ngxson/llama.cpp

Pull Requests Commits

ngxson committed 44 days ago

1fa084e7

ngxson committed 44 days ago

be101fc1

ngxson committed 44 days ago

8f1c6c02

ngxson committed 44 days ago

e79a1f75

ngxson committed 44 days ago

44bc40fd

jinja : do not pass empty tools and add some none filters (#19176)

CISC committed 45 days ago

Verified b45ef270

HIP: add mmf for CDNA (#18896)

zhang-hui-yulo committed 45 days ago

Verified f3dd7b8e

arg : add -kvu to llama-batched-bench (#19172)

ggerganov committed 45 days ago

Verified eed25bc6

ggml-zendnn : resolve ZenDNN backend cross-module symbol dependency (#19159)

z-vishal committed 45 days ago

Verified b33df266

CUDA: refactor topk-moe to enable more models (GLM 4.7, Nemotron etc.) (#19126)

am17an committed 45 days ago

Verified 3bcc9909

sycl: fix norm kernels: l2_norm, group_norm, rms_norm by remove assert to support more cases (#19154)

arthw committed 45 days ago

Verified d4964a7c

ci : find latest release with asset for winget (#19161)

CISC committed 45 days ago

Verified 50e8962f

Vulkan Flash Attention Coopmat1 Refactor (#19075)

0cc4m committed 45 days ago

Verified f6b533d8

spec : add self‑speculative decoding (no draft model required) + refactor (#18471)

srogmann committed 45 days ago

Verified 72d3b189

convert : yield Mamba2Model/GraniteMoeModel modify_tensors (#19157)

danbev committed 45 days ago

Verified ebf57258

ggml-sycl: remove unused syclcompat header (#19140)

PatKamin committed 45 days ago

Verified 0cd7032c

jinja : undefined should be treated as sequence/iterable (return string/array) by filters/tests (#19147)

CISC committed 45 days ago

Verified 60368e1d

vulkan: handle device dedup on MacOS + Vega II Duo cards (#19058)

okuvshynov committed 46 days ago

Verified 88d23ad5

doc: add build instruction to use Vulkan backend on macos (#19029)

chanbengz committed 46 days ago

Verified 0a95026d

ggml: new backend for Virglrenderer API Remoting acceleration (v2) (#18718)

kpouget committed 46 days ago

Verified b7feacf7

ggml-cpu: arm64: Q4_K scale unroll and vectorization (#19108)

Alcpz committed 46 days ago

Verified 6ad70c5a

cuda : fix "V is K view" check for non-unified KV cache (#19145)

ggerganov committed 46 days ago

Verified 631cbfcc

CUDA: tune GLM 4.7 Flash FA kernel selection logic (DGX Spark) (#19142)

ggerganov committed 46 days ago

Verified 2eee6c86

server : adjust spec tests to generate up to 16 tokens (#19093)

ggerganov committed 46 days ago

Verified b931f81b

llama : disable Direct IO by default (#19109)

ggerganov committed 46 days ago

Verified c5c64f72

sampling : remove sampling branching in output_reserve (#18811)

danbev committed 46 days ago

Verified eef375ce

ggml webgpu: Split shared state (webgpu_context) into global state and per-thread state (#18976)

nikhilJain17 committed 46 days ago

Verified 06961e28

ggml-zendnn : update ZenDNN git tag to main branch (#19133)

z-vishal committed 46 days ago

Verified f2571df8

jinja : implement mixed type object keys (#18955)

CISC committed 46 days ago

Verified 2b4cbd28

docs: Remove duplicated word on CUDA build section (#19136)

davelima committed 46 days ago

Verified 68ac3acb

Older