ngxson/llama.cpp

Pull Requests Commits

update dequant fn

ngxson committed 31 days ago

e5644781

data_torch.ndim

ngxson committed 31 days ago

b246a574

convert: allow using quantized Mistral weight

ngxson committed 31 days ago

83f3004c

ggml : Provide macos-specific backtrace printing to avoid terminal death (#17869)

gabe-l-hart committed 31 days ago

Verified 02e409a5

metal : print node names for debugging (#17882)

ggerganov committed 31 days ago

Verified 6b82eb78

ggml : allow fill node alloc inplace (#17870)

CISC committed 31 days ago

Verified 86a3f0fa

cmake: fix Mach-O current version number (#17877)

Rhys-T committed 31 days ago

Verified 63908b63

model : nit, DeepSeek V1 MoE is 16B and GigaChat is 20B (#12652)

CISC committed 31 days ago

Verified 42b12b56

console: allow using arrow left/right, home/end keys and history mode (#17836)

ngxson committed 31 days ago

Verified 4e842d51

CANN: add support for partial RoPE and Vision mode (#17543)

noemotiovon committed 31 days ago

Verified ca709e42

CUDA: fix FP16 overflow in tile FA kernel (#17875)

JohannesGaessler committed 31 days ago

Verified 0cdce38a

llama : add token matching support to llama-grammar (#17816)

aldehir committed 31 days ago

Verified e39502e7

model : support Rnj-1 (#17811)

philip-essential committed 31 days ago

Verified 1d2a1ab7

graph : use fill instead of scale_bias in grouped expert selection (#17867)

CISC committed 32 days ago

Verified c8554b66

model-conversion : add token ids to prompt token output [no ci] (#17863)

danbev committed 32 days ago

Verified 2fa51c19

server: delegate result_state creation to server_task (#17835)

ngxson committed 32 days ago

Verified 951520dd

ci : support bfloat16 SYCL release package (#17855)

arthw committed 32 days ago

Verified 68522c67

server: improve speed of speculative decoding (#17808)

ngxson committed 32 days ago

Verified f896d2c3

Make graph_max_nodes vary by ubatch size (#17794)

pwilkin committed 32 days ago

Verified e4e9c432

Fix Kimi-K2 tool-call parsing issues (#17376)

hksdpc255 committed 32 days ago

Verified 636fc17a

cuda : add FILL op support (#17851)

JayZenith committed 32 days ago

Verified 51e0c2d9

server : add development documentation (#17760)

ngxson committed 32 days ago

Verified 37a4f632

server : make cache_reuse configurable per request (#17858)

ggerganov committed 32 days ago

Verified 2bc96931

cuda: optimize SOLVE_TRI using registers and FMAF (#17703)

wsbagnsv1 committed 32 days ago

Verified 5814b4dc

ggml-cpu: add ggml_thread_cpu_relax with Zihintpause support (#17784)

ixgbe committed 32 days ago

Verified 79d61896

model: add llama 4 scaling for mistral-large (deepseek arch) (#17744)

ngxson committed 33 days ago

Verified 4d372627

Vulkan: improve mul_mat_vec_iq1_m (#16907)

lovedheart committed 33 days ago

Verified 08f9d3cc

ci : add windows-cuda 13.1 release (#17839)

CISC committed 33 days ago

Verified 0a540f9a

common : change --color to accept on/off/auto, default to auto (#17827)

CISC committed 34 days ago

Verified 22577583

sycl: add missing BF16 conversion support for Intel oneAPI (#17780)

yingying0906 committed 34 days ago

Verified d9e03db1

Older