ggerganov/llama.cpp

Pull Requests Commits

metal : extend l2_norm support for non-cont src0

ggerganov committed 2 hours ago

Verified 1f1cabc3

model : fix wavtokenizer embedding notions (#19479)

ggerganov committed 2 hours ago

Verified 6d957078

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

ggerganov committed 2 hours ago

Verified 89181c0b

metal : consolidate unary ops (#19490)

ggerganov committed 2 hours ago

Verified ceaa89b7

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

danbev committed 3 hours ago

Verified 2cce9fdd

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

ORippler committed 10 hours ago

Verified 612db618

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

nikhilJain17 committed 16 hours ago

Verified 57487a64

models : support qwen3.5 series (#19468)

JJJYmmm committed 16 hours ago

Verified fc0fe400

test: fix IMROPE perf test case (#19465)

ngxson committed 18 hours ago

Verified 9a963527

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

Alcpz committed 21 hours ago

Verified c03a5a46

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

k4ss4n committed 22 hours ago

Verified 6948adc9

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

pwilkin committed 1 day ago

Verified 854b09f0

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

danbev committed 1 day ago

Verified 66d403c4

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

rauletorresc committed 1 day ago

Verified f0bfe54f

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

hipudding committed 1 day ago

Verified 52e38faf

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

ggerganov committed 1 day ago

Verified a0d58553

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

ngxson committed 1 day ago

Verified 98e57ca4

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

tdakhran committed 1 day ago

Verified 262364e3

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

openingnow committed 1 day ago

Verified 820ebfa6

spec : remove check rate (#19377)

srogmann committed 1 day ago

Verified 292f6908

ci : add metal server workflows (#19293)

ggerganov committed 1 day ago

Verified 81ddc60c

revert : "[Model] Qwen3.5 dense and MoE support (no vision) (#19435)" (#19453)

ggerganov committed 1 day ago

Verified 972f323e

ggml-virtgpu: add backend documentation (#19354)

kpouget committed 1 day ago

Verified f5e7734f

cmake : add variable to skip installing tests (#19370)

WhyNotHugo committed 2 days ago

Verified 1e8924fd

[Model] Qwen3.5 dense and MoE support (no vision) (#19435)

pwilkin committed 2 days ago

Verified 39bf692a

CUDA: Fix non-contig rope (#19338)

ORippler committed 2 days ago

Verified e06088da

rpc : update from common.cpp (#19400)

angt committed 3 days ago

Verified 5fa1c190

server : improve context checkpoint logic (#19408)

ggerganov committed 3 days ago

Verified eb449cdf

llama-quantize : cleanup `--help` output (#19317)

ddh0 committed 3 days ago

Verified 5999b50e

ci : remove server job from webui and move slow test (#19424)

CISC committed 3 days ago

Verified 9a5f5779

Older