ggerganov/llama.cpp

Pull Requests Commits

metal : reduce register pressure

ggerganov committed 63 days ago

Verified 70b063a5

metal : fix typo [no ci]

ggerganov committed 64 days ago

Verified c6a1be6c

metal : reduce register pressure

ggerganov committed 64 days ago

Verified e7d14ab2

metal : mv q6_K support nr0 > 1

ggerganov committed 65 days ago

Verified fe12e20a

metal : fix nr constant [no ci]

ggerganov committed 65 days ago

Verified 51dea768

metal : fix comments [no ci]

ggerganov committed 65 days ago

Verified 982c82f1

metal : rename all_sum -> sum_all

ggerganov committed 65 days ago

Verified 24a9ea8b

metal : refactor mat-vec code

ggerganov committed 65 days ago

Verified fcca45c0

CI: fix SYCL build (#12546)

qnixsynapse committed 66 days ago

Verified 48d7021c

docs: update: improve the Fedoa CUDA guide (#12536)

teihome committed 66 days ago

Verified 3361e2de

llama-vocab : add SuperBPE pre-tokenizer (#12532)

compilade committed 66 days ago

Verified 00d53800

CUDA: Fix clang warnings (#12540)

yeahdongcn committed 66 days ago

Verified 7ea75035

mmap : skip resource limit checks on AIX (#12541)

mehendarkarprajwal committed 66 days ago

Verified c54f6b79

vulkan: fix mul_mat_vec failure in backend tests (#12529)

jeffbolznv committed 66 days ago

Verified 9b169a4d

server : Add verbose output to OAI compatible chat endpoint. (#12246)

mglambda committed 67 days ago

Verified 77f9c6bb

install : add macports (#12518)

IOOI-SqAR committed 67 days ago

Verified 18b663d8

llama : gemma3 : use output tensor if it exists in model weight (#12506)

ngxson committed 67 days ago

Verified fbdfefe7

ggml : fix quantized cpy op (#12310)

ggerganov committed 68 days ago

Verified ba932dfb

musa: refine compute capability (#12493)

yeahdongcn committed 68 days ago

Verified fac63a3d

vulkan: Optimize mul_mat_vec p021 and nc shaders (#12505)

jeffbolznv committed 68 days ago

Verified eddfb438

Vulkan: RTE rounding for cpy to quant (#12480)

stduhpf committed 68 days ago

Verified 4375415b

vulkan: workaround for AMD Windows driver 16 bit unpack8 bug (#12472)

netrunnereve committed 68 days ago

Verified 30c42ef5

model : do not repack if a GPU device is present (#12498)

ggerganov committed 69 days ago

Verified af04481e

chore : cleanup llama_model_loader::TENSOR_ usage (#12492)

CISC committed 69 days ago

Verified 960e7260

llama-tts : avoid crashes related to bad model file paths (#12482)

marcoStocchi committed 69 days ago

Verified ea1518e8

[SYCL] Fix build on Windows when ccache enabled (#9954) (#9976)

shou692199 committed 69 days ago

Verified 1aa87ee5

sycl: cleanup oneDNN related code (#12097)

sgeor255 committed 69 days ago

Verified 9ffcc9e3

webui : Prevent rerendering on textarea input (#12299)

woof-dog committed 70 days ago

Verified e0464306

llama : make Qwen2MoE QKV bias optional (#12477)

CISC committed 70 days ago

Verified dbb3a473

ggml : block interleaving support for Q4_K quantization for x86 AVX2 architecture (#12332)

Srihari-mcw committed 70 days ago

Verified 3d82dbcb

Older