ggerganov/llama.cpp

Pull Requests Commits

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

max-krasnyansky committed 60 days ago

Verified b1ff83bb

common : replace deprecated codecvt using parse_utf8_codepoint (#19517)

angt committed 60 days ago

Verified 4ae1b751

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

lhez committed 60 days ago

Verified 4d3daf80

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

ggerganov committed 61 days ago

Verified 914dde72

common : remove unused token util functions (#19506)

danbev committed 61 days ago

Verified 3136a849

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

AesSedai committed 61 days ago

Verified e463bbdf

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

danbev committed 61 days ago

Verified 53de59f6

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

ggerganov committed 61 days ago

Verified 9ab072eb

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

JohannesGaessler committed 61 days ago

Verified ada90bf2

common : improve download error reporting (#19491)

angt committed 61 days ago

Verified 0c1f39a9

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

max-krasnyansky committed 61 days ago

Verified 73cd5e1b

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

thecaptain789 committed 61 days ago

Verified 8ee538ce

model : fix wavtokenizer embedding notions (#19479)

ggerganov committed 61 days ago

Verified 6d957078

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

ggerganov committed 61 days ago

Verified 89181c0b

metal : consolidate unary ops (#19490)

ggerganov committed 61 days ago

Verified ceaa89b7

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

danbev committed 61 days ago

Verified 2cce9fdd

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

ORippler committed 61 days ago

Verified 612db618

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

nikhilJain17 committed 62 days ago

Verified 57487a64

models : support qwen3.5 series (#19468)

JJJYmmm committed 62 days ago

Verified fc0fe400

test: fix IMROPE perf test case (#19465)

ngxson committed 62 days ago

Verified 9a963527

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

Alcpz committed 62 days ago

Verified c03a5a46

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

k4ss4n committed 62 days ago

Verified 6948adc9

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

pwilkin committed 62 days ago

Verified 854b09f0

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

danbev committed 62 days ago

Verified 66d403c4

CANN: Remove unnecessary wrapper for `gml_backend_buft_is_cann` (#18968)

rauletorresc committed 62 days ago

Verified f0bfe54f

CANN: implement quantized MUL_MAT_ID for MoE models (#19228)

hipudding committed 62 days ago

Verified 52e38faf

cuda : extend GGML_OP_PAD to work with non-cont src0 (#19429)

ggerganov committed 62 days ago

Verified a0d58553

chat: fix case where template accepts type content only (#19419)

ngxson committed 62 days ago

Verified 98e57ca4

mtmd: Implement tiling for LFM2-VL (#19454)

tdakhran committed 63 days ago

Verified 262364e3

Server: log when converting requests to chat completions format (#19457)

openingnow committed 63 days ago

Verified 820ebfa6

Newer Older