ngxson/llama.cpp

Pull Requests Commits

ngxson committed 129 days ago

3a6ee0cf

Merge branch 'master' into xsn/scale_gemma4_3n_pad

ngxson committed 129 days ago

de546992

model: fix multimodal padding token for gemma3n/gemma4

ngxson committed 129 days ago

4a6366b0

gemma : perform per-layer projections in the first layer (#21612)

ggerganov committed 129 days ago

Verified 5764d7c6

examples : disable cb_eval callback for --save-logits (#21553)

danbev committed 129 days ago

Verified 87f4744a

parser: fix MiniMax handling (#21573)

pwilkin committed 129 days ago

Verified 85d482e6

tests : remove obsolete .mjs script (#21615)

ggerganov committed 129 days ago

Verified ae65fbdf

chore: Update labeler to have separate labels for `server/webui` and `server` changes (#21567)

allozaur committed 129 days ago

Verified 3bd9aa1f

chore: Remove legacy files (#21606)

allozaur committed 129 days ago

Verified ece522f9

model : support step3-vl-10b (#21287)

forforever73 committed 129 days ago

Verified 09343c01

webui: fix syntax highlighting lost after streaming for non-common languages (#21206)

hmblair committed 129 days ago

Verified 97508acb

devops: kleidiai: provide KleidiAI-Enabled ARM Release Artifact (#21259)

martin-klacer-arm committed 129 days ago

Verified 5c4aae66

CUDA: make cuda graphs props check faster (#21472)

am17an committed 130 days ago

Verified c5ce4bc2

ggml-cuda: ds_read_b128 for q4_0 and q4_1 mmq kernels (#21168)

iacopPBK committed 130 days ago

Verified 66c4f9de

gguf-py : fix missing comma after bad merge in tensor-mapping (#21558)

danbev committed 130 days ago

Verified 93bdc615

kv-cache : support attention rotation for heterogeneous iSWA (#21513)

ggerganov committed 130 days ago

Verified 4eb19514

ggml-webgpu: parameterize submission size and add iOS specific limits (#21533)

reeselevine committed 130 days ago

Verified 957d717c

CUDA: check for buffer overlap before fusing (#21566)

am17an committed 130 days ago

Verified de1aa6fa

llama-server: fix model params not propagated (#21509)

taronaeo committed 130 days ago

Verified 69c28f15

unicode : add custom Qwen2 regex handler to fix segfault on long input (#21257)

nhs000 committed 130 days ago

Verified 0d049d6a

llama: remove per-arch tensor name lists (#21531)

JohannesGaessler committed 130 days ago

Verified a8ec0df4

server : fix restore for checkpoints with pos_min == 0 (#21510)

ggerganov committed 130 days ago

Verified e8f50826

ggml : deprecate GGML_OP_ADD1 (#21363)

ggerganov committed 130 days ago

Verified 22fc7913

ggml: Vulkan build, Linux -- output error string for errno on fork failure (#20868) (#20904)

tomoverlund committed 130 days ago

Verified 2a619f6f

vulkan: add FA dequant for q4_1, q5_0, q5_1, iq4_nl (#21029)

mkoker committed 130 days ago

Verified edd4d9bc

webui : store reasoning_content so it is sent back in subsequent requests (#21249)

aldehir committed 130 days ago

Verified 482192f1

ggml-cuda : fix CDNA2 compute capability constant for gfx90a (MI210) (#21519)

aviallon committed 130 days ago

Verified 71a81f6f

fix: Detect streaming state in reasoning content blocks (#21549)

allozaur committed 130 days ago

Verified ecce0087

Fix rtl text rendering (#21382)

Kabir08 committed 130 days ago

Verified d1f82e38

[SYCL] Add Q8_0 reorder optimization (~3x tg speedup on Intel Arc) (#21527)

PMZFX committed 130 days ago

Verified 0988accf

Older