ggml-org/llama.cpp

Pull Requests Commits

server: (router) do not evict busy models

ngxson committed 4 hours ago

1621aade

mtmd: stop feeding the text stream again during Qwen3-TTS generation (#26706)

ServeurpersoCom committed 5 hours ago

Verified 217df17a

ggml : add aarch64 HWCAP fallbacks and fix fp16 variant detection (#25554)

kilian-hu committed 6 hours ago

Verified cb26014d

ui: read model modalities from the router model list (#26709)

ServeurpersoCom committed 7 hours ago

Verified 82bb4850

Mitigate crashing issue on Windows MSYS2 UCRT64 environment (GCC 16.1.0) (#26555)

rillomas committed 8 hours ago

Verified 42e98813

sycl: fix UE4M3 parsing (#25608)

clee committed 11 hours ago

Verified fc3f10b3

sycl: *glu flat path (#26354)

Titaniumtown committed 11 hours ago

Verified 6b5c2efb

sycl : Support DSv4 OPs: LIGHTNING_INDEXER,DSV4_HC_COMB,DSV4_HC_POST,DSV4_HC_PRE (#26568)

arthw committed 11 hours ago

Verified 31558dbb

sycl : update guide Q&A and script for device setting (#26442)

arthw committed 12 hours ago

Verified c1f41098

sycl : fix error Error OP FLASH_ATTN_EXT on arc770 (#26441)

arthw committed 12 hours ago

Verified eef5f3e3

sycl : enhance OP set_rows to support all missed data types (#26515)

arthw committed 12 hours ago

Verified c074cb3f

cuda: fix warnings for unused variable/function (#26688)

dfriehs committed 12 hours ago

Verified 5b87ed30

ci: abort if build requirements are missing (#26368)

nikwen committed 12 hours ago

Verified d8d98872

metal : avoid `threadgroup` matrix array instantiation in kernel_lightning_indexer (#26646)

JamePeng committed 12 hours ago

Verified e40bf886

mtmd: add chunk save/load function (#26645)

ngxson committed 23 hours ago

Verified 15586e2d

server: fix empty response for /cors-proxy (#26656)

ngxson committed 1 day ago

Verified 6a32c29a

convert : fix DeepseekV4 rope parameters with transformers 5.x (#26673)

CISC committed 1 day ago

Verified eb5667a1

model-loader : fix quantized reshaped tensor strides (#26672)

ggerganov committed 1 day ago

Verified 3db4ff87

convert : accept "ExaoneMoeForCausalLM" arch spelling (#26660)

csabakecskemeti committed 1 day ago

Verified e700bfb3

ci : onboard AMD ROCm CI with gfx1151 fixes (#26544)

jimw567 committed 1 day ago

Verified a1f96d4f

model-conversion : add --model-name to conversion scripts (#26665)

danbev committed 1 day ago

Verified 9de0fcf2

vulkan: fix submission batching size, add debug tools for diagnosing causes of DeviceLost drivers errors (#26371)

0cc4m committed 1 day ago

Verified 803b7fca

mtmd/ggml: add ggml_build_forward_order (#26649)

ServeurpersoCom committed 1 day ago

Verified c8e03ce8

server: harden the file_glob_search directory walk (#26626)

ServeurpersoCom committed 1 day ago

Verified f9e832c1

tests: re-enable MiniMax M3 in `test-llama-archs` (#26633)

nikwen committed 2 days ago

Verified 360e1349

mtmd: Unlimited-OCR fix max_tiles, setting in converter (#25614)

sfallah committed 2 days ago

Verified b06aa774

grammar : degrade max repetition >= 2000 to unbounded (#26613)

aldehir committed 2 days ago

Verified cd0fa605

mtmd: support multi-row batching for deepseek-ocr (#26154)

ngxson committed 2 days ago

Verified 717dad5c

fit: Fix memory allocation for MTP layers (#26605)

smalinin committed 2 days ago

Verified 9a688e51

security : clarify about AI-generated reports (#26579)

ngxson committed 2 days ago

Verified 9303cdd8

Older