ggml-org/llama.cpp

Pull Requests Commits

CUDA: split the PAD_REFLECT_1D grid to stay within the HIP grid limit

ServeurpersoCom committed 1 day ago

ff0e760c

chat : accept </ifm|think> as K2 Horizon closer at medium/low effort (#30335)

awss1i committed 1 day ago

Verified 5676dfbe

server, fix CI: read a small temp file in the read_file tool test [no release] (#32168)

ServeurpersoCom committed 1 day ago

Verified 0a66ac71

server: allow private media URLs in the vision tests [no ci] (#32167)

ServeurpersoCom committed 1 day ago

Verified 892b40b9

server: skip LAN address tests when the host cannot reach it [no ci] (#32166)

ServeurpersoCom committed 1 day ago

Verified e9701014

fix(hexagon): make host build work with clang-cl on Windows ARM64 (#32011)

SaidSultani committed 1 day ago

Verified 92d6c8a5

llama : prune (#30565)

ggerganov committed 1 day ago

Verified 8ca5a800

ggml : prune (#29801)

ggerganov committed 1 day ago

Verified 29dbbad2

hexagon: add Q5_0 and Q5_1 support (#32094)

aparmp-quic committed 1 day ago

Verified 2a4daf0e

server : cancel MCP tool calls on client disconnect or timeout (#32025)

hmblair committed 1 day ago

Verified d8c8cc4d

qwen3.5: merge tensors on convert also (#32102)

ORippler committed 1 day ago

Verified 3ba1ce0e

Title: CUDA: few-row MMQ for BS 2-24 (#31533)

anujj committed 1 day ago

Verified 0fb8f157

ci : build ui on non-slim runner (#32138)

CISC committed 1 day ago

Verified 827466b9

server: fix SSRF for CORS proxy and media URLs (#32115)

ngxson committed 1 day ago

Verified 8954f194

server: explicit opt-in tools via non-loopback addr, check cors host port (#32113)

ngxson committed 1 day ago

Verified a1a68eb7

server: disallow direct access child instance from browser (#32120)

ngxson committed 1 day ago

Verified e182fdad

server: keep the server awake while a request uses its state (#32133)

ServeurpersoCom committed 1 day ago

Verified 09452cde

mtmd: require an F32 position embedding for HunyuanVL (#32131)

ServeurpersoCom committed 1 day ago

Verified cfb9d4b7

mtmd: read the MiniCPM-V projection width from ne[1] (#32132)

ServeurpersoCom committed 1 day ago

Verified 72f4603c

mtmd: require the MLP projector and patch embedding tensors (#32129)

ServeurpersoCom committed 1 day ago

Verified 8c388364

mtmd: require a 2x2 spatial merge for the Qwen-VL projectors (#32130)

ServeurpersoCom committed 1 day ago

Verified 08f49598

ggml : route GATED_DELTA_NET to the SpacemiT RVV kernel (#31897)

gitGurugu committed 1 day ago

Verified a8575a3e

vulkan : Intel FA kernel for prefill. (#29357)

sliu39 committed 1 day ago

Verified a963f659

ggml-cpu: use the 4x4 simd_gemm tile on ARM NEON (#32128)

rjtokenring committed 1 day ago

Verified ab4a77b9

CUDA: Optimize prefil via fuse of w_s scale in epilogue MMQ for nvfp4 checkpoints (#26048)

kmorennv committed 1 day ago

Verified 0d69255c

ggml-backend : prefetch only inputs computed by earlier splits (#32100)

am17an committed 1 day ago

Verified 3bce8bd5

ggml : dispatch SpacemiT Q4_1 32x256 through the HP kernel (#31929)

gitGurugu committed 2 days ago

Verified 81f21f4b

ui: add search_chats tool (#32051)

ngxson committed 2 days ago

Verified b19f1d9e

ui: install llama-brand from a GitHub tarball instead of git (#32112)

allozaur committed 2 days ago

Verified d90cfd4a

server: when --tools-runtime is specified, ignore x-tool-runtime (#32119)

ngxson committed 2 days ago

Verified 60d46a5a

Older