ggerganov/llama.cpp

Pull Requests Commits

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

Alcpz committed 1 hour ago

Verified c03a5a46

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

k4ss4n committed 2 hours ago

Verified 6948adc9

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

pwilkin committed 4 hours ago

Verified 854b09f0

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

danbev committed 6 hours ago

Verified 66d403c4

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

rauletorresc committed 6 hours ago

Verified f0bfe54f

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

hipudding committed 6 hours ago

Verified 52e38faf

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

ggerganov committed 6 hours ago

Verified a0d58553

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

ngxson committed 15 hours ago

Verified 98e57ca4

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

tdakhran committed 20 hours ago

Verified 262364e3

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

openingnow committed 21 hours ago

Verified 820ebfa6

spec : remove check rate (#19377)

srogmann committed 23 hours ago

Verified 292f6908

ci : add metal server workflows (#19293)

ggerganov committed 23 hours ago

Verified 81ddc60c

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

ggerganov committed 23 hours 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 1 day ago

Verified 1e8924fd

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

pwilkin committed 1 day ago

Verified 39bf692a

CUDA: Fix non-contig rope (#19338)

ORippler committed 1 day ago

Verified e06088da

rpc : update from common.cpp (#19400)

angt committed 2 days ago

Verified 5fa1c190

server : improve context checkpoint logic (#19408)

ggerganov committed 2 days ago

Verified eb449cdf

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

ddh0 committed 2 days ago

Verified 5999b50e

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

CISC committed 2 days ago

Verified 9a5f5779

ci : use -j param correctly when building with sanitizers (#19411)

ggerganov committed 2 days ago

Verified 96441c95

metal : consolidate bin kernels (#19390)

ggerganov committed 3 days ago

Verified 8872ad21

metal : fix event synchronization in cpy_tensor_async (#19402)

ggerganov committed 3 days ago

Verified 34ba7b5a

model : support Step3.5-Flash (#19283)

forforever73 committed 3 days ago

Verified b8311181

gguf-py : bump sentencepiece version (#19319)

Ahajha committed 3 days ago

Verified 3228e772

ggml-webgpu: JIT compile binary operators and handle binding overlaps (#19310)

abhijitramesh committed 3 days ago

Verified 7fbd36c5

sycl: add F16 support for GGML_OP_CEIL (#19306)

NechamaKrashinski committed 3 days ago

Verified 537eadb1

tests: reduce number of FA test permutations (#19381)

jeffbolznv committed 3 days ago

Verified db6adb3c

common : add common_speculative_is_compat() (#19270)

ggerganov committed 3 days ago

Verified dfde5993

Older