ggerganov/llama.cpp

Pull Requests Commits

force f16 or f32 on step3-vl conv weights

CISC committed 3 hours ago

Verified f5059efa

CUDA: also store `node->src->data` ptrs for equality check (#21635)

am17an committed 6 hours ago

Verified d12cc3d1

fix: free ctx_copy in ggml_opt_free to plug per-training-session leak (#21592)

RealOrko committed 7 hours ago

Verified 2dcb7f74

server: respect the ignore eos flag (#21203)

ykhrustalev committed 7 hours ago

Verified 66060008

vocab : remove </s> eog token if gemma4 (#21492)

aldehir committed 8 hours ago

Verified d9a12c82

webui : send both backend_sampling == false/true (#18781)

ggerganov committed 8 hours ago

Verified 4a05e0c5

Propose fix a couple of typos (#21581)

jeis4wpi committed 8 hours ago

Verified e9fd9628

kv-cache : extend cache quantization checks (#21586)

Green-Sky committed 9 hours ago

Verified 3ba12fed

webgpu : Query for adapter support when registering WebGPU backend (#21579)

reeselevine committed 9 hours ago

Verified 54739490

metal: Q1_0 backend (#21528)

khosravipasha committed 9 hours ago

Verified dcdcbad4

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

ggerganov committed 9 hours ago

Verified 5764d7c6

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

danbev committed 10 hours ago

Verified 87f4744a

parser: fix MiniMax handling (#21573)

pwilkin committed 12 hours ago

Verified 85d482e6

tests : remove obsolete .mjs script (#21615)

ggerganov committed 12 hours ago

Verified ae65fbdf

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

allozaur committed 14 hours ago

Verified 3bd9aa1f

chore: Remove legacy files (#21606)

allozaur committed 15 hours ago

Verified ece522f9

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

forforever73 committed 15 hours ago

Verified 09343c01

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

hmblair committed 16 hours ago

Verified 97508acb

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

martin-klacer-arm committed 17 hours ago

Verified 5c4aae66

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

am17an committed 21 hours ago

Verified c5ce4bc2

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

iacopPBK committed 1 day ago

Verified 66c4f9de

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

danbev committed 1 day ago

Verified 93bdc615

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

ggerganov committed 1 day ago

Verified 4eb19514

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

reeselevine committed 1 day ago

Verified 957d717c

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

am17an committed 1 day ago

Verified de1aa6fa

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

taronaeo committed 1 day ago

Verified 69c28f15

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

nhs000 committed 1 day ago

Verified 0d049d6a

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

JohannesGaessler committed 1 day ago

Verified a8ec0df4

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

ggerganov committed 1 day ago

Verified e8f50826

ggml : deprecate GGML_OP_ADD1 (#21363)

ggerganov committed 1 day ago

Verified 22fc7913

Older