ggml-org/llama.cpp

Pull Requests Commits

metal : improve concurrency

ggerganov committed 15 hours ago

Verified a1ac46b4

WebUI Architecture Cleanup (#19541)

allozaur committed 16 hours ago

Verified 38adc7d4

metal : update sum_rows kernel to support float4 (#19524)

ggerganov committed 17 hours ago

Verified 3b3a9481

Add a workaround for compilation with ROCWMMA_FATTN and gfx9 (#19461)

superm1 committed 18 hours ago

Verified 6845f7f8

server : fix typo in README.md for features list (#19510)

RichardScottOZ committed 19 hours ago

Verified fa16e517

docs : update path in snapdragon README.md (#19533)

TriDefender committed 19 hours ago

Verified 313493de

hexagon: further optimization and tuning of matmul and dot kernels (#19407)

max-krasnyansky committed 20 hours ago

Verified b1ff83bb

common : replace deprecated codecvt using parse_utf8_codepoint (#19517)

angt committed 20 hours ago

Verified 4ae1b751

opencl: add general Q6_K mm and Q4_K mv (#19347)

lhez committed 1 day ago

Verified 4d3daf80

ggml : unary ops support non-cont src0 + metal F16 unary ops (#19511)

ggerganov committed 1 day ago

Verified 914dde72

common : remove unused token util functions (#19506)

danbev committed 1 day ago

Verified 3136a849

model: Add Kimi-K2.5 support (#19170)

AesSedai committed 1 day ago

Verified e463bbdf

build : fix case in dSYMs path for build-macos [no ci] (#19515)

danbev committed 1 day ago

Verified 53de59f6

metal : extend l2_norm support for non-cont src0 (#19502)

ggerganov committed 1 day ago

Verified 9ab072eb

docs: ban AI for issues and discussions [no CI] (#19512)

JohannesGaessler committed 1 day ago

Verified ada90bf2

common : improve download error reporting (#19491)

angt committed 1 day ago

Verified 0c1f39a9

hexagon: Add ARGSORT, DIV, SQR, SQRT, SUM_ROWS, GEGLU (#19406)

max-krasnyansky committed 1 day ago

Verified 73cd5e1b

llama : correct typos 'occured' and 'occurences' (#19414)

thecaptain789 committed 1 day ago

Verified 8ee538ce

model : fix wavtokenizer embedding notions (#19479)

ggerganov committed 1 day ago

Verified 6d957078

ggml : extend bin bcast for permuted src1 (#19484)

ggerganov committed 1 day ago

Verified 89181c0b

metal : consolidate unary ops (#19490)

ggerganov committed 1 day ago

Verified ceaa89b7

llama : refactor sampling_info to use buffer_view template (#19368)

danbev committed 1 day ago

Verified 2cce9fdd

CUDA : Update CCCL-tag for 3.2 to final release from RC (#19486)

ORippler committed 2 days ago

Verified 612db618

[WebGPU] Plug memory leaks and free resources on shutdown (#19315)

nikhilJain17 committed 2 days ago

Verified 57487a64

models : support qwen3.5 series (#19468)

JJJYmmm committed 2 days ago

Verified fc0fe400

test: fix IMROPE perf test case (#19465)

ngxson committed 2 days ago

Verified 9a963527

ggml-cpu: arm64: q6_K repack gemm and gemv (and generic) implementations (dotprod) (#19360)

Alcpz committed 2 days ago

Verified c03a5a46

ggml : use noexcept overload for is_regular_file in backend registration (#19452)

k4ss4n committed 2 days ago

Verified 6948adc9

convert : move experts permutation from Qwen2MoeModel to Qwen3VLMoeTextModel (#19445)

pwilkin committed 2 days ago

Verified 854b09f0

tts : fix typos in README.md [no ci] (#19463)

danbev committed 2 days ago

Verified 66d403c4

Older