ggerganov/llama.cpp

Pull Requests Commits

ggml : remove redundant src in ggml_cast

ggerganov committed 21 days ago

Verified 2a615b27

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

JohannesGaessler committed 21 days ago

Verified 0cdce38a

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

aldehir committed 21 days ago

Verified e39502e7

model : support Rnj-1 (#17811)

philip-essential committed 21 days ago

Verified 1d2a1ab7

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

CISC committed 21 days ago

Verified c8554b66

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

danbev committed 22 days ago

Verified 2fa51c19

server: delegate result_state creation to server_task (#17835)

ngxson committed 22 days ago

Verified 951520dd

ci : support bfloat16 SYCL release package (#17855)

arthw committed 22 days ago

Verified 68522c67

server: improve speed of speculative decoding (#17808)

ngxson committed 22 days ago

Verified f896d2c3

Make graph_max_nodes vary by ubatch size (#17794)

pwilkin committed 22 days ago

Verified e4e9c432

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

hksdpc255 committed 22 days ago

Verified 636fc17a

cuda : add FILL op support (#17851)

JayZenith committed 22 days ago

Verified 51e0c2d9

server : add development documentation (#17760)

ngxson committed 22 days ago

Verified 37a4f632

server : make cache_reuse configurable per request (#17858)

ggerganov committed 22 days ago

Verified 2bc96931

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

wsbagnsv1 committed 22 days ago

Verified 5814b4dc

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

ixgbe committed 22 days ago

Verified 79d61896

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

ngxson committed 22 days ago

Verified 4d372627

Vulkan: improve mul_mat_vec_iq1_m (#16907)

lovedheart committed 23 days ago

Verified 08f9d3cc

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

CISC committed 23 days ago

Verified 0a540f9a

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

CISC committed 23 days ago

Verified 22577583

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

yingying0906 committed 23 days ago

Verified d9e03db1

vulkan: perf_logger improvements (#17672)

jeffbolznv committed 24 days ago

Verified db978373

ggml-zendnn : add ZenDNN backend for AMD CPUs (#17690)

z-vishal committed 24 days ago

Verified 017761da

server: support multiple generations from one prompt (OAI "n" option) (#17775)

ngxson committed 24 days ago

Verified c42712b0

ggml : add circular tiling support to pad, for Vulkan, CUDA, and CPU (used for making seamless textures) (#16985)

Phylliida committed 24 days ago

Verified 09c7c50e

HIP: fix RDNA3 FP16/BF16 matrix multiplication (#17817)

JohannesGaessler committed 24 days ago

Verified f334b794

webui: Stop generation from chat sidebar (#17806)

allozaur committed 24 days ago

Verified a28e3c75

webui: Fix context available value in Multi-model Router mode (#17804)

allozaur committed 24 days ago

Verified e31b5c55

webui: Per-conversation system message with UI displaying, edition & branching (#17275)

allozaur committed 24 days ago

Verified 21f24f27

ggml : improve error handling for search path existence checks (#17653)

flyinskyin2013 committed 24 days ago

Verified 7b43f557

Older