ggml-org/llama.cpp

Pull Requests Commits

hexagon: add fp16 support for binary ops: add,sub,mul,div (#20139)

YardenTal44 committed 1 hour ago

Verified 2b10b626

models : kda chunk size = 16 (#19827)

ymcki committed 13 hours ago

Verified a0ed91a4

CUDA: Improve performance via less synchronizations between token (#17795)

aendk committed 16 hours ago

Verified 2cd20b72

model : update Qwen3.5 model type detection (#20126)

EZForever committed 16 hours ago

Verified 872646b3

cli : add command and file auto-completion (#19985)

CISC committed 18 hours ago

Verified b5ed0e05

convert : register Qwen 3.5 ForCausalLM for text only (#20119)

CISC committed 18 hours ago

Verified cf232515

webui: Improvements for Models Selector UI (#20066)

allozaur committed 20 hours ago

Verified 5e335ba1

chore : correct typos [no ci] (#20041)

marcelpetrick committed 20 hours ago

Verified 92f7da00

hexagon: Flash Attention optimizations (dma, mpyacc, multi-row) and MatMul updates (#20118)

max-krasnyansky committed 22 hours ago

Verified 7a99dc85

opencl: add `SET`, support i32 for `CPY`, minor refactor for cpy (#20101)

lhez committed 22 hours ago

Verified 69fd3453

hexagon: add llama-completion runner script (#20095)

tboinovski1 committed 1 day ago

Verified 1a29907d

[WebGPU] Fix wait logic for inflight jobs (#20096)

nikhilJain17 committed 1 day ago

Verified 24d2ee05

Add concat op to webgpu. (#20068)

yomaytk committed 1 day ago

Verified 541bf376

tools : add missing clocale include in mtmd-cli [no ci] (#20107)

CISC committed 1 day ago

Verified d969e933

ggml: fix ggml_is_contiguous_n for ne == 1 (#20092)

JohannesGaessler committed 1 day ago

Verified 7f5ee549

ggml : use a simple std::thread in AMX without OpenMP (#20074)

angt committed 1 day ago

Verified 66199c9f

impl : use 6 digits for tensor dims (#20094)

ddh0 committed 1 day ago

Verified c99909dd

Fix locale-dependent float printing in GGUF metadata (#17331)

ssam18 committed 1 day ago

Verified cb8f4fa3

completion : Fix a typo in warning message (#20082)

standby24x7 committed 1 day ago

Verified 54910bd4

docs: Fix intel documentation link (#20040)

mkdgs committed 2 days ago

Verified ecd99d6a

kleidiai : add sme fp16 compute path for q4_0 gemm on aarch64 (#20043)

chaxu01 committed 2 days ago

Verified 137435ff

opencl: add optimized q4_1 mm kernel for adreno (#19840)

shaofeiqi committed 3 days ago

Verified 24350fdf

ggml webgpu: fix workgroup dispatch limit for large batch sizes (#19965)

abhijitramesh committed 3 days ago

Verified 49a7564a

ggml webgpu: Clean up per-thread parameter buffer pool and job submission logic (#19772)

nikhilJain17 committed 3 days ago

Verified 4d828bd1

ggml-webgpu: Support non-contiguous `src0` and overlapping `src0/src1` in binary ops (#19850)

yomaytk committed 3 days ago

Verified 36a7a658

vulkan: tune MMVQ for Intel Windows (#19988)

0cc4m committed 3 days ago

Verified feefb928

scripts : improve get-wikitext-2.sh (#19952)

angt committed 3 days ago

Verified ec88c3ce

ggml-cpu: optimise s390x multiply extend instructions (#20032)

taronaeo committed 3 days ago

Verified 2afcdb97

vulkan: improve partial offloading performance on AMD (#19976)

0cc4m committed 4 days ago

Verified 31914624

cuda: cap grid.y at 65535 in non-contiguous dequantize/convert kernels (#19999)

oobabooga committed 4 days ago

Verified 66d65ec2

Older