 49 days ago

Verified 0ccbfdef

github : add missing backends to issue templates (#19603)

mengshengwu committed 49 days ago

Verified 94a602db

Newer Older