ngxson/llama.cpp

Pull Requests Commits

vulkan: Implement topk_moe fused shader, ported from CUDA (#16641)

jeffbolznv committed 157 days ago

Verified e56abd20

CUDA: use registers instead of smem in topk-moe (#16647)

am17an committed 157 days ago

Verified 38355c6c

opencl: transposed gemm/gemv moe kernel with mxfp4,f32 (#16602)

shawngu-quic committed 157 days ago

Verified 81387858

llama-model: fix insonsistent ctxs <-> bufs order (#16581)

JohannesGaessler committed 157 days ago

Verified 66b0dbcb

rpc : report actual free memory (#16616)

rgerganov committed 158 days ago

Verified 41386cf3

vulkan: Add State Space Model (SSM) Operations Support (#16463)

giuseppe committed 158 days ago

Verified 3d4e86bb

ggml : fix SpaceMit IME array out-of-bounds in task assignment (#16629)

muggle-stack committed 158 days ago

Verified 342c728d

webui: reorganize settings layout (#16607)

ServeurpersoCom committed 158 days ago

Verified ababae7e

vulkan: fix debug build (add_rms_len/data not found) (#16624)

jeffbolznv committed 158 days ago

Verified b1949159

metal : add `CONV_TRANSPOSE_2D` (#16542)

iliailmer committed 158 days ago

Verified 9ad4f193

grammar : use int64_t to avoid int overflows in int schema to grammar conversion logic (#16626)

ochafik committed 158 days ago

Verified 79967ec5

SYCL SET operator optimized for F32 tensors (#16350)

GittyBurstein committed 158 days ago

Verified ceff6bb2

mtmd : support home-cooked Mistral Small Omni (#14928)

ngxson committed 158 days ago

Verified 1bb4f433

fix: added a normalization step for MathJax-style \[\] and \(\) delimiters (#16599)

ServeurpersoCom committed 159 days ago

Verified 683fa6ba

sycl : add ARANGE operator (#16362)

GittyBurstein committed 159 days ago

Verified b22572e9

CANN: format code using .clang-format (#15863)

noemotiovon committed 159 days ago

Verified 7a50cf38

common : Update the docs on -t --threads (#16236)

takasurazeem committed 159 days ago

Verified 6f5d9246

ggml-cpu: replace putenv with setenv for const-correctness (#16573)

otegami committed 159 days ago

Verified adc9b60f

SYCL: Add GGML_OP_MEAN operator support (#16009)

yael-works committed 159 days ago

Verified ee50ee1e

gguf-py : add support for endian conversion of BF16 data (#16594)

AlekseiNikiforovIBM committed 159 days ago

Verified 7adc79c0

cpu : add FLOOR, CEIL, ROUND and TRUNC unary operators (#16083)

safranowith committed 159 days ago

Verified 466c1911

opencl: add q8_0 mm support (#16469)

lhez committed 159 days ago

Verified 0cb7a068

opencl: fix FA for f32 (#16584)

lhez committed 159 days ago

Verified d93f8439

Add server-driven parameter defaults and syncing (#16515)

allozaur committed 160 days ago

Verified f9fb33f2

metal: optimise `GGML_OP_SUM` (#16559)

cern1710 committed 160 days ago

Verified f4ce81c4

server : fix img token logs (#16595)

ggerganov committed 160 days ago

Verified 17304cbc

llama-quant: add support for mmproj (#16592)

ngxson committed 160 days ago

Verified 3e3cb19f

CUDA: Changing the CUDA scheduling strategy to spin (#16585)

JTischbein committed 160 days ago

Verified 5acd4554

server : fix mtmd checkpoints (#16591)

ggerganov committed 160 days ago

Verified 554fd578

metal : avoid using Metal's gpuAddress property (#16576)

ggerganov committed 160 days ago

Verified fa882fd2

Newer Older