ggerganov/llama.cpp

Pull Requests Commits

CUDA: skip graph compat check when graph is being reused

am17an committed 28 minutes ago

f53582ed

CUDA: move graph and concurrent stream state out of common.cuh (#32001)

am17an committed 1 hour ago

Verified 17701e48

ui: use a live-preview markdown editor for the chat form input (#31955)

allozaur committed 1 hour ago

Verified ac007054

model : stop renaming the recurrent cache tensors in plamo2 (#31984)

sanmai committed 1 hour ago

Verified 940a30ee

common/json-schema : fix allOf with plain types (e.g. integer) generating an empty object grammar (#31834)

chinmaykamjith001 committed 2 hours ago

Verified 45c8ea83

common : escape control characters in tool argument strings (#31937)

vvsotnikov committed 2 hours ago

Verified c90aa996

cann: fail early when checking if buffer type is supported (#31270)

rauletorresc committed 2 hours ago

Verified a2c942dc

sycl: Add device-side MoE scheduler (#30156)

cwriter committed 2 hours ago

Verified 18c5a95e

vulkan: add tiled-transpose concat shader for Qwen models (#30591)

0cc4m committed 2 hours ago

Verified ae4893a0

ggml-rpc : fail gracefully on unreachable endpoints and failed handshakes (#30415)

figkcoci-debug committed 2 hours ago

Verified 6122b566

vulkan: Divide GROUP_NORM workgroup to 2D to avoid maxComputeWorkGroupCount limit (#31930)

rillomas committed 2 hours ago

Verified 3837bf0b

vulkan: Fix out of bounds reads/writes on MoltenVK (#31890)

EAddario committed 2 hours ago

Verified c1a4febe

vulkan: cap scalar flash attention at 8 rows for large heads on Qualcomm (#30829)

wanghqc committed 3 hours ago

Verified 78ffbaa1

llama : apply the iGPU lazy-mode fallback only when mmap is off (#31957)

oscarmherrera committed 3 hours ago

Verified 58e38bdf

ci: update oneAPI 2026.2, oneDNN 2026.1 and Level Zero SDK 1.34.0 (#31540)

Asahi-Prv committed 3 hours ago

Verified d5bbcd1f

sycl: un-gate oneDNN fused SDPA for Alchemist at head_dim=64 (#31574)

Gotya-OFFCIAL committed 4 hours ago

Verified 8766c027

chat : don't use lazy grammar when response_format is set with tools (#31956)

rmz-24 committed 9 hours ago

Verified 3b4d6ab3

hexagon: fix 64-bit FastRPC handle truncation on close (#30952)

dmnobunaga committed 9 hours ago

Verified cf6463ac

server: responses api: support conversion of text.format.json_schema to completions (#26145)

boondocklabs committed 10 hours ago

Verified 711f9f4c

server : use common_chat_input in realtime api (#31978)

aldehir committed 11 hours ago

Verified 26d52b0f

server: add oai-compat /v1/audio/speech endpoint (#31763)

ngxson committed 11 hours ago

Verified 08f945ff

mtmd: support gemma 4 input audio stream (#31965)

ngxson committed 11 hours ago

Verified 20c2e966

mtmd: add mtmd_helper_resampler (#31964)

ngxson committed 12 hours ago

Verified 3d94ffe4

ui : retry opening the sidebar when a demo imports its traces (#31952)

allozaur committed 14 hours ago

Verified 0a94a3f2

ui: make the app feel native on phones (#31949)

allozaur committed 14 hours ago

Verified 610f7ab4

feat: external PLE file loading (#31185)

pwilkin committed 15 hours ago

Verified 5f490d7b

Windows unbuffered model load (#26014)

JTischbein committed 15 hours ago

Verified 9d340ec3

server : functionality test for new spec dec (#31027)

nv-vankit committed 16 hours ago

Verified 359dc585

ggml-zdnn: accept non-zdnn tensors, invalidate compute tensors after every op (#31824)

taronaeo committed 20 hours ago

Verified a4d4f2fd

server : fix response_format json_schema when schema is not nested in the json_schema wrapper (#28697)

linsen458-spec committed 20 hours ago

Verified 1945793e

Older