ggerganov/llama.cpp

Pull Requests Commits

hparams : add comment [no ci]

ggerganov committed 175 days ago

Verified 21fe0ce4

llama : fix Gemma3 SWA KV cache shift

ggerganov committed 175 days ago

Verified de9d18fa

arg : no n_predict = -2 for examples except for main and infill (#12364)

ngxson committed 175 days ago

Verified be7c3034

llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181)

ggerganov committed 175 days ago

Verified e0dbec0b

server : fix crash when using verbose output with input tokens that are not in printable range (#12178) (#12338)

ishaangandhi committed 175 days ago

Verified 2048b591

Update build.yml for Windows Vulkan builder to use Vulkan 1.4.304 SDK for VK_NV_cooperative_matrix2 support (#12301)

oscarbg committed 176 days ago

Verified f08f4b31

llama.swiftui : fix xcframework dir in README [no ci] (#12353)

danbev committed 176 days ago

Verified 80a02aa8

sycl : variable sg_size support for mmvq kernels (#12336)

Alcpz committed 176 days ago

Verified 363f8c5d

CUDA/HIP: Fix fattn-vec-* when device warp size is not 32 (#12315)

IMbackK committed 176 days ago

Verified 34c961b1

llama : Add Gemma 3 support (+ experimental vision capability) (#12343)

ngxson committed 176 days ago

Verified 7841fc72

vulkan: fix bug in coopmat1 mul_mat_id (#12316)

jeffbolznv committed 176 days ago

Verified bf69cfe6

CUDA/HIP: refractor mmqv to unify the calculation of nwarps and rows per block between host and device code. (#12177)

IMbackK committed 177 days ago

Verified 10f2e818

ggml-backend : fix backend search path (#12330)

jklincn committed 177 days ago

Verified ba765438

metal : Cache the Metal library at the device context level (#12265)

BB-fat committed 177 days ago

Verified 6ab2e476

clip : bring back GPU support (#12322)

ngxson committed 177 days ago

Verified 96e12808

mat vec double buffer (#12188)

netrunnereve committed 178 days ago

Verified 2c9f833d

musa: support new arch mp_31 and update doc (#12296)

yeahdongcn committed 178 days ago

Verified 25136454

opencl: use OpenCL C standard supported by the device (#12221)

linehill committed 178 days ago

Verified 8acdacb3

readme: added Sidekick to available UIs (#12311)

johnbean393 committed 178 days ago

Verified 89b2b56e

tests : fix test-quantize-fns to init the CPU backend (#12306)

ggerganov committed 178 days ago

Verified e128a1bf

common : refactor '-o' option (#12278)

marcoStocchi committed 178 days ago

Verified 6ef79a67

`server`: extract <think> tags from qwq outputs (#12297)

ochafik committed 178 days ago

Verified 4e39a3c3

`tool-call`: ensure there's always a non-empty tool call id (#12292)

ochafik committed 178 days ago

Verified be421fc4

allow missing content in message if tool_calls provided (#12293)

ochafik committed 178 days ago

Verified 87c26305

`sampler`: fixes trigger tokens + lazy grammars (fix typo cast from token to string) (#12291)

ochafik committed 178 days ago

Verified 2b3a25c2

llava : fix bug in minicpm-v code (#11513)

tc-mb committed 178 days ago

Verified 8352cdc8

server : add speculative decoding presets for FIM (#12287)

ggerganov committed 179 days ago

Verified 1e2f78a0

authors : update (#12271)

ggerganov committed 180 days ago

Verified 0fd7ca7a

ggml-backend : make path_str compatible with C++20 (#12269)

ctrysbita committed 180 days ago

Verified 6fefc05a

server : infill gen ends on new line (#12254)

ggerganov committed 181 days ago

Verified 7ab36439

Older